Overclock.net › Forums › Intel › Intel CPUs › The Intel Devil's Canyon Owners Club
New Posts  All Forums:Forum Nav:

The Intel Devil's Canyon Owners Club - Page 2089

post #20881 of 22515
Quote:
Originally Posted by CL3P20 View Post

@ tolis626 - Your not very accurate in your statements advising against 'higher voltage'

*up to 1.25v on air for >8GBs of RAM is quite common even for XMP settings .. and especially when running full bank (all slots populated)
*"stock" VCCSA is independent for each IMC/CPU.. your 'stock' is not others 'stock'
*VCCSA depends greatly on the type of RAM IC.. ie - set Auto using Samsung and compare to Auto running PSC etc.
*RAM voltage below 1.85v for IB and HW is just fine so long as your IC arent getting instability from heat during extended loads.

I run 1.8v daily for 4x2 BSGE (garbage Kingston 1333mhz set) @ 2050mhz. Its stable using HCI memtest so far, >500% coverage & there is no risk in damaging the IC at this speed/latency/voltage


Hmmm... Regarding the VCCSA part, I must admit I didn't know. All Haswell's I've seen started off with 0.82V VCCSA. But now that you mention it, that was always at a bone stock "Load Optimized Defaults" state, so you're probably right. I remember some people with Sandy Bridge-E CPUs having trouble with high VCCSA, but I can't remember the details. Unless I can find something to contradict you, I have to take your word for it. smile.gif

But on the RAM voltage side... I don't see where you disagree with me. I said the same thing, that DDR3 DIMMs can take that kind of voltage no problem from what I know. When I said something about frying the DIMMs, I meant that in the worst case scenario it's just a RAM swap and the problem is fixed. It's not like you'll ruin your mobo or CPU at even higher DRAM voltages. And when I said it's not recommended, I meant that it's not manufacturer recommended. Manufacturers only suggest up to 1.65V from what I know. I guess I should have been clearer on that part, though.

Since you seem to know your stuff when it comes to RAM overclocking, what do you do with VTTDDR (or whatever that voltage that follows exactly half of VDIMM is called) when going over 1.65V on VDIMM? I mean, most guides say leave it do its thing, but only up to a certain point. On my mobo, the number will turn red at over 0.825V, or when running 1.65V VDIMM. Which is fine, as 1.65V is all I need for my memory, but still I'm curious. I also may want to further overclock my memory down the road just for fun (Although I doubt it's going to do much, as anything above 2400MHz doesn't seem to boot anyway), so it'd be nice to know.
post #20882 of 22515

VTTDDR is the most finicky voltage there is. It is from my experience completely independent from set to the next. A different timing combination or speed could mean using a different VTT.. some of which may refuse to even POST. Frankly, its not worth the time tuning unless your running SKL and trying for all out speed. Its never produced anything for me other than mystery.

 

*only RAM I have every damaged is PSC..and with less than 1.77v air. Samsung and Hynix can run over 2v air for life without worry. Its all in what your trying to accomplish. Obviously vitamin 'V' isnt going to cure everything and thats where folks get things twisted I think. Its ok to add voltage as long as results are scaling and temps are under control. Fall outside of those lines and your asking for serious issues regardless of what component your talking about.

post #20883 of 22515
Quote:
Originally Posted by CL3P20 View Post

VTTDDR is the most finicky voltage there is. It is from my experience completely independent from set to the next. A different timing combination or speed could mean using a different VTT.. some of which may refuse to even POST. Frankly, its not worth the time tuning unless your running SKL and trying for all out speed. Its never produced anything for me other than mystery.

*only RAM I have every damaged is PSC..and with less than 1.77v air. Samsung and Hynix can run over 2v air for life without worry. Its all in what your trying to accomplish. Obviously vitamin 'V' isnt going to cure everything and thats where folks get things twisted I think. Its ok to add voltage as long as results are scaling and temps are under control. Fall outside of those lines and your asking for serious issues regardless of what component your talking about.

