• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

SR-2 Optimization Thread

I've been rocking the following settings stable for the entire week so far. These were borrowed and tweaked from Musky's original post. I am at 3.6 (temps at 53) on a pair of L5640's w/Noctua HD-D14's. RAM is 6GB (2x (3x 1GB)) Kingston HyperX DDR3-1600 (8-8-8-24). I did try blck of 205, but couldn't even get it to boot and didn't have time to muck with it. For now the settings are stable and I am not going to work much harder to eak out a little bit more.

I've crunched a few WUs this week:
Normal A3 SMP: 1:09 for 76k PpD
6040: TPF 3:30 for 70.8k PpD
6701: TPF 3:09 for 44.7K PpD
2685: TPF 13.04 for 130.4k PpD (currently working on)
2692: TPF 13:09 for 129.2k PpD


***
Bios settings
Frequency/Voltage Control - main screen
CPU Frequency - 200
PCIE Frequency - 101
CPU Multiplier - 18
QPI Frequency - 4.800GT
Memory Frequency - DDR-1066
MCH Strap - Auto
CPU Uncore = 18x

CPU Configuration screen
Max CPUID Limit - Disabled
Coherency Support - Disabled
A20M - Disabled
Speedstep - Enabled
Turbo - Enabled
C-State - Disabled
ACPI T State - Disabled
Everything else is Enabled

Memory screen
8-8-8-24 - auto the rest

Voltage
VDroop (both) - Without VDroop
Vcore (all) - 1.350 V
Vtt (all) - 1.350 V
Memory CPU0 - 1.65 V
Memory CPU1 - 1.65 V
IOH - 1.400 V
Turn off NUMA under Power Management

Signal Tweaks screen
IOH QPI 0 = Auto
IOH QPI 1 = Auto
 
Is your uncore actually 18 though? (ie if you pull up CPU-Z or ELEET what is the NB freq listed as? it should be 3600 if your uncore is 18. If it's 3200 you're having the same uncore problems I'm having. :()
 
zero2dash said:
Is your uncore actually 18 though? (ie if you pull up CPU-Z or ELEET what is the NB freq listed as? it should be 3600 if your uncore is 18. If it's 3200 you're having the same uncore problems I'm having. :()
Click to expand...

It IS showing 3200. I am using BIOS A50!!
 
:( I think we're realizing now the L5640's are locked to 16x uncore....they may possibly go lower, but it seems like 16x is the highest they'll go. I tried raising in A47 and A50 both and even though I put uncore at 18x and 20x, it still booted at 16x no matter what else I tried (including raising Vtt). Ah well, I did PM MIBW about it, and as he said....higher uncore is nice but it's not really a deal breaker. Out of the two I'd definitely be happier getting NUMA disabled, I could (pretty much honestly) care less about running a higher uncore. :eek:

I still need to work on disabling NUMA though...worked for a bit on it last night, then it got late so I had to quit. I tried raising Vtt to 1.35 to disable NUMA and I still wouldn't boot (LED goes to 0C then 2A and restarts the boot process again). My bios settings are pretty identical to yours (and musky's :)) with the exception of me adding a few other fine tunes....I wonder if the QPI freq 0 and 1 tweaks are the issue. (Mine are set at -85, -16 as shamino recommended). I see you guys are on auto for both but you've had luck disabling NUMA. :confused:
 
Try leaving the QPI Frequencies on AUTO and see if you can disable NUMA. I was going to do the QPI changes last, but frankly it's rock solid stable and TPF's are pretty damn good (13:03 on a P2685). Temps are rock solid (53 or so on CPU cores and PWM is 60) so I don't see any reason to play around with the QPI frequencies.
 
Ghanakis: Did you ever figure out how to check which BIOS version you are on?

I did do the update to A50, but don't really have a way to checking to make sure that the update took :confused:
 
Isn't it on the first screen when you enter the bios, System Info or whatever it is called? I found it once (that is where I got the A41 version as being the shipped version), but I can't remember where I saw it.
 
musky said:
Isn't it on the first screen when you enter the bios, System Info or whatever it is called? I found it once (that is where I got the A41 version as being the shipped version), but I can't remember where I saw it.
Click to expand...

