Overclock.net banner
2421 - 2440 of 2443 Posts

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by $ilent View Post

Folks, just a quick heads up.

I am looking at doing an Intel section overclocking competition very soon. This is a good chance for you all to put your nicely overclocked chips to the test!

I have created a discussion thread and would really appreciate your input.

http://www.overclock.net/t/1539659/official-ocn-intel-overclocking-competition-discussion-thread-we-need-your-thoughts/0_100

Thanks!
Interesting, I might be willing to go on my three CPUs (i3-550, X3460 and X5650).
Quote:
Originally Posted by mapesdhs View Post

Hard to decide how to judge such things I expect, eg. which is the greater achievement, a Clarkdale i3 at 4.8GHz, or a
G3258 at 4.8GHz. Much easier to do the latter, and it'll give better performance, but the Clarkdale is the tougher challenge.

Will have a look at the thread later; thanks for the notice & link!!

Ian.
I have had several ideas on how to bench some of these "budget" and not-so-budget CPUs. Too bad real life has stopped me to getting all of the hardware.

This is what i have been itching to test. I might ditch the E5 Xeon, but the rest would be good to really see the improvements Intel has done through the years. The X5698 would be the "best-scenario" case of a dual-core CPU which is not limited at all in base clock, cache, PCI-E lanes nor memory bandwidth.


Quote:
Originally Posted by mapesdhs View Post

In theory it should be able to handle around 4.2 to 4.3 max, but of course that means extra voltage,
more heat, etc. If you want to keep it reasonably power efficient, somewhere around 3.6 to 3.8 is
indeed optimal.

Ian.
I also have partly replaced my i3-550 with an X3460 (the Xeonized i7 860). The performance has been virtually the same, especially now that i have gotten back to a single GPU. If there's an area where the Clarkies really need some improvement, it's on IMC and PCI-E performance. My motherboard (EVGA P55 SLI Micro) even ditches the PCI-E switches and directly routes 8 lanes to each PCI-E x16 slot, so i can even discard the switches in any performance hits i might have had on SLI/Xfire.

However i do miss the crazy clocks of my 550. My X3460 only gets to 4.2 GHz on air and it hits 85c.
 

·
Registered
Joined
·
542 Posts
Starbomba wites:
> The X5698 would be the "best-scenario" case of a dual-core CPU which is not limited at all in base clock,
> cache, PCI-E lanes nor memory bandwidth. ...

That's a very interesting CPU, I had no idea it existed. Strange though it does not exist on Intel's ark site.

> I also have partly replaced my i3-550 with an X3460 (the Xeonized i7 860). The performance has been
> virtually the same, especially now that i have gotten back to a single GPU. ...

That doesn't surprise me, Clarkdale was already enough of a decent speedup from Wolfdale to make the
dual-cores pretty beefy for most tasks. The 860 had a somewhat lower clock as a baseline, which is why
I went with an 870 initially. Indeed, quite a few 860s won't safely stay at 4GHz, whereas 4.2GHz is easy
for most 870s (I have three atm, one at 4.3, two not yet oc'd, but also an 875K not yet oc'd).

> If there's an area where the Clarkies really need some improvement, it's on IMC and PCI-E performance.
> My motherboard (EVGA P55 SLI Micro) even ditches the PCI-E switches and directly routes 8 lanes to
> each PCI-E x16 slot, so i can even discard the switches in any performance hits i might have had on SLI/Xfire.

That depends on the mbd though, and how the board maker has chosen to use the lanes. Asrock's P55
Deluxe is a good example of how it can be done rather well without breaking the bank (very nice board,
only 75 UKP at one point, I bought several), while boards like the ASUS P7P55 WS Supercomputer (I have
two) show what can be done when PLEX-type switches are employed. I've tested one with an 870 and
three GTX 980s, it now holds many of the 3DMark records for any P55 platform.
biggrin.gif
Also worked just fine
with four 580s for CUDA acceleration in apps like AE. For example, see:

http://www.3dmark.com/fs/4099529
http://www.3dmark.com/fs/4099561
http://www.3dmark.com/fs/4099641
http://www.3dmark.com/3dm11/9448492
http://www.3dmark.com/3dm11/9448534
http://www.3dmark.com/3dmv/5224914
http://www.3dmark.com/3dmv/5224919

Naturally, there is a performance hit compared to using the same 980s on a newer board, eg. comparing
to a 5GHz 2700K on an ASUS M4E and a 4.8GHz 3930K on an ASUS R4E:

http://www.3dmark.com/compare/fs/4099529/fs/4024488/fs/3971612

but the speed difference is nothing like as large as many would expect, and in one case (gfx test 2) it's
tiny. The overall scores are skewed heavily by the higher CPU results and Combined scores.

P55 is a lower latency platform, which perhaps helps a bit aswell. It's another reason why I went P55
instead of X58 way back (the other being a P55 build was about a 3rd cheaper than an i7 920 setup).

When originally hunting for P55 options, I just flat out avoided any board which only offered x16/x4 with
its two main slots, no thanks. The P55 Deluxe though allows x8/x8/x4 with two GPUs employed, so the
x4 slot can be used for RAID cards, sound card, wifi, etc. Asrock were also really good at supporting the
board, even sending me a prototype BIOS at one point which had updates to support certain RAID cards.

Have a look, it's easily one of the best mainstream P55 boards made IMO, especially with the 3-slot
spacing for the PCIe slots, allowing good cooling even with hot cards (I ran mine with two 580s for a while
before switching to P67/X79). The 16-phase power made oc'ing much easier than many lesser designs. See:

http://www.asrock.com/mb/Intel/P55%20Deluxe/

And btw, notice how it does have a debug POST LED, yet loads of modern Z97 mainstream boards don't
have one, which is pretty bad IMO. I don't think I'd get any board now which didn't have a debug LED. Here's
a good pic:

http://www.ixbt.com/mainboard/asrock/p55-deluxe/board.jpg

> However i do miss the crazy clocks of my 550. My X3460 only gets to 4.2 GHz on air and it hits 85c.[/quote]

Yes, that's what I found with my 860, it got very hot once over 4GHz. Eventually it died, whereas 870s handle
the same 4.2+ at a usefully lower temp. I don't think I'd want to run a Lynnfield at 85C 24/7.

There is of course the elusive i7 880, and I've seen a few come & go on eBay (three in the last week or so,
part of a CPU/RAM/mbd combo which all went for good prices), but so far not at a low enough price for me
to chance a bid, not when I already have 870s, etc.

In the dual-core stakes though, I think the most potent normal consumer-type CPU is the i5 680. I could buy
one anytime for not really that much (50 to 70 UKP), but I already have an i5 670 and I'm reasonably confident
I'll be able to get it to 5GHz, when I can find the time to mess about with the other P7P55 WS-S board. I'll still
keep an eye out for a cheap 680 though, ya never know. Atm there's a used one on eBay UK for 50 (if I didn't
already have a 670, I'd buy it for sure):

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=181731741857

... another used one for 65 with Make Offer...

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=251925014958

... and a couple of new 680s listed for 70 each:

http://cgi.ebay.co.uk/ws/eBayISAPI.dll?ViewItem&item=361240768640

In the meantime, I have an i3 550 running fine at 4.7 on a P55 Deluxe. Not yet done anything with the 670,
not had the time. It's sitting on an EVGA P55 FTW.

I bought a bunch of boxed new 550s a while back for only 25 each, a few more used; some show good potential
for oc'ing in terms of low vcore requirements aswell as nicely even/low stock load temps. Does vary though,
eg. I had a 540 that showed a huge variance in stock load temps, one core being 10C higher than another,
but this does comply with Intel specs (I checked) so there's no real grounds to moan at Intel, just not a chip
I'll be using for oc fun.

Oh for more time to play... haven't done any P55 oc'ing in many months. :\

Ian.
 