Glad to see I've been right to avoid messing with it. thumb.gif

To be honest, RAM is the most "safe" thing to mess with for me. The only times I've seen RAM fail aren't from overclocking, but rather aging or because of defects. Plus, it's the easiest and cheapest thing to replace if it breaks. Thing is, most of the time problems with RAM are a pain in the bum to diagnose. Sure, a boot loop or a failure to post is one thing, but random BSODs out of nowhere can be RAM related but RAM is usually the last thing people have in mind. And that's why I never ever suggest overclocking RAM to anyone, unless they're someone who knows what they're doing. Testing for hours upon hours with HCIMemtest, only for it to find an errors 4 hours in, after having made a 400% pass is infuriating. But it's part of the fun for me. Others usually don't find this fun. At all. tongue.gif

Since you're here... One more question. What's your suggestions on what to do with tertiary timings? My mobo has them labeled in a quite confusing way, so I can't know what is what. But I do know they can make a large difference performance-wise.

Also, @everyone. What would you say is the maximum safe VCCIN on air/AIO water? Seeing it go over 2V kinda scares me, but different guides imply that up to 2.2V should be "safe".
post #20884 of 22515
Quote:
Originally Posted by tolis626 View Post

Since you're here... One more question. What's your suggestions on what to do with tertiary timings? My mobo has them labeled in a quite confusing way, so I can't know what is what.

You can know, and you'll need to. I've been at this for weeks and I feel like I've barely scratched the surface of memory timings. It helps to learn the terms each MB maker uses so you can draw from as many sources of information as possible.

IIRC your motherboard is pretty friendly for memory OCing. There's a ROG guide or two out there for your motherboard that covers some of the basics for memory overclocking with it. You might be able to just change a couple options and get 90% of the way there.

Beyond that it gets hairy, and you'll find that the rabbit hole goes as deep as you want. Try leaving memory timings on auto and see what tertiaries it sets for different ram speeds. Look at other people's timings with the same ICs as your RAM at the speed you want.

If you want to manually tune them, get a few different memory benchmarks together, a few RAM stress tests at the ready, and a big notebook. Bigger.
post #20885 of 22515
Quote:
Originally Posted by MIXEDGREENS View Post

You can know, and you'll need to. I've been at this for weeks and I feel like I've barely scratched the surface of memory timings. It helps to learn the terms each MB maker uses so you can draw from as many sources of information as possible.

IIRC your motherboard is pretty friendly for memory OCing. There's a ROG guide or two out there for your motherboard that covers some of the basics for memory overclocking with it. You might be able to just change a couple options and get 90% of the way there.

Beyond that it gets hairy, and you'll find that the rabbit hole goes as deep as you want. Try leaving memory timings on auto and see what tertiaries it sets for different ram speeds. Look at other people's timings with the same ICs as your RAM at the speed you want.

If you want to manually tune them, get a few different memory benchmarks together, a few RAM stress tests at the ready, and a big notebook. Bigger.

By "I can't know", I meant there's some confusing labeling of timings. I don't know why, but Asus HAD TO do it differently. Tertiary timings' names are all over the place. And I think not even all of them are there, ready to be adjusted. Primary and secondary timings are ok, at least, although tweaking secondary timings didn't provide that large of a performance benefit, even during memory benchmarks. I think they just improved the consistency of the RAM's performance, but that could easily be placebo.

Another problem I have is that I don't know what ICs I'm using. That I have a pair of 8GB Corsair Vengeance Pro sticks rated at 2133MHz 11-11-11-27-2T at 1.5V doesn't say much. It's your run of the mill, mid to high end memory module. I couldn't find the ICs used anywhere online, though. The way I see it, the only way to know is to take the DIMMs apart, but I'm kind of scared of taking apart things that weren't meant to be taken apart (That's the reason I haven't delidded yet). This is also the reason I don't risk going over 1.65V on the RAM. Sure, Haswell isn't that sensitive (Or sensitive at all) to high memory voltages like some earlier CPUs were (*cough*socket 1366*cough*), but my system is on the brink of stability and having it thrown off by RAM going bad is infuriating just to think of. Then again, I don't believe my RAM will go bad at, say, 1.75V, but I also don't believe I can achieve any meaningful gains in performance by doing so. 2400MHz 10-12-12-30-1T is more than respectable in my book and having tightened secondary timings also helps. Maybe I could do like 2666MHz, but 8GB DIMMs don't predispose me for high clocks anyway, so I'm happy. In fact, for the longest time, I could only do 2200MHz because I didn't know about tweaking VCCSA. Turns out that my IMC crapped out and not the RAM itself in the beginning. Having it at about 0.95V let me go 2400MHz no problem.