Yup. Confirmed I am on A50
 
I still cannot disable NUMA and for all intents and purposes I'm throwing in the towel.
I've tried running only 1 gpu, lowering PCIe from 101 to 100, turning pretty much everything on auto (like capreppy's settings), running 1.35 Vcc and Vtt both. I just don't get it. I don't have enough info to troubleshoot this....I can't find any info on what NUMA relies on to run or not run; whether it needs Vcore, Vtt, certain timings, PCIe frequencies, voltage elsewhere....there's not enough info out there on it.
 
After running through a -configonly client setup tonight I realized that my SR-2 is only reporting "10231 MB" of available memory, so after checking my control panel I see that my system is only reporting 10GB of ram.

I recall seeing this issue before but the thread is now 29 pages long and "missing ram" didn't produce results for me. Is there a particular setting that affects this? It was reporting 12GB when I initially set it up, but I've made multiple changes since then. Thanks guys.
 
zero2dash said:
I still cannot disable NUMA and for all intents and purposes I'm throwing in the towel.
I've tried running only 1 gpu, lowering PCIe from 101 to 100, turning pretty much everything on auto (like capreppy's settings), running 1.35 Vcc and Vtt both. I just don't get it. I don't have enough info to troubleshoot this....I can't find any info on what NUMA relies on to run or not run; whether it needs Vcore, Vtt, certain timings, PCIe frequencies, voltage elsewhere....there's not enough info out there on it.
Click to expand...

All the things you mentioned are less likely to work than memory related issues - forgive my foggy brain today, I can't remember if I was talking with you before about this or someone else, but I would try loosening memory timings first - 2T command rate, MCH to 1600 etc.

Aix. said:
After running through a -configonly client setup tonight I realized that my SR-2 is only reporting "10231 MB" of available memory, so after checking my control panel I see that my system is only reporting 10GB of ram.

I recall seeing this issue before but the thread is now 29 pages long and "missing ram" didn't produce results for me. Is there a particular setting that affects this? It was reporting 12GB when I initially set it up, but I've made multiple changes since then. Thanks guys.
Click to expand...

You have to loosen memory timings a tad until it shows up reliably (This is also seen in NUMA disabled situations).. in my case missing memory was intermittent with 12 dimms even before the NUMA thing - but my looser settings were probably why I had no dramas with NUMA disabled.

More info here with probably a lot more specific info than my solution, which was simply to loosen every timing by approx 10%
:
http://www.evga.com/forums/tm.aspx?m=649017
 
I went through the missing memory issue a few days ago. Even after resetting to default BIOS I wasn't able to get it to show up in BIOS/Windows. I ended up removing and testing the ram 1 dimm at a time. When I got to the point where I was ready to test all 6 dimms, it was all showing up.

I switched back to my overclocked BIOS with NUMA disabled. All the dimms were still showing up.
 
Aix. said:
s there a particular setting that affects this?
Click to expand...

Manually adjust your Back-to-Back CAS Latency; doing that solved my disappearing ram issue. Found out about doing that from some NewEgg reviews referencing the Asus x58 boards and their forums.

Put it 1 lower than your timing. IE if your ram is 9-9-9 set it to 8. (It boots at 1 digit higher than what is in the bios.) Mine is 9-9-9-24, I set that to 8, I have not had randomly disappearing ram since then.


may i be worthy said:
All the things you mentioned are less likely to work than memory related issues - forgive my foggy brain today, I can't remember if I was talking with you before about this or someone else, but I would try loosening memory timings first - 2T command rate, MCH to 1600 etc.
Click to expand...

Actually I went even better than that and left all ram settings on auto and tried MCH 1600 as well.
Still no post with NUMA disabled. Same thing...LED reads 0C (which is check for keyboard), then 2A (which is initialize primary devices) and...restart. Does this 3x, shuts off for a second, turns back on and tries again.
Digging up info on the 2A bios code can have something to do with expansion slots, so I tried taking out a gpu (running 1 gpu only), I also tried putting the second gpu in the PCIe 16x #3 slot (normally I have it in 16x #5 to give the cards breathing room). Still didn't work.

After basically throwing the kitchen sink at disabling NUMA (running practically everything on auto and raising voltages a lot higher than they have been stable at so far), hoping one of those settings would fix it....it still hasn't fixed it. :(
 
Last edited:
Running 2684's with a TPF of 16:29. How is that comparing with other SR-2's?

Not exactly problem free yet since I came back an found a blue screen today. On a 6701 of all things! A little more cpu vcore to bring it up to 1.375 we'll see how that works out.