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by mapesdhs View Post

That's a very interesting CPU, I had no idea it existed. Strange though it does not exist on Intel's ark site.
It was only launched for OEM servers built for high-speed trading (it depends more on single-core IPS, hence the high clock) which are not that many. Hell, most of the chips i can find are ES versions, not even retail. It is a rare chip indeed, but that only makes it more awesome.
Quote:
Originally Posted by mapesdhs View Post

That doesn't surprise me, Clarkdale was already enough of a decent speedup from Wolfdale to make the dual-cores pretty beefy for most tasks. The 860 had a somewhat lower clock as a baseline, which is why I went with an 870 initially. Indeed, quite a few 860s won't safely stay at 4GHz, whereas 4.2GHz is easy for most 870s (I have three atm, one at 4.3, two not yet oc'd, but also an 875K not yet oc'd).
Indeed. I'm awed at how Intel could squeeze so much noticeable performance from a simple Tick. Changing from my W3530 @ 4 GHz to my X5650 @ 2.9 GHz was a huge increase in overall performance on my X58 build. I would like to get an 875k, but lack of time (and money) would make me waste that chip, so got to be happy with what i have for now.
Quote:
Originally Posted by mapesdhs View Post

That depends on the mbd though, and how the board maker has chosen to use the lanes. Asrock's P55 Deluxe is a good example of how it can be done rather well without breaking the bank (very nice board, only 75 UKP at one point, I bought several), while boards like the ASUS P7P55 WS Supercomputer (I have two) show what can be done when PLEX-type switches are employed. I've tested one with an 870 and three GTX 980s, it now holds many of the 3DMark records for any P55 platform.
biggrin.gif
Also worked just fine with four 580s for CUDA acceleration in apps like AE.
Well, i did notice a performance increase on the Xeon against the i3, on BOINC. Distributed computing really uses the PCI-E lanes, unlike any kind of benchmark. A card would have anywhere between a 5-20 minute reduction in completion time on a x8 lane, and it was without any CPU load. Hell, the faster i3 550 should have beaten the Xeon on raw clocks and architectural differences as most tasks use up to 1 full thread per GPU, but it wasn't the case. The improvement between my W3530 and X5650 was minimal though, almost negligible, if the project i selected used the GPU almost exclusively.
Quote:
Originally Posted by mapesdhs View Post

P55 is a lower latency platform, which perhaps helps a bit aswell. It's another reason why I went P55 instead of X58 way back (the other being a P55 build was about a 3rd cheaper than an i7 920 setup).

When originally hunting for P55 options, I just flat out avoided any board which only offered x16/x4 with its two main slots, no thanks. The P55 Deluxe though allows x8/x8/x4 with two GPUs employed, so the x4 slot can be used for RAID cards, sound card, wifi, etc. Asrock were also really good at supporting the board, even sending me a prototype BIOS at one point which had updates to support certain RAID cards.

And btw, notice how it does have a debug POST LED, yet loads of modern Z97 mainstream boards don't have one, which is pretty bad IMO. I don't think I'd get any board now which didn't have a debug LED.
True that. Not having to depend on the QPI and/or a third chip (in this case, the IOH) really lowers overall latency, not just in PCI-E lanes, but in the ICH/PCH connections too. I also went to P55 for my build back then due to pricing. That i3 550 was just too tempting of a price compared to an i7 920, and the motherboard (EVGA P55 Micro) was a steal.

I have always avoided mainstream boards with X16/X4 layout. Got too used to SLI/Xfire, despite the issues these systems have.That, and that just screams "budget", and if they cut costs in that way, i bet other (more critical) areas got less beefing up as well (like the VRM). I mean, PCI-E lane switches are a dime a dozen, it can't be too hard to use them, but you can get a cheaper PCB with less layers if you do not use them, lowering the overall motherboard quality

POST LEDs are awesome too. My last four motherboards have had it (Asrock P67 Extreme6, GB Z77X-UP5-TH, RIVBE and EVGA X58 Classy3). and now i cannot live without it. Hell, even my RIIIE had a useful "boot status" LED row below the 24 pin connector which was dead useful in troubleshooting the most common issues.

http://www.pureoverclock.com/wp-content/uploads/images/review/motherboards/asus_r3e/asus_r3e_13.jpg
Quote:
Originally Posted by mapesdhs View Post

Yes, that's what I found with my 860, it got very hot once over 4GHz. Eventually it died, whereas 870s handle the same 4.2+ at a usefully lower temp. I don't think I'd want to run a Lynnfield at 85C 24/7.

In the meantime, I have an i3 550 running fine at 4.7 on a P55 Deluxe. Not yet done anything with the 670, not had the time. It's sitting on an EVGA P55 FTW.
My X3460 is at 4 GHz too. Can get to 65-70c, which is more tolerable. It's a quite big contrast to my W3530, where it got to 4 GHz and under 60c, and 4.6 GHz hitting 80c. Maybe i just got lucky though, or simply the bigger area of the s1366 chip plays on its favor.

My i3 550
Quote:
Originally Posted by mapesdhs View Post

Oh for more time to play... haven't done any P55 oc'ing in many months. :\
Same here. IMO, X58 and P55 are the ultimate OC platforms. No kiddy wheels, no artificial caps, and a LOT of variety.
 

·
Registered
Joined
·
542 Posts
Quote:
Originally Posted by Starbomba View Post

It was only launched for OEM servers built for high-speed trading (it depends more on single-core IPS, hence the high clock) which are not that many. ...
Ah, I see! Oddly enough, I know someone who used to write CUDA code for trading acceleration, a most intriguing area, typically Tesla (ECC essential, etc.)

Quote:
Originally Posted by Starbomba View Post

Hell, most of the chips i can find are ES versions, not even retail. It is a rare chip indeed, but that only makes it more awesome.
Oh, that explains why it's not on Intel's site.

Quote:
Originally Posted by Starbomba View Post

Indeed. I'm awed at how Intel could squeeze so much noticeable performance from a simple Tick. Changing from my W3530 @ 4 GHz to my X5650 @ 2.9 GHz was a huge increase in overall performance on my X58 build. ...
Small thought, is it possible that beyond just having more cores, perhaps the X5650 has better Turbo levels which mean that when it is using 4 cores, more of them are at a higher clock compared to when the W3530 is using all 4 cores? Rather annoying that Intel's Ark pages do not list the Turbo levels.

Quote:
Originally Posted by Starbomba View Post

I would like to get an 875k, but lack of time (and money) would make me waste that chip, so got to be happy with what i have for now.
Yes, they do tend to sell for quite high amounts.

However, having read so many reviews & posts, what I say to people about the 875K is that it's best not to think of the chip as being a route to a higher clock than one could get with an 870, since review tests of the 875K typically show it has more or less the same thermal ceiling anyway, ie. what an 870 can do pretty much already reveals the natural limit of Lynnfield.

Rather, the main plus of the 875K is the customisable Turbo levels one can set in the BIOS, something which is not possible with the 870.

Thus, for example, suppose one has an 870 and an 875K which can both run with all 4 cores active (eg. wPrime, etc.) at a max of 4.3GHz with the same voltage. That's it for the 870, one can't really tweak it any more. But with with 875K one can then configure different multiplier settings for using 3, 2 and 1 cores, since this means higher thermal headroom with fewer cores active, perhaps allowing 4.4 to 4.5 with 3 cores, 4.6+ with 2 cores (not sure yet, haven't had time to try and replicate what I've read on forums), and then maybe 4.6 to 4.8 with 1 core (who knows). This is what I'm looking to explore; it was certainly a pleasant surprise to see the extra options appear in the BIOS when I swapped out an 870 for an 875K I won on eBay. I don't think many 875K reviews mentioned it.

Heck, maybe I'm erring too much on the side of caution here; I mean, in theory an 875K with 2 cores active should behave like an i3 550 or i5 670, in which case perhaps it's possible to have a 2-core Turbo level for an 875K as high as 4.7 to 4.9+, with 1-core at 5+ (depends how the presence of the rest of the chip die affects what has to be kept active). Argh, just need the time to play...

Quote:
Originally Posted by Starbomba View Post

... Distributed computing really uses the PCI-E lanes, unlike any kind of benchmark. ...
This is why Tesla cards sometimes have such a huge advantage over gamer cards for professional CUDA. I've helped many users gain access to low cost CUDA power by exploiting (for example) GTX 580 cards (I have many, just bought another ten 3GB models), but the problem with gamer cards is they all only have a single-speed PCIe return path, ie. data going back from the GPU to the CPU is comparatively slow, so in some cases the tradeoff is worth investigating to ascertain whether investing in a Quadro/Tesla is worthwhile (just might be, depends on the app/task), ie. Quadro/Tesla cards have a full speed return path.