Back to tweaking timings on my mobo. The only redeeming aspect of tertiary timings tweaking on the M7F is that it has a "Latency boundary" option (two of them actually, A and B) that roughly set tertiary timings in one swoop, so you don't have to tweak them all one by one. Thing is, I had it set to 1 and 5 and my timings didn't look that much lower than auto anyway. I also had HCI return a few errors (4 I think) after a 400% pass and 4 hours+ of testing. Thing is, that was at 1.6V (Yes, my RAM works at 2400MHz CL10 and 1.6V), but I have since moved to 1.65V for peace of mind. So I may try lowering them again just 'cause I can. Bad thing is, I don't know exactly what they correspond to. redface.gif
post #20886 of 22515
http://www.xtremesystems.org/forums/showthread.php?285767-DDR3-IC-thread&p=5255150&viewfull=1#post5255150

http://www.xtremesystems.org/forums/showthread.php?285767-DDR3-IC-thread&p=5256015&viewfull=1#post5256015

Looks like Hynix AFR. No clue how those compare to the BFR I've become intimately familiar with, but that's a good springboard. Poke around for some overclocking posts with 4gbit AFR sticks and you'll at least be able to find some secondary timing examples that might shed some light on an incorrect timing causing that instability.

http://pdf1.alldatasheet.com/datasheet-pdf/view/533432/HYNIX/H5TQ4G83AFR-H9C.html

That's the AFR datasheet.

Your CWL for CAS 10 is 7, and CAS 11 needs 8. Double check that setting. Incorrect timing right there will make full stability incredibly difficult to achieve.
post #20887 of 22515
Quote:
Originally Posted by MIXEDGREENS View Post

http://www.xtremesystems.org/forums/showthread.php?285767-DDR3-IC-thread&p=5255150&viewfull=1#post5255150

http://www.xtremesystems.org/forums/showthread.php?285767-DDR3-IC-thread&p=5256015&viewfull=1#post5256015

Looks like Hynix AFR. No clue how those compare to the BFR I've become intimately familiar with, but that's a good springboard. Poke around for some overclocking posts with 4gbit AFR sticks and you'll at least be able to find some secondary timing examples that might shed some light on an incorrect timing causing that instability.

http://pdf1.alldatasheet.com/datasheet-pdf/view/533432/HYNIX/H5TQ4G83AFR-H9C.html

That's the AFR datasheet.

Your CWL for CAS 10 is 7, and CAS 11 needs 8. Double check that setting. Incorrect timing right there will make full stability incredibly difficult to achieve.

At least I got CWL right (or tWCL as my mobo calls it). If I remember correctly I had searched for higher binned kits of Vengeance Pro and copied some timings from there. That's how I ended up with my primaries, at least. I remember getting some erratic behavior with CWL at 8 and then I think it completely failed to boot at 6.

Anyways, thanks for your help man, much appreciated. I will take a closer look at all those later in the evening when I have time. +rep!

PS : I remembered now. My kit is Vengeance Pro v4.29. I had asked about it in the forums a while back and it seems that it uses Micron ICs. Those tend to not overclock that well if I'm not mistaken. I welcome being told I'm wrong, but 2400 CL10 seems to be about as high as it'll go without adding ridiculous amounts of voltage and needing to loosen the timings too much. thumbsdownsmileyanim.gif
Edited by tolis626 - 6/27/16 at 6:26am
post #20888 of 22515

Hynix can take some voltage safely though. If they scale with voltage, you should be safe to run ~1.8v for daily use. I'd try 11-12-12-36 2T and see how high they will scale with voltage. Once you find a limit you can try to tune them in a bit.