X5680
CPU - Both running 4225mhz (so far)
Memory - 1825mhz, 6-9-6-24-CR1

Seems to have some headroom left on the CPUs. I know the RAM does.

Need more time from people and work to continue the testing iterations.
 
Last edited:
[H]ecklerKoc[H];1036397148 said:
Running 2684's with a TPF of 16:29. How is that comparing with other SR-2's?

Not exactly problem free yet since I came back an found a blue screen today. On a 6701 of all things! A little more cpu vcore to bring it up to 1.375 we'll see how that works out.
Click to expand...

I run around 18 on my fastest one, so you have me beat by quite a bit.

Edit: Mine are L5640s at 3654.
 
Last edited:
[H]ecklerKoc[H];1036397148 said:
Running 2684's with a TPF of 16:29. How is that comparing with other SR-2's?

Not exactly problem free yet since I came back an found a blue screen today. On a 6701 of all things! A little more cpu vcore to bring it up to 1.375 we'll see how that works out.
Click to expand...

Hiya HK, what GHz and mem speed are you at right now?

From HFM: P2684

Name: SR2-1 - X5650 - 4200mhz
Number of Frames Observed: 300

Min. Time / Frame : 00:15:34 - 100,275 PPD
Avg. Time / Frame : 00:16:08 - 95,039 PPD


Name: SR2-2 - X5660 - 4300mhz
Number of Frames Observed: 300

Min. Time / Frame : 00:15:21 - 102,406 PPD
Avg. Time / Frame : 00:15:42 - 99,001 PPD
 
My SR-2 w/L5640's @ 3.6, I'm getting 18:25 TPFs for 78Kk PpD
 
[H]ecklerKoc[H];1036397148 said:
Running 2684's with a TPF of 16:29. How is that comparing with other SR-2's?

Not exactly problem free yet since I came back an found a blue screen today. On a 6701 of all things! A little more cpu vcore to bring it up to 1.375 we'll see how that works out.

X5680
CPU - Both running 4225mhz (so far)
Memory - 1825mhz, 6-9-6-24-CR1

Seems to have some headroom left on the CPUs. I know the RAM does.

Need more time from people and work to continue the testing iterations.
Click to expand...

6701s seem to be very sensitive. I can get 4.1Ghz working with full stability in 6 hours of LinX and BigAdv units, but I occasionally have gotten BSODs from that setting at 186x22. They are very RAM timing sensitive, so you can try loosening them a bit, as they seem pretty aggressive. I'm sure your RAM has more headroom (as you've said) but the IMCs on your chips might need more voltage or looser timings.

That's pretty good in terms of TPFs though. I'm getting 16:55 - 17:05 on my 3.9Ghz x5650s but that is with straight up 9-9-9-24 1T timings with my RAM at 1560mhz and no Back-to-Back CAS delay set up. Funny enough if I reduce unCore to a slower speed like 15x, and set B2B CAS delay to 3 (minimum value) and get the chips to 4.0Ghz (20x200) I actually get slower TPFs.

Aix. said:
After running through a -configonly client setup tonight I realized that my SR-2 is only reporting "10231 MB" of available memory, so after checking my control panel I see that my system is only reporting 10GB of ram.

I recall seeing this issue before but the thread is now 29 pages long and "missing ram" didn't produce results for me. Is there a particular setting that affects this? It was reporting 12GB when I initially set it up, but I've made multiple changes since then. Thanks guys.
Click to expand...

There are three things, and a combination thereof that cause this outside of hardware problems.

1) As MIBW said, RAM timings. Loosen them up a bit and see if the issue goes away. Adding 3 to B2B CAS delay might help without changing the RAS, CAS, and Command Rate timings (1 or 2T).

2) VTT voltage on the chips. You don't want to run past 1.40V VTT to be safe, but the IMC is powered by VTT and can have DIMMs drop off the bus. POST may even show memory as 12280, as can CPU-Z, but Windows may show 10GB and this may be the cause of it.

3) CPU UnCore speed is too high. Lower it one notch if you can in the BIOS and see if that helps the issue, if you can. Otherwise re-visit items 1 and 2.

Many here are fortunate enough to run 190+ mhz BCLKs and not have to loosen timings, but others are not. The chips can vary a lot.