Tesla cards also have a better caching structure, ECC and other features lacking in gamer cards. People often assume it's just a difference in drivers, but there are key hw differences which mean PCIe bandwidth, in both directions, can make a big difference for some applications. I've certainly noticed some AE CUDA tests not scaling as well as one might expect with 3 or 4 gamer GPUs, but atm I don't know whether this is a general PCIe issue, a return path issue, or just inefficiency in how AE works (I don't have enough of the right kind of Quadros to compare, only one K5000 atm). Testing with Arion might be more revealing as it's known to scale better, even with multiple GPUs of different types in the same system.

I keep meaning to test this in more detail using a heavy After Effects render test, something a friend of mine constructed which takes half an hour even on a Titan for just one frame. Not had time to sort it out yet though, still writing up the page.

Quote:
Originally Posted by Starbomba View Post

True that. Not having to depend on the QPI and/or a third chip (in this case, the IOH) really lowers overall latency, not just in PCI-E lanes, but in the ICH/PCH connections too.
I remember when P55 was released, there were a lot of X58/920 owners acting very defensively on forums because results showed the 870 repeatedly beating the 920, including SLI/CF tests. Back then, PCIe 2.0 x8 was plenty to feed current GPUs, so a x8/x8 P55 performed very well, and then of course there are PLEX models like the ASUS WS which offer x16/x16 for two GPUs, or x8/x8/x8/x9 for four (though SLI/CF is typically limited to 3-way on such boards). X58 offers more than 2-way natively of course, but not that many people were buying multiple 480s, etc. at the time.

I couldn't get why they were so upset; P55 was obviously a lower latency platform, given the 870 had a higher base clock and higher Turbo levels, without the more complex RAM structure of Nehalem... at the expense of less expansion potential, and most P55 boards didn't have SATA3 either (though as it turned out, most SATA3 ports on X58 boards use Marvell controllers, which are utterly awful). Hmm, maybe the potency of Clarkdale annoyed them more, beats me.

Quote:
Originally Posted by Starbomba View Post

... I also went to P55 for my build back then due to pricing. That i3 550 was just too tempting of a price compared to an i7 920, and the motherboard (EVGA P55 Micro) was a steal.
Indeed, some of the P55 models were very well priced. I think the P55 Deluxe was a tad on the higher side of P55 options when it launched (must confess one reason why I bought it was the SATA3 PCIe option card; alas it's a Marvell-based card and performs really badly), but over time it came way down, to the point where 75 for such a board was just a no-brainer. I even won a new one on eBay for 45 a little later.
biggrin.gif
(today it would likely sell for 4X that much) Anyway, a friend of mine had an ASUS P6X58D Premium with an i7 930, and we both had a GTX 460 at the time; my P55/870 setup was consistently quicker, for significantly less platform cost.

Of course, X58 owners had the option of a 6-core upgrade, but I expect many never bothered doing that given the advent of SandyBridge which jumped over Gulftown once oc'd, at far less power, cost, etc.

Quote:
Originally Posted by Starbomba View Post

I have always avoided mainstream boards with X16/X4 layout. ...
I wish review articles would start criticising this a lot more, it's lazy/cheap design as you say.

Quote:
Originally Posted by Starbomba View Post

POST LEDs are awesome too. ...
Yup! The only thing which really annoys me about many boards is the way the maker often puts the LED, power/reset buttons & FP connector along the bottom edge of the board, where a 3rd or 4th GPU would completely block them. Grrr! For all that I really like the ASUS P9X79 WS (used five of them so far), the LED being on the bottom is crazy, given the board is targeted at users who'd very likely use all the slots, eg. my AE monster has four 580s, but I had to bend over the FP pins, cut away plastic bits surrounding USB/etc. ports, and of course the power/reset buttons can't be accessed anymore (have to rely on the case buttons).

ASUS clearly learned from this, given the later E version of the board has the power/reset buttons moved to the top of the board, but they left the LED near the bottom! (duh), though at least it's been moved to the edge where it's more likely to be visible with a 4th GPU fitted. I just wish the positioning of boards would be more like the ASUS M4E, where the power/reset buttons and LED are all at the top, hence no combination of PCIe slot usage can block them. Indeed, I'd like to get one of the E boards at some point as it has an improved PCIe provision (4-way all at x16).

Re i3s, it's a shame there are no easily overclockable options for S1155 like there are for P55, re the shift from bclk to multiplier based oc'ing brought in by SB. That's one of the really nice things about S1156, the ability to take a low cost i3 (or indeed i5) and really ramp it up to impressive levels, eg. an i3 550 @ 4.7 matches the stock speed of an i5 760 for threaded tasks, while its single-thread speed is through the roof (actually matches a stock speed 2700K!). I'm certain I can get a 550 to 5GHz on one of the boards I have, just not tried it yet (the P55 Deluxe is a good board, but I doubt it's as capable as the ASUS WS or the EVGA FTW).

Maybe this is why the value of used P55 boards has gone up so much in the last few years, ie. people who enjoy the whole challenge of oc'ing for its own sake find it boring these days with current tech, it's just too easy, whereas messing about with an old P55 board, or indeed X58, is a proper challenge. Have a look at recent completed listings on eBay for P55 boards, some of the winning bids are amazingly high. Likewise, certain X58 boards go for really high amounts, inparticular the Gigabyte X58A-UD3R (perhaps because of its strong XEON support, not sure).

I certainly enjoy the process of oc'ing, but doing it with SB is just no challenge whatsoever (ditto SB-E). This week was typical; I received a used 2700K), took one of my M4E/Z boards out of storage (bought boxed/new for ~80), fitted the chip, stuck on a TRUE that only cost 17 on eBay, attached a single NDS fan, GSkill 8GB/2133 kit (40 from eBay), quick test, all up & running ok @ stock, then dialed in the same settings used on my main system (takes only a couple of minutes, most settings are fine with Auto), and voila, 5GHz once again, with almost no noise and decent temps/voltage; so, from an oc'ing 'experience' point of view, a total yawn. But getting the most out of the i5 670 and the 875K, that's going to be a proper challenge.

Atm I have a pile of i3 550s, but not any extra boards to try them.
biggrin.gif
Just have to keep waiting for a lucky P55 win.

Quote:
Originally Posted by Starbomba View Post

... My last four motherboards have had it (Asrock P67 Extreme6, GB Z77X-UP5-TH, RIVBE and EVGA X58 Classy3). and now i cannot live without it. ...
I couldn't help smiling when I read P67 Extreme6; for obscure reasons after having P55 as my main system for a while, I ended up getting an ASUS Maximus IV Extreme way back for the ridiculously cheap sum of 87 UKP when it was still very current (it's what I'm using to type this, still going strong today with a 5GHz 2700K). Later though, I really wanted to have a decent X58 benching system; I bought an Asrock X58 Extreme6 because of its excellent slot spacing, a board so good that for a while it became pretty much the board used by tech sites for new GPU releases.

Alas I haven't had time to do much with it yet, and atm it's fitted with a XEON X5570 (I have a Dell T7500 with two of these, they run quite well, normally at 3.2). However, waiting on a shelf is a 990X which is still in its packet. I want to get the X5570 properly oc'd first, then maybe try an i7 950 if I can find one cheap, finally trying the 990X. Anyway, key thing about the X58 Extreme6 is it's the only mbd I have of any type which has 3-slot spacing for 3-way CF/SLI, and is thus the one board I have which should allow for some mad oc's on the MSI 580 LXs I've obtained (2-slot spacing really limits what one can do with GTX 580 oc'ing).

Never enough hours in the day!

Quote:
Originally Posted by Starbomba View Post

Hell, even my RIIIE had a useful "boot status" LED row below the 24 pin connector which was dead useful in troubleshooting the most common issues.
Years ago a lot of this sort of thing was treated as extras found only on top-end boards, but really by now these features should be standard on all boards IMO. There are a number of what ought to be quite potent boards, using various chipsets, which are badly held back in their practical usefulness by the lack of a debug LED & suchlike. For example, the Gigabyte GA-X79-UD3 has been shown to permit very decent oc's (5.6GHz with a 3930K) and it has 4-way SLI/CF support, but it has no LED or mbd power/reset switches, so definitely not a board I'd want to use for high-end oc/GPU fiddling, even though it's known the base hw can handle it ok.