 

*VCCIN- depends on CPU. Some just need more. Its not a voltage you want to increase unless its needed. Not all FIVR's are created equal.. and some HW CPU need higher VCCIN to produce the same amount of stable vcore compared to others. I generally just watch VCCIN with a MultiMeter and once both my vcore and VCCIN start to droop under load, I raise VCCIN a little bit until I see only the vcore fluctuating a small amount under load. I feel this technique helps achieve the lowest stable VCCIN for said clocks.

post #20889 of 22515
How do i know who made my cheap-ass kit of ram?
Dorje
(12 items)
 
  
CPUMotherboardGraphicsRAM
4690k Gigabyte z97x-sli MSI TF 280 G Skill Sniper 1600/9 
Hard DriveCoolingOSMonitor
Seagate 4 tb hybrid Corsair h80i Windows 8.1 SE39UY04 
PowerCaseMouseAudio
EVGA NEX-G 750W S340 (white) Bloody ZL5A Razer Blackshark 
  hide details  
Reply
Dorje
(12 items)
 
  
CPUMotherboardGraphicsRAM
4690k Gigabyte z97x-sli MSI TF 280 G Skill Sniper 1600/9 
Hard DriveCoolingOSMonitor
Seagate 4 tb hybrid Corsair h80i Windows 8.1 SE39UY04 
PowerCaseMouseAudio
EVGA NEX-G 750W S340 (white) Bloody ZL5A Razer Blackshark 
  hide details  
Reply
post #20890 of 22515
Quote:
Originally Posted by CL3P20 View Post

Hynix can take some voltage safely though. If they scale with voltage, you should be safe to run ~1.8v for daily use. I'd try 11-12-12-36 2T and see how high they will scale with voltage. Once you find a limit you can try to tune them in a bit.

*VCCIN- depends on CPU. Some just need more. Its not a voltage you want to increase unless its needed. Not all FIVR's are created equal.. and some HW CPU need higher VCCIN to produce the same amount of stable vcore compared to others. I generally just watch VCCIN with a MultiMeter and once both my vcore and VCCIN start to droop under load, I raise VCCIN a little bit until I see only the vcore fluctuating a small amount under load. I feel this technique helps achieve the lowest stable VCCIN for said clocks.
Yeah, Hynix (and I think also Samsung) ICs can take a beating. I've seen people go crazy with them no problem. As long as the IMC on your CPU can keep up, they keep going. I don't know anything about Micron ICs though. They're like the least interesting of the bunch or something. No one seems to talk about them. tongue.gif

For now I'll leave it be. It's plenty fast as it is and I don't think it's worth the hassle to push it farther.

Anyways, I know how VCCIN works and should be tweeked. Among the first things I had to learn as my CPU wouldn't really go far at its stock VCCIN. And sadly it's not among those that work better with low VCCIN. 1.5-1.6V crashes consistently faster than higher voltages. Oh well... I was just curious to see what the max recommended voltage is. For now, I'm using 1.95V as that allows me to run 4.7GHz at 1.275V (Cache at 42x, VRing at 1.2V, VCCSA at 0.96V, VCCIO-D at 1.08V, VCCIO-A at 1.05V, VDIMM at 1.65V). This doesn't seem terribly high, but I've seen people go way higher. Thing is, what's safe for 24/7 use? I mean, some people push over 1.4V on their CPUs. That doesn't mean everyone should, and I wouldn't do it on a daily basis. I think you get my point. smile.gif
Quote:
Originally Posted by jdorje View Post

How do i know who made my cheap-ass kit of ram?

I guess the quickest way would be to search online for your exact kit. I think G.Skill is more consistent with their ICs, unlike Corsair that have 3 or 4 different ICs in their Vengeance Pro line alone. XtremeSystems should be a good place to start. smile.gif

Or I guess you could always take a RAM stick out of your PC, remove the heatsink and just look on the ICs directly.
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Intel CPUs
Overclock.net › Forums › Intel › Intel CPUs › The Intel Devil's Canyon Owners Club