i7-2600K build: Overclocking: GPU Unlock/OC

[darkware]

For the most part, the heat spreaders are unnecessary. There are some situations (fanless liquid cooling boxes, builds which largely obstruct any air flow to the RAM) where they are helpful, but so long as you've got air circulating and aren't over-volting them, then there's no real benefit to them.

That said, I do remember seeing warnings from both Mushkin and Corsair that removing the heat spreaders was a dangerous task, as the ICs can be damaged by the force of pulling the spreaders off. At the same time, some people reported that the spreaders on some RAM were connected only by some weak thermal tape and in some cases, nothing at all.

So... I guess its up to you to see if you want to gamble.

As for the amounts of RAM, I currently have 8GB (and a Core2 Quad). Even running some huge games (Civ V seems to be particularly hungry), I don't fill up physical memory. Still, I'll be using 16GB on the new build because I plan on doing some development with multiple VMs and its nice to be able to give 4GB physical to a VM instance. The other applications for 16GB of RAM seem to be video editing, photo editing, professional graphics/rendering, and high-complexity CAD. Lots of those applications will run faster with lots of RAM. Of course... they'll only see a notable benefit if they are compiled as 64-bit applications. If not, some people with 16GB use large chunks of their physical memory as a RAM disk to further speed things up. However, in the age of SSDs RAM disks aren't as impressive as they used to be. They're faster, just not mind-blowingly so anymore.

 

[tersagun]

no worries, are you thinking of getting two?

I do, yeah.

Though I'm running a fan-less, 12 year old case, with it's top and side wide open. Meaning there isn't any special air ventilation in the case (though I still got average temps for my stock 6950 and overclocked Phenom 720).

In order to crossfire, I have to upgrade my board, as it's only 16x/4x but I'm worried about the heat. Because 2 Twin Razr's on top of each other without heavy cooling might result in 100% GPU fans

 

[DrBoss]

IntelBurnTest @4.4ghz


Not sure on the accuracy of IntelBurnTest for stability compared to 12 hours of Prime95, but it certainly is convenient. 5 tests only took around 25minutes.

Think i'm done playing with the cpu, time to move onto the gpu's

 

[Termie]

This ram talk got me thinking. Does anyone have any experience removing the heat spreader from Corsair Vengeance ram? Considering my case is well ventilated i can't image the heat spreader is doing much for me (the ram is not OC'd).

It is still my interest to have 16gb of ram. The next question is what would actually make use of the added capacity. General consensus is that no more than 8gb of ram is needed for gaming, but would i likely see improvement in CAD and photo/image editing software if i had 16gb?

TechPowerUp looked at the heat spreaders on the Vengeance and found that they could not get them apart: http://www.techpowerup.com/reviews/Corsair/Vengeance_1600_MHz_CL9_12GB/2.html.

IntelBurnTest @4.4ghz


Not sure on the accuracy of IntelBurnTest for stability compared to 12 hours of Prime95, but it certainly is convenient. 5 tests only took around 25minutes.

Think i'm done playing with the cpu, time to move onto the gpu's

Nice...I'm not sure either, but I've never really liked the idea of running my rig at full tilt for 12 hours. You really got your temps nailed now - max of 60, idle around 30. That's really impressive for that overclock.

 

[DrBoss]

That test was performed with an ambient room temperature of 70˚F (21˚C). The Archon is a solid heat-sink, no doubt. Last night I briefly played around with some higher clocks, and one point taking the processor to 5Ghz. It would take a significant amount of tweaking to get stable, but I think this system would do 5Ghz without breaking 70˚C under load… 4.4Ghz is good for now though.


GPU OCing
With the stock bios the Twin Frozr III is limited in both Catalyst Control Center and MSI Afterburner to 900Mhz Core clock and 1325Mhz Memory Clock, at those settings I saw a 3% average performance increase (vs 850Mhz, 1300Mhz) in Crysis… Good, but not good enough. I spent a bit of time looking into unlocking the cards. There is still no concrete consensus about the additional shaders unlocking on these cards (considering they are non-reference). I’ve read a few threads that say high frequencies can be unlocked, but not the shaders… Techpowerup, however; claims the additional (R6970) shaders can be accessed.
http://www.techpowerup.com/reviews/MSI/HD_6950_Twin_Frozr_III/27.html
http://www.techpowerup.com/forums/showthread.php?t=142870
I guess I am just going to have to try.

I’ve never flashed a GPU bios. Aside from what is outlined in various online guides, any advice? Is there a preferred program to use? Russian Sensation posted a bios (shaders unlocked, voltages not changed) in another forum. Does anyone have alternate bios profiles? I’d like to compare multiple profiles using Techpowerups RBE (radeon bios editor) prior to selecting one.