Quote:
Well whaddya know, that layout is virtually identical to the M4E.
biggrin.gif


Quote:
Originally Posted by Starbomba View Post

My X3460 is at 4 GHz too. Can get to 65-70c, which is more tolerable. ...
Hmm, interesting... I assume the X3460 is basically the XEON version of the i7 860, but I notice from Intel's specs something rather interesting, namely that the X3460 officially supports up to 32GB RAM; have you tried this on any P55 board? Maybe it has to be a server board with more slots? I also see it officially supports a 4-way PCIe split of 4 lanes each, whereas the 860 just offers the usual 1x16 or 2x8. I wonder which if any boards have expoited this XEON's 4x4 provision (no idea)...

Quote:
Originally Posted by Starbomba View Post

It's a quite big contrast to my W3530, where it got to 4 GHz and under 60c, and 4.6 GHz hitting 80c. Maybe i just got lucky though, or simply the bigger area of the s1366 chip plays on its favor.
Hard to say... I do wonder though whether some other factor means certain XEONs run hotter than their similar specs to an i7 might allow one to infer, eg. so far I've not been able to get an X5570 anywhere near 4GHz stable, whereas a 950 can handle 4GHz+ with ease (doubly strange given the XEON has a lower TDP and higer TCase). Or maybe I have a poor X5570 sample, though I was trying with 24GB RAM, so a tad heavy on the RAM setup.

Quote:
Originally Posted by Starbomba View Post

Same here. IMO, X58 and P55 are the ultimate OC platforms. No kiddy wheels, no artificial caps, and a LOT of variety.
Referring to my earlier post, that's why I reckon it could be hard to do an oc contest in a manner that's not just based on performance, since speed-wise modern CPUs & chipsets will win easily. I'm way more impressed by someone managing to get an i3 550 running at 5+ vs. any SB/IB/HW/etc. running at a high clock. For newer CPUs, especially HW, it's become entirely a battle of heat, whereas with P55/X58 it's much more of a balancing issue with the bclk impacting on various parts of the system, dealing with bclk holes, especially re X58's other subsystems that must be carefully managed. For a contest, I'd suggest initially grouping it by chipset category, best P55, best X58, etc.

How ironic then that with Intel's chosen cap filler resulting in lower absolute clocks (and thus less of an overall speed boost than people had hoped would be possible), oc'ing in general today seems to have become rather boring, made worse by the top-end expensive 5960X having such a stupidly low base clock. HW/HW-E can handle 'good' clocks, but it needs water at a minimum to do it, even with a mild oc.

The only newer CPU I'm currently interested in testing is a 4820K, because its full 40-lane PCIe provision means in theory it should be better than a 5820K for 3/4-way SLI/CF, something I'm eager to try out because I thoroughly dislike the lane-cripping of the 5820K. Atm I have my 4820K on an ASUS R4E, but (again) no time atm to meddle with it.

Hoping to do some decent oc meddling in July.

Ian.

PS. Here's my i7 870 on the ASUS P7P55 WS Supercomputer (suspect I just happen to have a good sample), fastest SLI P55 around I reckon.
biggrin.gif
(though it's down to two 980s now, one has moved to its permanent home, my main gaming PC)
 

·
Registered
Joined
·
603 Posts
Hi guys, my kids have an AIO HP Touchscreen that runs an i3-2100. The fans have been on quite a bit more than when we first got it a long while ago. But now, it is lagging and the fans are screaming. I looked at Hardware Monitor and it is showing idle temps of 61c. I don't remember Sandy Bridge being this hot at idle. Does this look normal to you guys? What are idle temps supposed to be?
 

·
Registered
Joined
·
603 Posts
Yes, it is possible, they're basically built like a laptop, anyone have temps for their Sandy or 2100?
 

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by SynchronicBoost View Post

Hi guys, my kids have an AIO HP Touchscreen that runs an i3-2100. The fans have been on quite a bit more than when we first got it a long while ago. But now, it is lagging and the fans are screaming. I looked at Hardware Monitor and it is showing idle temps of 61c. I don't remember Sandy Bridge being this hot at idle. Does this look normal to you guys? What are idle temps supposed to be?
That does not look normal at all. Hell, anything over 50c on idle is not normal on any CPU i've had, be it laptop or desktop.

You should look around for a disassembly guide, it might take a lot of screws but should be doable. Then, a simple cleaning with canned air/compressor and new TIM will do wonders.

I never had a Sandy i3, but both my i3 550 and i3 3220 never went over 40c on idle with air cooling. Hell, even 35c was pushing it when on the desktop.
 

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by mapesdhs View Post

PS. Here's my i7 870 on the ASUS P7P55 WS Supercomputer (suspect I just happen to have a good sample), fastest SLI P55 around I reckon.
biggrin.gif
(though it's down to two 980s now, one has moved to its permanent home, my main gaming PC)
While i might never have tri SLI (my P55 Micro can only hold 2 cards at once), my own X3460 is no pushover. BCLK-wise, i might be able to push it higher than 4 GHz:

http://valid.canardpc.com/eflmvt

I am kinda cheating with the RAM though, it's Registered ECC, which gives waaay more stability than regular modules when OCed. The only downside is that it needs a TON of voltage.

Lately i have modified that computer to be a low noise, low power PC. CPU undervolts very well to 1.05v (in BIOS, real voltage is like 0.970-0.980 with a voltimeter) with HT On and Turbo Off, and the GTX 650 i have has been downclocked in the idle speeds, undervolted in idle and load speeds, and locked the fan to have 0-50% usage. I have even made a custom fan profile in PrecisionX to not have it spin unless it gets over 55c, and it almost never does, unless i'm playing, or watching really high quality videos (with madvr), but regular usage and even Youtube do not tax it that much.

http://valid.canardpc.com/e2b6u5
 

·
Registered
Joined
·
82 Posts
WOW hold crap these chips are amazing! Really can't tell the diff between this & my 4670k rig in general (when i3 540 is highly overclocked) . Even in gaming it keeps up with every game that I play without even cracking a sweat
thumb.gif
Also run super cool. This is my 24/7 under an old TRUE with silent fans.

4.64Ghz
thumb.gif
biggrin.gif

 

·
Registered
Joined
·
542 Posts
Ah the venerable TRUE. 8) I have loads of those now, get them off eBay when I can. Always annoys me when a review of a new cooler doesn't include a TRUE in their resuts as a baseline reference (are modern coolers really that much better?).

Yup, the old Clarkdales are pretty good when oc'd. I have quite a few, plus various Lynnfield CPUs, SB, SB-E, etc. Have to say I can tell the difference, but then I'm playing a couple of games which do benefit from having more than 2 physical cores (eg. Elite Dangerous), as do numerous newer games. P55 as a platform always was a lot more potent than most people gave it credit for. I even tested a P55 with two and three 980s, it runs surprisingly well, eg. for 3DMark11 it's typically only a couple of percent slower in the Graphics tests than the same cards used on a much faster platform (4.8GHz 3930K), except for Test 1 which does depend on CPU power. See:

http://www.3dmark.com/compare/3dm11/9296340/3dm11/9381010

This was comparing to an oc'd i5 760 on the P55. The Physics/Combined results are quite different of course, but anything GPU -bound is amazingly close. Here's the 3-way 980 result, currently holds a few P55 3DMark records:

http://www.3dmark.com/3dm11/9448492

Your vcore freaks me out though, don't think I've ever gone as high as 1.5.
biggrin.gif
I do have a bunch of unused/new i3 550s I was hoping to push to the max some day, and also an i5 670 which I've not yet oc'd (that may just give the highest P55 oc of all, as it has a good 3.46GHz starting point), but atm they sit unused because it's become so difficult to find used P55 boards for a good price. Anyone else noticed that old P55 boards have become very highly valued? They go on eBay for way more than a used Z68 or other kind of board. I wonder if overclocking fans who are bored with the newer CPUs now regard P55 as perhaps a challenge they never took up when it was new. Even on a 'reasonable' P55 like my Asrock P55 Deluxe, it's pretty easy to get an i3 550 to 4.7 or so. I'm sure 5+ is possible with a decent board, never gotten round to trying yet (I do have two better boards, but not the time to experiment atm).

