Overclock.net - An Overclocking Community - Reply to Topic
Thread: Overclock i7-920 or upgrade CPU? Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
04-01-2019 08:53 AM
Blameless If it's a memory issue, it's almost always QPI/VTT (which is the supply voltage for the L3 and IMC) or VDIMM. QPI PLL and IOH are closely related because QPI is how the CPU communicates with the IOH; shouldn't have much to do with memory though and being closely related doesn't imply they need be the same or that they scale the same way...just that a high QPI clock can require adjusting both because of how the IOH is connected. Edit: It's worth mentioning that many PLL voltages don't need to be brute forced as too much voltage can introduce considerable noise.

Anyway, you can stress the IOH with anything that sends a lot of traffic between the CPU/memory and the rest of the system, as the only path from the CPU to SB or any of the PCI-E lanes is through the IOH.

GPU heavy benchmark tests that frequently change scenes/load new assets work well, as does stressing the southbridge. I'd recommend looping Unigine (Valley or Superpossition) or the Final Fantasy IV Stormblood (making sure the frame rate is uncapped in settings) benchmarks while simultaneously running random read benchmarks on some of your drives. If your NIC is attached to the IOH PCI-E lanes, you could run a network test at the same time.
04-01-2019 08:36 AM
OCmember
Quote: Originally Posted by Blameless View Post
Did a quick test at 3.8GHz QPI with 1.2 QPI PLL and 1.1v IOH with my wife's computer this morning (it's an h.265 mkv file, about six minutes long): https://www.sendspace.com/file/etuxkr

Didn't want to fiddle with it too much as I didn't want to risk corrupting anything and having to reload the backups. I never really pushed high BCLKs with that X5670, but I'm sure I could get a bit higher at 1.2/1.1v if I spent some time on it.

I've only got two working X58 boards left, the one in my wife's computer is a Gigabyte X58A-UD5 2.0 and my seed box still has my working EX58-UD3R r1.1 in it. I sold one other EX58-UD3R a long time ago and killed a third. I've also killed an ASRock X58 Extreme, an ASUS P6X58D and P6TD Deluxe.

The ASRock board was a bit of a lemon, but all the other boards could hit ~220 BCLK at the 18/36x QPI multiplier with little or no IOH voltage increases. They all had the ~4GHz QPI hard wall though, as does pretty much every unmodded X58 board, except some of the later OC focused models, like the Gigabyte X58-OCA.

Most of my LGA-1366 CPUs would scale to ~3.8GHz QPI with 1.1v QPI PLL, but much past that I recall there being a lot of variance. Past 3.9GHz most of them started to need a lot of voltage, but I rarely never needed to run such high BCLKs. Since 2011 most of my LGA-1366 parts have been 32nm Westmeres and Gulftowns, so maybe I'm miss remembering how much QPI PLL my i7-920s required...but 1.4v still sounds quite extreme to me. I'll check through some of my old threads here to see what exactly I was doing with my last 920.

Anyway, I know I've never experienced the need to keep QPI PLL and IOH voltage the same, nor have I generally needed to touch IOH voltage very much, unless the bulk of the directly attached PCI-E lanes were loaded. Even then my GTX 480 SLI and R9 290 CFX setups still only needed 1.2v IOH at 3.8GHz QPI.
Seems you might be right on this one. Looking back on what I did with some recent instabilities while including your information it makes it clear that 850MHz cl6.6.6.18 1T wasn't cutting it cause the last thing I ended up doing was raising my timings to 7.7.7.20 1T It's odd because 6.6.6.18 would clear 100% Memtest ICH but occasionally would have trouble starting games but I never pinned the issue on the memory. I was convinced more QPI PLL and IOH Core was helping but never looked back at it when 7.7.7.20 didn't display any issues. I've also seen some articles talking about the relationship between the QPI PLL and IOH Core and how they are best kept in sync, voltage wise.

I've booted into Windows at QPI PLL & IOH Core at 1.1v Is there any test that pressures the NB Subsystem that I can use to double check my stability?
04-01-2019 07:58 AM
Blameless
Quote: Originally Posted by OCmember View Post
Post screen shots to prove that.

EDIT: With the UD7 I'm using 24 PCIe lanes. The 1070, a sound card, and a NIC.

EDIT2: What board are you using, again? IIRC I had issues with this UD7 booting beyond 3.6Ghz before and the QPI PLL and IOH core helped. I can't remember what my EVGA 760 was set to but it was able to reach close to 4GHz but I'm not so sure it was stable.. it's been a while since using that board and I think I sold it off. Bought an 141 E770 Rev 1.0 a while ago and haven't even tested it. That might be a fun board to mess around with.
Did a quick test at 3.8GHz QPI with 1.2 QPI PLL and 1.1v IOH with my wife's computer this morning (it's an h.265 mkv file, about six minutes long): https://www.sendspace.com/file/etuxkr

Didn't want to fiddle with it too much as I didn't want to risk corrupting anything and having to reload the backups. I never really pushed high BCLKs with that X5670, but I'm sure I could get a bit higher at 1.2/1.1v if I spent some time on it.

I've only got two working X58 boards left, the one in my wife's computer is a Gigabyte X58A-UD5 2.0 and my seed box still has my working EX58-UD3R r1.1 in it. I sold one other EX58-UD3R a long time ago and killed a third. I've also killed an ASRock X58 Extreme, an ASUS P6X58D and P6TD Deluxe.

The ASRock board was a bit of a lemon, but all the other boards could hit ~220 BCLK at the 18/36x QPI multiplier with little or no IOH voltage increases. They all had the ~4GHz QPI hard wall though, as does pretty much every unmodded X58 board, except some of the later OC focused models, like the Gigabyte X58-OCA.