Can't read Chinese, so this could certainly be bogus... but here is another apparent successful unlock
http://blog.danawa.com/prod/communi...76&cate_c3=971&depth=3&prod_c=1326716&sTab=28

 

[Termie]

That test was performed with an ambient room temperature of 70˚F (21˚C). The Archon is a solid heat-sink, no doubt. Last night I briefly played around with some higher clocks, and one point taking the processor to 5Ghz. It would take a significant amount of tweaking to get stable, but I think this system would do 5Ghz without breaking 70˚C under load… 4.4Ghz is good for now though.


GPU OCing
With the stock bios the Twin Frozr III is limited in both Catalyst Control Center and MSI Afterburner to 900Mhz Core clock and 1325Mhz Memory Clock, at those settings I saw a 3% average performance increase (vs 850Mhz, 1300Mhz) in Crysis… Good, but not good enough. I spent a bit of time looking into unlocking the cards. There is still no concrete consensus about the additional shaders unlocking on these cards (considering they are non-reference). I’ve read a few threads that say high frequencies can be unlocked, but not the shaders… Techpowerup, however; claims the additional (R6970) shaders can be accessed.
http://www.techpowerup.com/reviews/MSI/HD_6950_Twin_Frozr_III/27.html
http://www.techpowerup.com/forums/showthread.php?t=142870
I guess I am just going to have to try.

I’ve never flashed a GPU bios. Aside from what is outlined in various online guides, any advice? Is there a preferred program to use? Russian Sensation posted a bios (shaders unlocked, voltages not changed) in another forum. Does anyone have alternate bios profiles? I’d like to compare multiple profiles using Techpowerups RBE (radeon bios editor) prior to selecting one.

Can't read Chinese, so this could certainly be bogus... but here is another apparent successful unlock
http://blog.danawa.com/prod/communi...76&cate_c3=971&depth=3&prod_c=1326716&sTab=28

Unlocking the shaders will give you only about a 3% increase in performance. See our own MrK6's review, particularly post #24: http://forums.anandtech.com/showthread.php?t=2144158.

You will get MUCH more out of overclocking past the CCC limits, particularly considering a BIOS flash might not work. It's as simply as unlocking "unofficial overclocking" in MSI Afterburner (which you obviously have since you have MSI cards). Here's how to do it: http://www.overclock.net/amd-ati/641299-guide-enabling-unofficial-overclocking-msi-afterburner.html. As you'll see from MrK6's review, the 6950 can go well beyond 900Mhz.

And in case you're wondering, my current HD5850 overclock is most definitely not "official"!

P.S. If you really want to know more about HD6950 overclocking, you should post a new thread in the video card forum, or just scan recent threads there.

 

[DrBoss]

You will get MUCH more out of overclocking past the CCC limits, particularly considering a BIOS flash might not work. It's as simply as unlocking "unofficial overclocking" in MSI Afterburner (which you obviously have since you have MSI cards). Here's how to do it: http://www.overclock.net/amd-ati/641299-guide-enabling-unofficial-overclocking-msi-afterburner.html. As you'll see from MrK6's review, the 6950 can go well beyond 900Mhz.

Yea, after doing a bit more digging, it seems the shader unlock isn't worth the hassle (but it would be nice for "badass" factor).

I am going to run with the "unofficial overclocking" you linked above - seems easy and i am pretty sure the R6950 TFiii will go to 950Mhz w/ stock voltage.

 

[DrBoss]

The build has been running flawlessly for months...
Until today.

I came home from work and went to start the system,
it didn't post.

The fans and hard-drive spun up, but the motherboard didn't beep, and no video came through the monitor.

I opened the case and began to diagnose.

The Sabertooth Mobo has a series of LED intended to inform you if any particular system (CPU, HD, Memory, GPU's) isn't working. All lights ran through their cycles as normal (indicating there was no problem), except for the GPU LED. The GPU LED didn't even turn on. The manual says that if any system throws an error the corresponding light will come on, and STAY ON. Usually they just light up, and then turn off (indicating no problem). The fact that the GPU LED didn't even illuminate had me scratching my head.

So i unplugged everything, removed both GPU's to examine them. I saw no defects. I then put GPU #1 in, and attempted to start the system... NO LUCK. So i put GPU #2 in and attempted to start the system... NO LUCK. I was worried my cards were fried.

Before throwing in the towel, i decided to clear the CMOS via the jumper. Once i did this, i started the computer without any GPU installed and sure enough the GPU LED illuminated and stayed on (indicating a problem - no gpu installed). So i installed GPU #1 and the system turned over. I booted to windows... everything looked normal. Shut it down, installed GPU #2, hooked up the crossfire bridge. The system booted. I figured i was out of the woods.