Newer dual-cores like the unlocked G3258 do have an efficiency edge of course, ie. better performance per clock, but Clarkdale's HT makes up for that somewhat in threaded tasks; see:

http://www.sgidepot.co.uk/misc/tests-jj.txt

The Blender/BMW1M result is odd, the i3 550 is quicker at stock. It's the oc'd i3 results which are fun though, but of course it does lose out for threaded stuff. i3 isn't the best for rendering or video encoding, but at least one has the option of an i5 760, i7 870, etc., which are not too bad even today once oc'd.

Ian.
 

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by C-BuZz View Post

WOW hold crap these chips are amazing! Really can't tell the diff between this & my 4670k rig in general (when i3 540 is highly overclocked) . Even in gaming it keeps up with every game that I play without even cracking a sweat
thumb.gif
Also run super cool. This is my 24/7 under an old TRUE with silent fans.

4.64Ghz
thumb.gif
biggrin.gif
Yeah, less cores/cache, even OCed, will amount to less power and less heat produced. Plus the 1st gen i3's have 4 MB of L3 cache which is an advantage as well against other dualies. AFAIK, only until Haswell the i3's started again having 4 MB L3 cache, before they had 3 MB.

That is awesome btw. I also got 4.6 GHz stable back in the day. Might dust off my i3-550 and do a more updated validation and do some benches with my 780 Classy instead of my SLI's GTX 470's
tongue.gif


http://valid.canardpc.com/2382710
Quote:
Originally Posted by mapesdhs View Post

Ah the venerable TRUE. 8) I have loads of those now, get them off eBay when I can. Always annoys me when a review of a new cooler doesn't include a TRUE in their resuts as a baseline reference (are modern coolers really that much better?).
I kinda want to test a TRUE against my Ultima 90. The Ultima is the best 92mm tower cooler i've ever had, bar none. Tried Noctua, Artic, Coolermaster and none can beat the power of the Ultima with the same fans.Too bad other coolers have been lent to me, i'd make one hell of a review. To bad they discontinued that series waaaaaay back (last supported socket without extra add-ons was 775, not even 1366 or 1156 in the image xD).

Good thing Thermalright also made the VX BTK II. An Ultima 90 working on socket 2011 CPUs? Yespls.
yessir.gif


I might try and look for a TRUE Black, or even better, a TRUE Copper in the future. But i still want to try a TRUE with my GTs, or even better, my 38mm Panaflos.

Quote:
Originally Posted by mapesdhs View Post

P55 as a platform always was a lot more potent than most people gave it credit for.
It still is a powerful platform to game on. Same can be said about X58.

Quote:
Originally Posted by mapesdhs View Post

Your vcore freaks me out though, don't think I've ever gone as high as 1.5.
biggrin.gif
I did once go like 1.525v, i cannot get my ambient temps low enough to OC higher without dumping more volts
biggrin.gif


I was underwater on a custom loop though, that helped a bit. Even for a dualie, 1.5v is a bit too much. I got scared and just backed down to 4.6 GHz after validating.
Quote:
Originally Posted by mapesdhs View Post

it's become so difficult to find used P55 boards for a good price. Anyone else noticed that old P55 boards have become very highly valued? They go on eBay for way more than a used Z68 or other kind of board.
That, and once they are on sale on any forum, they get snatched faster than you can say "Paypal". That is the main reason i refuse to get rid of mine. Plus, as a mATX, it is even rarer.
Quote:
Originally Posted by mapesdhs View Post

I do have two better boards, but not the time to experiment atm.
Same here, wish i had the time to really tweak my platforms. I only have one motherboard, but for a mATX, it's been even more stable and rock-solid than most other, full ATX boards.
 

·
Registered
Joined
·
542 Posts
Clarkdale lives!! 8) You still out there, Starbomba?

A while ago I managed to bag an i5 680 on the cheap, though I'd leave it for a rainy (yes, it's raining), then last week someone put an ASUS P7P55 WS Supercomputer board on ebay for 70 + 5 shipping, with make offer. I put in an offer of 60... then I changed my mind and thought sod that, 70 is already very good (P55 boards are still selling for high amounts these days), so I used the BIN anyway.
biggrin.gif
This means I now have three of these WS boards, woot!

The i5 680 is up and running, and good heavens does this board love high bclcks! I'd never really tried it before, but it's whooshed up to 220 no problem. Don't think I'll bother going any higher, no need, but instead I'll leave it at 210 and start reducing VTT and PLL, see how low they can go and still keep 210 stable (at this stage all the other multis are low, just doing bclk isolation atm).

Anyone still using their Clarkdale setups? My main daily rig is still a 5GHz 2700K, though I am in the process of upgrading an X79 setup in preparation for a GPU setup (currently has a 3930K, but want to change it to 4930K to get PCIe 3.0), gonna fit a 960 Pro to hold game data (tested with a 4820K on a different R4E, gives 3GB/sec).

Quote:
Originally Posted by Starbomba View Post

Yeah, less cores/cache, even OCed, will amount to less power and less heat produced. Plus the 1st gen i3's have 4 MB of L3 cache which is an advantage as well against other dualies. AFAIK, only until Haswell the i3's started again having 4 MB L3 cache, before they had 3 MB.
Yesterday I was researching exactly what Intel did with the 2-core Clarkdale i5s, looks like the main difference visible to the user is the presence of Turbo on the i5s, the other feature differences being things like VPro, though I noticed the i5s have AES whereas the i3s don't. See:

http://ark.intel.com/compare/48504,48505

Quote:
Originally Posted by Starbomba View Post

I kinda want to test a TRUE against my Ultima 90. ...
Did you ever get hold of a TRUE? I have a bunch spare btw, including some new; don't really use them now except for testing, as I've switched to AIOs (I bought a mountain of H80s and H110s when Scan was doing them cheap a la proper ASUS refurbs).

Quote:
Originally Posted by Starbomba View Post

Good thing Thermalright also made the VX BTK II. An Ultima 90 working on socket 2011 CPUs? Yespls.
yessir.gif
biggrin.gif


I collected a few VenomousX coolers, they work very well (though for some reason on certain mbds they don't end up with a firm mounting pressure, just seem kinda loose, no idea why). They're ideal for testing X79 setups, and I also got a couple of Ryzen compatibility kits direct from Thermalright so that I can use them when I eventually start meddling with Ryzen (not had time yet). If one isn't going AIO, a VX with 2x NDS PWM is a very potent cooling solution.

Quote:
Originally Posted by Starbomba View Post

I might try and look for a TRUE Black, or even better, a TRUE Copper in the future. ...
I have several TRUE Blacks (one of them is on a P7P55 WS Super with an 875K, which I used to bag all the P55 3DMark records using three 980s), can't say I noticed any real difference. The Copper might be better, but I suspect it'd just be on a par with a VX anyway.

Quote:
Originally Posted by Starbomba View Post

It still is a powerful platform to game on. Same can be said about X58.
It's showing its age in some areas, but there's still some scope for keeping up that tech sites often ignore, such as M.2 devices on addin cards, or using models with their own boot ROMs as C-drives (to avoid the BIOS support issue for normal M.2 NVMe products).

Shocking thought, I still haven't gotten round to testing a 990X I obtained, which was ages ago. Bad me! I was also sent an i7 975, though for "typical user" benchmarking I'd rather get hold of an i7 950 (sorta middle of the pack).

Quote:
Originally Posted by Starbomba View Post

I did once go like 1.525v, i cannot get my ambient temps low enough to OC higher without dumping more volts
biggrin.gif
Blimey, I've never gone anywhere near that.
biggrin.gif


Quote:
Originally Posted by Starbomba View Post

Same here, wish i had the time to really tweak my platforms. ...
If only one didn't have to sleep...

Ian.
 