Most of my LGA-1366 CPUs would scale to ~3.8GHz QPI with 1.1v QPI PLL, but much past that I recall there being a lot of variance. Past 3.9GHz most of them started to need a lot of voltage, but I rarely never needed to run such high BCLKs. Since 2011 most of my LGA-1366 parts have been 32nm Westmeres and Gulftowns, so maybe I'm miss remembering how much QPI PLL my i7-920s required...but 1.4v still sounds quite extreme to me. I'll check through some of my old threads here to see what exactly I was doing with my last 920.

Anyway, I know I've never experienced the need to keep QPI PLL and IOH voltage the same, nor have I generally needed to touch IOH voltage very much, unless the bulk of the directly attached PCI-E lanes were loaded. Even then my GTX 480 SLI and R9 290 CFX setups still only needed 1.2v IOH at 3.8GHz QPI.
04-01-2019 05:02 AM
OCmember
Quote: Originally Posted by Blameless View Post
I've got half a dozen Nehalem and Westmere parts that will do near 4GHz QPI with under 1.2v QPI PLL and I've never needed more than stock 1.1v IOH core on any X58 system that wasn't running crossfire/SLI.
Post screen shots to prove that.

EDIT: With the UD7 I'm using 24 PCIe lanes. The 1070, a sound card, and a NIC.

EDIT2: What board are you using, again? IIRC I had issues with this UD7 booting beyond 3.6Ghz before and the QPI PLL and IOH core helped. I can't remember what my EVGA 760 was set to but it was able to reach close to 4GHz but I'm not so sure it was stable.. it's been a while since using that board and I think I sold it off. Bought an 141 E770 Rev 1.0 a while ago and haven't even tested it. That might be a fun board to mess around with.
04-01-2019 03:10 AM
Blameless
Quote: Originally Posted by OCmember View Post
The QPI PLL & IOH Core need to be the same volts. For 3740MHz QPI I need about 1.4v on my QPI PLL & IOH Core.
I've got half a dozen Nehalem and Westmere parts that will do near 4GHz QPI with under 1.2v QPI PLL and I've never needed more than stock 1.1v IOH core on any X58 system that wasn't running crossfire/SLI.
03-31-2019 06:17 AM
OCmember If he does any overclocking his biggest voltage concerns are with the Vcore (CPU), VTT (UNCORE), DRAM (RAM), QPI PLL (Quick Path Inter-connect), and IOH Core (North-Bridge). The QPI PLL & IOH Core need to be the same volts. For 3740MHz QPI I need about 1.4v on my QPI PLL & IOH Core. I think that's around the limit for most QPI speeds on any X58 board. I personally wouldn't recommend going any higher with either of those voltage settings. Uncore and DRAM also have a relationship. Uncore can reach up to around 3.4GHz with 1.33v 1.35v is the absolute max the uncore should be pushed for longevity. And usually the bios settings don't match what shows on a multi meter so it's best not to set it at or above 1.35v. With 1.33v Uncore it's safe to push DRAM voltages to 1.83v but with fluctuations it's best to keep those two within .40v, e.g. 1.33v Uncore 1.73v DRAM. 32nm 1366 socket chips can easily do 4Ghz with 1.32v vcore, maybe even less vcore.

My first X58 board was an EVGA 760 A1 bought back when they first came out (2008 or 2009?). Also picked up an Asrock X58 (first gen/iteration) for my main rig. I'm currently on an X58 UD7 for gaming with a GTX 1070 and it's great for the UE3 & UE4 engine, CS:GO runs very well on it too (1080p 144Hz)
03-11-2019 09:07 AM
Cloudforever
Quote: Originally Posted by acquacow View Post
Different CPUs in the same socket can also have different stack heights, so depending on mounting mechanism, you can easily over-tighten the heatsink on the CPU and break it, or not have enough clamping force to maintain good contact with the CPU.

this series is the same stack heights / nothing's changed other than adding the 2 core's in the CPU itself.

That was one thing i definitely made sure on when I swapped my 930 > X5675

as for looking to see if it'll work - if a 980 or 990x will work (providing the most up to date bios) , a xeon will work in it.
03-06-2019 04:39 PM
Firehawk
Quote: Originally Posted by acquacow View Post
Different CPUs in the same socket can also have different stack heights, so depending on mounting mechanism, you can easily over-tighten the heatsink on the CPU and break it, or not have enough clamping force to maintain good contact with the CPU.
This is news to me. I was under the impression the stack height standardized for any given socket. For instance, all AM4 based chips have the same height.

Can you provide an example where this isn't the case? I'm genuinely curious.
03-06-2019 02:27 PM
acquacow Different CPUs in the same socket can also have different stack heights, so depending on mounting mechanism, you can easily over-tighten the heatsink on the CPU and break it, or not have enough clamping force to maintain good contact with the CPU.
03-06-2019 10:18 AM
Firehawk CPU coolers are based on the socket they can be applied to, not the CPU inserted in said socket.

Having said that, different CPUs on the same socket can have drastically different thermal properties, so you do need to take that into account. For instance an i7 9900k versus an i5 8400.

In your case, going from a 930 to something like a x5675 or w3670 you'll gain 2 cores, 4 threads, slight IPC improvements, and possible frequency gains as well. That, and more RAM since you're only running 6GB, will give your 1080 Ti some breathing room. BUT don't buy anything until you're sure it'll work on your motherboard. You don't want to spend money on parts you can't use when you could instead put it to a platform upgrade, to say a Ryzen 2600.
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off