I started watching a video file and ten minutes into it, the system had a hard shutdown. It just went black. I flipped off the power switch on the PSU, then turned it back on, and pressed the power button. The system came up.... and here i am typing this.

I am not sure what the problem is. I assume its GPU related, or related to the PCI-E slots on the Mobo.

Any thoughts?

I've tested crossfire on Deus Ex Human Revolution, both cards are functioning as you would expect. When the system is running, it seems to run fine. But now i am fearful of another hard reset.

Any help/suggestions would be greatly appreciated.

 

[Termie]

Might want to post this in the video cards section, but my guess is that you have one bad video card. Why not try running the system with just one card to troubleshoot?

Did you end up trying to unlock the cards? If so, that could have something to do with it.

I'm just glad it wasn't that SSD that I talked you into.

 

[DrBoss]

No worries, the SSD hasn't faltered.

It seems odd to me that a GPU would fail (prevent the system from booting), and then revert back to normal after the CMOS was reset...

I just ran through another 30 minutes session of Deus Ex with crossfire enabled, both cards operating at 100%... the game ran without flaw.

When i took the system apart earlier tonight, I examined both cards closely and saw no sign of defect or circuit failure (burns on the PCB, etc)

Haven't had another hard reset.

But i will take the advice and post in the video card section tomorrow morning.

Cheers

*EDIT, as for Unlocking/Overclocking the GPU's, i never did it. They are running at the factory OC'd spec

 

[Termie]

No worries, the SSD hasn't faltered.

It seems odd to me that a GPU would fail (prevent the system from booting), and then revert back to normal after the CMOS was reset...

I just ran through another 30 minutes session of Deus Ex with crossfire enabled, both cards operating at 100%... the game ran without flaw.

When i took the system apart earlier tonight, I examined both cards closely and saw no sign of defect or circuit failure (burns on the PCB, etc)

Haven't had another hard reset.

But i will take the advice and post in the video card section tomorrow morning.

Cheers

A lot of things could be wrong that you couldn't see from looking at the PCB. For instance, a damaged memory chip. I agree it's a bit odd that the system booted after a CMOS clear, but not that odd. I still suggest you try running on one card for a while to see if you can replicate it on one card or the other (like running that video that brought it down last time). If neither card produces the problem on its own, it could be a crossfire problem, but that's when you'd have to get help from people with crossfire.

I highly doubt it's the mobo.

 

[DrBoss]

makes sense,
i will isolate the cards and try to replicate the problem

thanks termie

 

[jackstar7]

Just another thing to consider, it could be the PSU going. Since dual GPUs would likely be your biggest draw.

I noticed I had failures start up and ended up needing to RMA and replace my PSU (same model as yours it looks like).

Hope you find the problem and it's not a huge pain.

 

[DrBoss]

is there anyway to test a PSU to see if it is meeting the rated output?

 

[DrBoss]

Its been a while.

Still running the same system in the OP aside from having replaced graphics cards several times (currently MSI GTX 1080), increased ram to 16GB (same RAM as listed in OP, just twice as much), and added several SSD's.

I've gotta say, the i7 2600K has really gone the distance. Over the last 6 years I’ve routinely upgraded GPU’s to improve performance to great effect… it’s only been recently that I’ve noticed the CPU being the bottleneck on some games. A new build is on the horizon, but I’m going to wait until AMD gets Ryzen 2 and Threadripper 2 on the shelf.


But that’s not the reason for this post. I’ve recently started getting hard resets while gaming on the system. No graphical anomalies, no perceived slow-down in performance, just abrupt hard resets (only while gaming). I don’t get a BSOD, the system just shuts off, and immediately reboots itself. The system event viewer only logs a Kernel 41 error “The system has rebooted without cleanly shutting down first”. No other faults.

I’m currently running 10 passes of MemTest86 to see if the RAM is perhaps the culprit.


Any bright ideas as to what could be causing this?


Perhaps the PSU.

Perhaps the CPU needs more voltage for the 4.4 Ghz OC (it’s been undervolted for years and run flawlessly).

Perhaps some kind of driver conflict, but I doubt that, I keep everything up to date, and there have been no new software installs on the system in months.


If the memory test is clean when I get home, I think I’m going to reduce my offset voltage and see if a bit more juice to the CPU irons out the issue.

+++edit+++ I should add, CPU and GPU temps as reported by MSI afterburner are all solid. Nothing is running hot... except perhaps the PSU, which I'm unable to monitor.

 

[Markfw]

My guess is PSU. Your next build may need a bigger one, so get like am RM1000 Corsair, and see if that fixes it. When you do the new build, but the 850 back here with only one video card.

Also, CPU's can degrade over time. Instead of more voltage, take the OC down a couple hundred mhz.