·
Watercooled Railgun
Joined
·
4,638 Posts
Quote:
Originally Posted by mapesdhs View Post

Clarkdale lives!! 8) You still out there, Starbomba?
Nehalem/Clarkdale Master Race, still alive and kicking
biggrin.gif

Quote:
Originally Posted by mapesdhs View Post

A while ago I managed to bag an i5 680 on the cheap, though I'd leave it for a rainy (yes, it's raining), then last week someone put an ASUS P7P55 WS Supercomputer board on ebay for 70 + 5 shipping, with make offer. I put in an offer of 60... then I changed my mind and thought sod that, 70 is already very good (P55 boards are still selling for high amounts these days), so I used the BIN anyway.
biggrin.gif
This means I now have three of these WS boards, woot!
Now that's one nice deal. I still have my P55 Micro (that mobo also loves high BCLKs, but needs some love to break the 210 MHz mark, since there's a HUGE bclock hole between 208 and 215 MHz), my i3-550 and a Xeon X3460, which is what my mobo currently has covering the socket. That setup has been moved mostly to storage, as i use more my X79 and X58 systems due to my RAM needs, but i get it out from time to time just to test it and see if it still works. It does, to my pleasant surprise. Sadly, i have nowhere else to put it, but it remains unsold as it is the first relatively high end platform i ever owned.
Quote:
Originally Posted by mapesdhs View Post

Anyone still using their Clarkdale setups? My main daily rig is still a 5GHz 2700K, though I am in the process of upgrading an X79 setup in preparation for a GPU setup (currently has a 3930K, but want to change it to 4930K to get PCIe 3.0), gonna fit a 960 Pro to hold game data (tested with a 4820K on a different R4E, gives 3GB/sec).
My daily rig is my X79 setup running an E5-2670, though i will replace it with a Ryzen 7 for better IPC and single core performance. So far i haven't been able to get another Nano nor a second 290X with waterblock, so there went my plans of Crossfiring those beasts. My Nano is one mighty card though, it can single-handedly run all games i play @ 4K with very nice frames, and the fact that my monitor also has Freesync greatly helps, so it's not like i'm in a hurry to get good performance.
Quote:
Originally Posted by mapesdhs View Post

Yesterday I was researching exactly what Intel did with the 2-core Clarkdale i5s, looks like the main difference visible to the user is the presence of Turbo on the i5s, the other feature differences being things like VPro, though I noticed the i5s have AES whereas the i3s don't.
Those are basically the main differences. Intel being Intel and stuff, disabling very minor stuff to make you pay more.
Quote:
Originally Posted by mapesdhs View Post

Did you ever get hold of a TRUE? I have a bunch spare btw, including some new; don't really use them now except for testing, as I've switched to AIOs (I bought a mountain of H80s and H110s when Scan was doing them cheap a la proper ASUS refurbs).
biggrin.gif
I've mainly kept myself to custom WC, but i did get a TRUE Copper, currently it is sitting on my X58 rig. Someday I need to test that beast with my P55.


Quote:
Originally Posted by mapesdhs View Post

I have several TRUE Blacks (one of them is on a P7P55 WS Super with an 875K, which I used to bag all the P55 3DMark records using three 980s), can't say I noticed any real difference. The Copper might be better, but I suspect it'd just be on a par with a VX anyway.
It would be good to test that theory, with the same fans. The TRUE Copper responds VERY positively to high pressure, high CFM fans.
Quote:
Originally Posted by mapesdhs View Post

It's showing its age in some areas, but there's still some scope for keeping up that tech sites often ignore, such as M.2 devices on addin cards, or using models with their own boot ROMs as C-drives (to avoid the BIOS support issue for normal M.2 NVMe products).
True that, it might not have that much single threaded performance nowadays, however for multithreaded performance is still up there with the newest architectures, almost 7 years after.
Quote:
Originally Posted by mapesdhs View Post

Shocking thought, I still haven't gotten round to testing a 990X I obtained, which was ages ago. Bad me! I was also sent an i7 975, though for "typical user" benchmarking I'd rather get hold of an i7 950 (sorta middle of the pack).
Dayum, i'd like to get a 990X myself, just for the lulz. I have a W3530 (i7 930 equivalent) which is still kicking somewhere.
Quote:
Originally Posted by mapesdhs View Post

If only one didn't have to sleep...
Or work...
rolleyes.gif
 

·
Registered
Joined
·
542 Posts
Quote:
Originally Posted by Starbomba View Post

Nehalem/Clarkdale Master Race, still alive and kicking
biggrin.gif
Yay!! 8)

Quote:
Originally Posted by Starbomba View Post

Now that's one nice deal. ...
Yup! Ironically I bagged an Asrock P55 Deluxe just a month later (item 112517277720), though it's not arrived yet. Also a very good board (have one with an i3 550 @ 4.7), though not quite as good as the ASUS WS. Mind you, the Deluxe was selling for only 75 UKP at one point, which was awesome. Mine with an i7 870 easily outperformed a friend's ASUS X58 with an i7 920, which was kinda funny (we both had GTX 460 SLI so it made direct comparisons easy). The lower latency and higher Turbo bins of P55 really made a difference.

Quote:
Originally Posted by Starbomba View Post

I still have my P55 Micro ...
What brand is that?

Quote:
Originally Posted by Starbomba View Post

since there's a HUGE bclock hole between 208 and 215
I've been lucky, so far I've not encountered any such holes.

Quote:
Originally Posted by Starbomba View Post

Sadly, i have nowhere else to put it, but it remains unsold as it is the first relatively high end platform i ever owned.
Alas I wrecked my initial Deluxe board when an 870 oc went wrong (CPU died, took the mbd with it). Talking to Intel, their view was it was most likely something odd with the CPU, which they replaced no prob. The mbd was out of warranty, but as I say a replacement by then was only 75, so it's up & running again but I only use it for benchmarking.

Btw, this low price of the Deluxe is why it really gets on my nerves that debug LED, on-board power and reset buttons are not totally standard by now. Asrock proved it doesn't cost much to include them. I was amazed to see so many Z97 and newer chipset boards without them.

Quote:
Originally Posted by Starbomba View Post

My daily rig is my X79 setup running an E5-2670, ...
That's funny, my X79 2nd gaming PC (GPC2) has the next mbd down, the standard R4E. When I move up to a 4K monitor (probably next year), this will become the primary machine, with a 960 Pro for game data and either a 1080 Ti or I might wait for Volta.

I went with a 3930K though because it was fairly cheap and oc's so easily. At 4.8 is has approx. 4% better IPC than the 1800X (ref), and the threaded performance isn't that much less (an 1800X is about 30% quicker for multithreaded loads, though of course it can be oc'd aswell). This is why my next system will most likely be a Threadripper, as what I have is fine for gaming, but in the future what I'll need is something more beefy for video encoding. Not buying anything yet though, no time to use it, maybe next year.

Quote:
Originally Posted by Starbomba View Post

though i will replace it with a Ryzen 7 for better IPC and single core performance. ...
I goofed up in that regard. Late last year I built an X99 system for a guy in Norway (R5E, 6850K); the cost of the mbd/RAM/CPU was crowd funded for his YT channel, and I was providing the case, PSU, etc. for free. But alas, when asking his local Customs, he was told that anything crowd funded must be purchased locally (which he did eventually do), so I was stuck with the system. It wasn't long before Ryzen came out, so selling the parts was pointless, I would never have recouped the cost; the mbd cost 414 UKP, but recently a new/boxed R5E sold for just 178 (cm_newtech has them listed without backplate for only 145). The CPU was 557, best price I could find at the time, but by heck did Ryzen kick that SKU in the goolies.
biggrin.gif
Thus, instead of a 4960X and ASUS P9X79-E WS setup I'd already built for editing, it's now an X99 with a 950 Pro, etc., so the 4960X is just to sell off, no point in keeping it.

I could of course use the X99 for gaming, but I just don't need something that powerful, I'm not playing modern games. Most common thing I play is Elite Dangerous.

Quote:
Originally Posted by Starbomba View Post

So far i haven't been able to get another Nano nor a second 290X with waterblock, so there went my plans of Crossfiring those beasts. ...
Is CF still worth bothering with? SLI isn't. NV has been deliberately playing it down, and modern games don't scale much now as devs put less and less effort into supporting CF/SLI. Still great if you're playing older games though.

Hmm, I wonder if the Etherium mining craze has meant used Nanos just get gobbled up.

Quote:
Originally Posted by Starbomba View Post

the fact that my monitor also has Freesync greatly helps, ...
Cunning plan. 8)

Quote:
Originally Posted by Starbomba View Post

Those are basically the main differences. Intel being Intel and stuff, disabling very minor stuff to make you pay more.
Quite funny how TR has kicked Intel sideways with the whole product stack segmentation thing they normally do. 60 PCIe lanes, ECC, etc., with any TR CPU is a very smart move on AMD's part.

Quote:
Originally Posted by Starbomba View Post

I've mainly kept myself to custom WC, ...
So far I've only dealt with AIOs (just Corsair). I talked to a German company that builds AIOs and custom kit, one of the suppliers that makes stuff for certain brands, the tech there said it wasn't worth going custom water unless one was willing to spend a decent amount, it wouldn't be enough better than a good AIO like the H110.

One of these days I'll get into custom loops, though it'd need to be when I have some serious spare time, and it's a bit daunting to know where to start, I've noticed most tech sites rarely have any articles about how to begin with custom loops.

Quote:
Originally Posted by Starbomba View Post

but i did get a TRUE Copper, currently it is sitting on my X58 rig. Someday I need to test that beast with my P55.
I'd be surprised if it behaved any better than a VenomousX (which was much cheaper), though I've noticed on some P55 boards the VX does have an odd issue with its mounting firmness, something about the mechanism means it can rotate too easily. Had this very issue with that P55 WS I bought, so I just fitted a TRUE with two NDS PWM instead. No such issues with the VX on other mbds like Z68 or X79.

Quote:
Originally Posted by Starbomba View Post

It would be good to test that theory, with the same fans. The TRUE Copper responds VERY positively to high pressure, high CFM fans.
I do have some insane Scythe and Yate Loon fans for crazy cfm testing, ye gods they move some air, with suitable noise to match.
biggrin.gif


Speaking of testing, lay your bets, will three 780 Tis on my P55 WS beat the three 980s for the various 3DMark tests?...

I think the 980s will be quicker, just about, though the 780 Tis should win out for CUDA tests such as Octane Bench, Arion and Blender/BMW.

Quote:
Originally Posted by Starbomba View Post

True that, it might not have that much single threaded performance nowadays, however for multithreaded performance is still up there with the newest architectures, almost 7 years after.
Actually I think if I can get the 680 running at 5.0, it'll only be about 10% slower than a stock 1800X for single-threaded.
biggrin.gif
Should be a laugh. I'm guessing at a CB10 1T score of around 7016 (stock 1800X gets approx. 7650), though it might be higher as I'm extrapolating from my i3 550 @ 4.7 which I could never get running on the Deluxe with fast RAM at the same time, so who knows.

At least back in the days of bclk oc'ing, one could still have some fun with i3s and i5s; Intel let that oc market slip away. I've often wondered whether this is what is driving the high bidding on ebay for P55 boards, those who like oc'ing for the fun of it but find modern mbds just too easy and boring (Z68 was insanely easy, I can get any 2700K to 5.0 on an ASUS M4E in less than 5 mins, done it 7 times so far), ie. oc'ing a P55 or X58 (especially the latter) is much more of a challenge.

Quote:
Originally Posted by Starbomba View Post

Dayum, i'd like to get a 990X myself, just for the lulz. I have a W3530 (i7 930 equivalent) which is still kicking somewhere.
I know, bad me!! 8\ Just not had the time. Real irony is, it only cost me half what it normally would as I had a friend at Intel who helped me get it. And I do have a great board to try it with too, Asrock X58 Extreme6 (currently fitted with a XEON X5570), the board which a lot of sites used for reviews because its slot spacing is perfect even for 3-way SLI/CF.

Quote:
Originally Posted by Starbomba View Post

Or work...
rolleyes.gif
Blows my mind how much tech tinkering stuff I've never been able to do 'cos one has to pay the bills first. dammit, there's a C64 adventure game I want to finish off! If anyone ever invents a drug that means one does not have to sleep, with zero side effects, the nerds of the world are going to go crazy.
biggrin.gif


Hmm, so how high do you reckon I could push the 680? I'm sure it can reach 5.0, but I've never gone that high before in terms of the bclk levels required (I could use a low bclk and high multi, but that would mean rather slow RAM), certainly not while trying to maintain a good RAM speed at the same time. With the bclks the mbd can handle, and the high multipliers available (max 29 I think), in theory it could run well over 6.0, but something's going to be the limit somewhere. Come to think of it, I've never run any CPU past 5.0 before, would be hillarious if it was a 680.

Ian.
 

·
Watercooled Railgun
Joined
·
4,638 Posts
rolleyes.gif
Quote:
Originally Posted by mapesdhs View Post

The lower latency and higher Turbo bins of P55 really made a difference.
That is true. I had first my i3-550, then i "upgraded" to a W3530. That felt more of a downgrade than an upgrade, until i got my X5650. Then, i made the stupid decision (again) to "upgrade" to a 2500k. That was the last time i used a "mainstream" platform, ever. I've never came down nor had even the thought to go to anything Z based (or worse, H), coming from X58 and X79.

Quote:
Originally Posted by mapesdhs View Post

What brand is that?
EVGA, old school mobo. For a mATX mobo, it is a beast with 6+2 phases (each phase as DrMos instead of regular mosfets), a well designed heatsink, decent layout and a very solid BIOS. RAM OCing isn't bad either, could do almost 2133 MT/s on my Xeon, and even ran 1800 MT/s on my i3 and it's frankly atrocious IMC.

Quote:
Originally Posted by mapesdhs View Post

I've been lucky, so far I've not encountered any such holes.
You lucky. I am almost sure the hole is on the mobo, because both my i3 and Xeon have the same hole, which i greatly doubt it is a coincidence. However, with only 2 CPUs and 1 Mobo to test, i was never able to fully validate my findings.

Quote:
Originally Posted by mapesdhs View Post

Btw, this low price of the Deluxe is why it really gets on my nerves that debug LED, on-board power and reset buttons are not totally standard by now.
You never know you need them until you use them. I usually don't test outside cases, so they are an afterthought for me, until i bench lazily. I do have a spare "board" with buttons i scavenged out of an old blade server, so i use that instead. A minor inconvenience.

Quote:
Originally Posted by mapesdhs View Post

When I move up to a 4K monitor (probably next year), this will become the primary machine, with a 960 Pro for game data and either a 1080 Ti or I might wait for Volta.
Get the 1080ti next year, it'll be waaay cheaper than a Volta, for the performance it gives. If my Nano, which is worse than a 1070, can play everything i play (most demanding games i play are DOOM, Total War Warhammer, UT2016) at 40 FPS or over that, the 1080ti will give you way better performance (and more if you play older games, those i can get like 200 FPS).

Quote:
Originally Posted by mapesdhs View Post

This is why my next system will most likely be a Threadripper, as what I have is fine for gaming, but in the future what I'll need is something more beefy for video encoding.
I will be building Ryzen 7 this year, and Threadripper maybe on December or next year (depending on funds). I NEED COARS because i do distributed computing (BOINC), and while those tasks are not really parallelized, you can have one task per core/thread, thus boosting the point production. I also use AutoCAD, Photoshop and Sketch-up sparingly, but only as my sisters need them (architects whose job and personal computers have like 1/4 of the power mine does).

Also, it's a good thing to wait, early adopters get bitten by all the bugs. It's better to wait while all issues are ironed out then the experience should be flawless.

Quote:
Originally Posted by mapesdhs View Post

I goofed up in that regard.
Now that is some bad luck there. While i'd take those cores off you if i could, the RAM support for X99 is dodgier than Ryzen, even today. Plus, Ryzen made X99 look downright ridiculous. And, the only upgrade path for me on X79 is an E5-2697 v2, which isn't that expensive anymore on the bay. At least you can sell the mobo for an insane price to recoup some of the loss there
biggrin.gif


Quote:
Originally Posted by mapesdhs View Post

Is CF still worth bothering with? SLI isn't. NV has been deliberately playing it down, and modern games don't scale much now as devs put less and less effort into supporting CF/SLI. Still great if you're playing older games though.
Eh, it's a hit-and-miss thing, but i don't really play that much either. I'd use that much GPU power for BOINC mainly. But i used several multi GPU setups from Nvidia and AMD, and never had an issue with them.