 

[DrBoss]

After the memory test is complete, In an effort continue isolating what might be causing the crash, I assume stress testing the CPU with Prime95 (or similar software) would be a good way to rule out the CPU...

Perhaps I can replicate the crash by running Unigine Heaven to stress both the CPU and GPU...

But yeah, PSU seems likely.

 

[Markfw]

After the memory test is complete, In an effort continue isolating what might be causing the crash, I assume stress testing the CPU with Prime95 (or similar software) would be a good way to rule out the CPU...

Perhaps I can replicate the crash by running Unigine Heaven to stress both the CPU and GPU...

But yeah, PSU seems likely.

The more I think about it, the CPU degrading is more of a possibility than the PSU.

Put the VCORE at stock, and the OC to 4.2, if that does not work, than 4.0.

 

[DrBoss]

I will test this evening and report back.

 

[DrBoss]

Its not the CPU/OC. I've rolled back performance significantly and am still getting crashes. Prime95 was stable for 2 hours, max CPU temp 62c. MemTest86, after 8 passes, found zero errors. I've taken out my audio card, reinstalled majority of drivers, performed a clean boot with all accessory software disabled... keep crashing while gaming.

....

this is the typical log dump after each crash (3 different events,timeline relatively typical):

ERROR (upon reboot)
9:26.18
- System
- Provider
[ Name] Microsoft-Windows-Eventlog
[ Guid] {fc65ddd8-d6ef-4962-83d5-6e5cfe9ce148}
EventID 1101
Version 0
Level 2
Task 101
Opcode 0
Keywords 0x4020000000000000
- TimeCreated
[ SystemTime] 2018-03-08T02:26:18.168820500Z
EventRecordID 160893
Correlation
- Execution
[ ProcessID] 1296
[ ThreadID] 1820
Channel Security
Computer Vicious
Security
- UserData
- AuditEventsDropped
Reason 0

_______________________________________________________________________

CRITICAL (power loss event - undefined)
9:26:09
The system has rebooted without cleanly shutting down first. This error could be caused if the system stopped responding, crashed, or lost power unexpectedly.

- System
- Provider
[ Name] Microsoft-Windows-Kernel-Power
[ Guid] {331C3B3A-2005-44C2-AC5E-77220C37D6B4}
EventID 41
Version 2
Level 1
Task 63
Opcode 0
Keywords 0x8000000000000002
- TimeCreated
[ SystemTime] 2018-03-08T02:26:09.448405200Z
EventRecordID 745343
Correlation
- Execution
[ ProcessID] 4
[ ThreadID] 8
Channel System
Computer Vicious
- Security
[ UserID] S-1-5-18
- EventData
BugcheckCode 0
BugcheckParameter1 0x0
BugcheckParameter2 0x0
BugcheckParameter3 0x0
BugcheckParameter4 0x0
SleepInProgress false
PowerButtonTimestamp 0

____________________________________________________________________

ERROR (wtf - note this happens several minutes in advance of the crash)
9:23:57
Event filter with query "SELECT * FROM __InstanceModificationEvent WITHIN 60 WHERE TargetInstance ISA "Win32_Processor" AND TargetInstance.LoadPercentage > 99" could not be reactivated in namespace "//./root/CIMV2" because of error 0x80041003. Events cannot be delivered through this filter until the problem is corrected.

- System
- Provider
[ Name] Microsoft-Windows-WMI
[ Guid] {1edeee53-0afe-4609-b846-d8c0b2075b1f}
[ EventSourceName] WinMgmt
- EventID 10
[ Qualifiers] 49152
Version 0
Level 2
Task 0
Opcode 0
Keywords 0x80000000000000
- TimeCreated
[ SystemTime] 2018-03-08T02:23:57.000000000Z
EventRecordID 122576
Correlation
- Execution
[ ProcessID] 0
[ ThreadID] 0
Channel Application
Computer Vicious
Security
- EventData
//./root/CIMV2
SELECT * FROM __InstanceModificationEvent WITHIN 60 WHERE TargetInstance ISA "Win32_Processor" AND TargetInstance.LoadPercentage > 99
0x80041003

________________________________________________________________

 

[DrBoss]

I'm back to thinking it's the PSU.

System only drops when both CPU and GPU are being strained. Lately I've been playing FarCry3 and Prey, downsampled from 2880x1600 @ 60fps.... Both run flawlessly until its lights out.

Guess I'll have to pull the Amiga out of the closet and boot up Shadow of the Beast II to pass the time until a new PSU arrives.

 

[DrBoss]

Long story short: It was the PSU.

After taking the CPU back to factory settings, unplugging and re-plugging all connections, testing all memory sticks independently, & sighing a lot... I replaced my Corsair HX850 w/ a Corsair AX860, now i'm back to regular operation with my original CPU OC.