Quote:
Originally Posted by mapesdhs View Post

Hmm, I wonder if the Etherium mining craze has meant used Nanos just get gobbled up.
Not really, Fiji in general offers a bad hash rate for the price the cards command and the power they eat, and cannot be tweaked much on the memory side as it is HBM. Mining (from what i understand) needs fast memory and tight timings. HBM is inherently slower than GDDR5, however since it is wider, it can transfer more data. But that slowness is what kills hashrates. Something similar happens to GDDR5X, however its due to a timings issue. As the timings are looser, hashrate suffers.

That is why most miners go for the fastest GDDR5-based cards (RX 570/580 and GTX 1060), as those can be memory tweaked to be faster than the top-end cards (a GTX 1060 has a 25 MH/s rate and a 1080 has a 23MH/s rate, and the 1080 eats more power)

The reason i haven't been able to do that Crossfire is lack of funds, due to Murphy's Law.

Quote:
Originally Posted by mapesdhs View Post

Quite funny how TR has kicked Intel sideways with the whole product stack segmentation thing they normally do. 60 PCIe lanes, ECC, etc., with any TR CPU is a very smart move on AMD's part.
Not only TR, even Ryzen has some features that you don't see unless you get Xeons, like ECC memory. Hell, even the amount of cores for the price is what got Intel aching in the proverbial crown jewels.

Quote:
Originally Posted by mapesdhs View Post

So far I've only dealt with AIOs (just Corsair). I talked to a German company that builds AIOs and custom kit, one of the suppliers that makes stuff for certain brands, the tech there said it wasn't worth going custom water unless one was willing to spend a decent amount, it wouldn't be enough better than a good AIO like the H110.
I did some dabbling with some of the 1st gen Corsair AIOs (both the H50 and H70) and while better than budget aircoolers (like the Corsair A50 which i had at the start), did worse than my full loop. However, it is a small improvement. My full loop improved temps on my i3-550 @ 4.6 GHz by 5-7c over the H70 (thick rad, dual fan), and like 15c over my H50 (slim rad, single fan). Right now those deltas are decreased to like 3-4c with more modern AIOs.

Thee aspect where custom loops shine is multi-device support, and noise levels. You can connect the VRM, GPUs, Chipsets to the loop to make everything cooler, while choosing better fans that make less noise or even reducing on the fan amount than you would be otherwise using.

A normal CPU heatsink or AIO has at least 2 fans, then if you're using 2 GPUs with aftermarket coolers, that's 4-6 fans right there, plus the case fans you need to circulate air so the other fans do their job. On a full loop, you can cut that down to 6-8 fans with better sound profiles, so in the end you are reducing noise while improving cooling, especially on GPUs.

Quote:
Originally Posted by mapesdhs View Post

One of these days I'll get into custom loops, though it'd need to be when I have some serious spare time, and it's a bit daunting to know where to start, I've noticed most tech sites rarely have any articles about how to begin with custom loops.
It can be a little difficult, however EK started to sell some very nice kits, which are not nearly as expensive as what i spent on my loop. It's only when you get fancy (using rigid tubing, getting angled fittings, adding more GPUs, getting QDs) when you see prices ballooning out of control.

Quote:
Originally Posted by mapesdhs View Post

Speaking of testing, lay your bets, will three 780 Tis on my P55 WS beat the three 980s for the various 3DMark tests?

I think the 980s will be quicker, just about, though the 780 Tis should win out for CUDA tests such as Octane Bench, Arion and Blender/BMW.
980's all the way in everything but compute workloads. There are some tasks where the 780Tis and the original Titan still beat Pascal-based Quadros, Titans and the 1080Tis. Hell, some Fermi cards are still very competitive on compute workloads, rivaling the 1080.

After Kepler, i stopped using all Nvidia cards since most of the workloads i do are compute driven, and not pixel pushing. That is why i greatly prefer AMD, since they're like the jack of all trades, but master of none.

Quote:
Originally Posted by mapesdhs View Post

Actually I think if I can get the 680 running at 5.0, it'll only be about 10% slower than a stock 1800X for single-threaded.
biggrin.gif
Should be a laugh. I'm guessing at a CB10 1T score of around 7016 (stock 1800X gets approx. 7650), though it might be higher as I'm extrapolating from my i3 550 @ 4.7 which I could never get running on the Deluxe with fast RAM at the same time, so who knows.
Yeah, Clarkies do have a really weak IMC, and the distance between dies does not help latencies at all. However, 1600-1700 MT/s should be easily attainable for any clarkdale CPU, and a nice performance uplift. I was never able to push anything over 1800 MT/s on my 550 tho, and i never saw any validation or bench made at anything higher than my speeds, so i'd say 2000 MT/s or above is pretty much impossible.

Considering the IMC part is based on the P45 chipset, and Intel's Official Specs for that chipset specified DDR3 up to 1066 MT/s, means the Clarkdale "1333" spec is an overclock already (since die shrinking anything does not improve performance directly). It's surprising it can reach speeds north of 1600 MT/s.

Quote:
Originally Posted by mapesdhs View Post

At least back in the days of bclk oc'ing, one could still have some fun with i3s and i5s; Intel let that oc market slip away.
I agree completely.My 2500k was boringly easy to OC, had a 4.8 GHz in like 10 minutes. Heck, i didn't even stress test beyond 5 minutes, yet i had no BSODs or any errors for months. My old 3930k posed a biiit more of challenge due to the straps, but still, i had a 4.4 GHz OC in like 30 minutes (before i burned it trying to eke a bit more RAM speed). That is the reason i still have my P55 and X58 rigs, some say taking a week to perfect an overclock is too tedious, but for me it's really fun.

Quote:
Originally Posted by mapesdhs View Post

Real irony is, it only cost me half what it normally would as I had a friend at Intel who helped me get it.
Those are the kind of friends i need, friends working at Intel/AMD xD

Quote:
Originally Posted by mapesdhs View Post

Blows my mind how much tech tinkering stuff I've never been able to do 'cos one has to pay the bills first.
Want to buy new hardware to tweak, but have to work to buy said hardware. Then, life happens and you have no money nor time. I know that feel.

Quote:
Originally Posted by mapesdhs View Post

Hmm, so how high do you reckon I could push the 680? I'm sure it can reach 5.0, but I've never gone that high before in terms of the bclk levels required (I could use a low bclk and high multi, but that would mean rather slow RAM), certainly not while trying to maintain a good RAM speed at the same time. With the bclks the mbd can handle, and the high multipliers available (max 29 I think), in theory it could run well over 6.0, but something's going to be the limit somewhere. Come to think of it, I've never run any CPU past 5.0 before, would be hillarious if it was a 680.
Considering the high 27x multi of that chip, i do not think that would be your limiting factor. 210-215 BCLK seems like a nice, stable high ceiling, however combined with the multi, theoretically, if you don't want to push bclk too high, you could do 185x27, 193x26 or 200x25 for 5 GHz. However, my 550 whined too much to reach 5 GHz via BCLK (209x24), however that was due to the massive hole i have. I found about it later, but back then i already had my X58, so i made no further effort to test that theory. That is one of the plans i have for the foreseeable future on P55, find the true max OC of that CPU, and maybe get a 680 or 560 to get more multi.
 

·
Registered
Joined
·
3 Posts
Bought this little beast about 2 months ago after reading some reviews
It can maxed out my GTX 960 2GB
Plow trough some games at 1080p with max/mixed settings
BF1
The Witcher 3
Assassin's Creed Syndicate
Titanfall 2
Fallout 4

i3 540 @4.4GHz @1.308v
Biostar TH55B-HD
Thermalright Archon SBE-X2 (DeepCool Z5 Thermal Paste)
6GB (3x2GB) DDR3 @770MHz
Zotac GTX 960 2GB Core @1300MHz/Mem @1953MHz
SuperFlower Atlas 650W 80+ Bronze

 

·
Registered
Joined
·
2,739 Posts
Yup still decent performance for very little money. My first play through of Dark Souls 3 was on an over clocked 540 and a 260x(temp machine for a few months) and played through it fine.
 
2421 - 2440 of 2443 Posts
Top