For me I can reliably hit up to 200 mhz, but one mhz higher and DIMMs drop off, and/or I have stability issues regardless of other settings, even with VTT at 1.425/1.45V.

Good luck.
 
Last edited:
zero2dash said:
Manually adjust your Back-to-Back CAS Latency; doing that solved my disappearing ram issue. Found out about doing that from some NewEgg reviews referencing the Asus x58 boards and their forums.

Put it 1 lower than your timing. IE if your ram is 9-9-9 set it to 8. (It boots at 1 digit higher than what is in the bios.) Mine is 9-9-9-24, I set that to 8, I have not had randomly disappearing ram since then.
Click to expand...

Thanks for that - and sorry to hear you have still not had luck witht he NUMA disable trick. :(

I have been testing extensively this back to back setting, and will do a write up shortly - headline news: it can affect booting with/without all mem, but can have a large performance penalty.
 
BACK TO BACK CAS delay testing. [H]ard Data.™

Summary for those who are in a hurry:

Adjusting this setting can help disappearing memory, but with a potentially large performance penalty. I had better results simply loosening CAS timings a few percent across the board.

The details: Clearly the best thing to do while waiting around for a client to approve a render is to test memory settings for the good of the [H]orde.:D This is another great example of a lot of work finding out something is off by default for a reason ;).

Problem: disappearing memory at boot - intermittent. (eg 12GB of ram boots up as 10GB etc) The SR2 is very picky about RAM, and particularly when overclocked (esp with NUMA disabled) will sometimes boot without all memory. I experienced this on both SR2 builds so far. One because I was running with all 12 DIMMs filled, the other because I was running DDR 1600 memory at DDR 1870, requiring in both cases to loosen timings.

All values mentioned here for back to back cas delay refer to the value set in the bios - as Zero2dash noted CPU tweaker reads it as one value higher.

SR2#2 - 12GB (6 x 2GB) @ 1333 divider 2:10 = DDR 1870 at CAS 9, 10, 9, 24, 1T

This rig did not suffer much from missing memory - it was more a case of running it faster than spec, so I was already pretty loose on minor subtimings. (Running 9 10 9 24 rather than 7 8 7 20 etc)

So what happens when we change Back to Back CAS delay from AUTO (0) in bios? (which reads in CPU tweaker as Disabled)

original.jpg

The vertical scale on graph is mucked - 1.04 = 104% etc

DO NOT COPY SETTINGS for this graph blindly and assume it will work. This graph shows B2B cas delay setting of 8 as the least worst - but on my other rig it was 3!

At 3 (minimum possible) through 5 we have problems booting with all memory or being stable. 5 passes 300% HCI memtest, but one bootup in 3 attempts did not boot with all memory, so is disqualified. 8 was the best for this set of timings, with a penalty of only half a percent. The only way to be sure a setting is fine to to boot a few dozen times and see - not practical for a test like this. I really moved on to testing frame times - no point testing long term stability if there is no benefit. No options gave faster times than the existing settings, which are already loose enough for 100% reliable bootups with no missing memory.

SR2#1 - 24GB (12 x 2GB) @ 1066 divider 2:8 = DDR 1528 at CAS 8, 9, 8, 24, 1T

Ok, memory timings. This is voodoo shit, so if you are lost follow the recipe: First worry about divider speed. Do 1333 / 2:10 if you can. Then worry about if you can keep 1T command rate. Lastly lower your timings.

If I set all timings to auto at 1066 mem divider and MCH of 1067 I get
7 7 7 20 8 59 4 4 4 20 2T Auto 63 64 66
With Numa disabled this will bluescreen on Windows boot, although it does pass 10 mins of the very nearly useless memtest86+. As the RAM is sold as having timings of 7 8 7 20 it does not take a rocket scientist to work out one should change it to that and boots. TPF of 11:19 mins - almost as good as the settings I have been using, but too close to unstable to bother testing more.

Back to back cas delay in yellow:

Command rate 1T and loose:
8 9 8 24 10 67 5 5 5 24 1T Auto 73 74 75 = TPF of 11:18 (averaged 4 frames) (BASELINE - stable settings I have been using for a month)
8 9 8 24 10 67 5 5 5 24 1T 3 73 74 75 = TPF of 11:21 ** (averaged 4 frames) PASSED 100% memtest HCI, 10 reboots
8 9 8 24 10 67 5 5 5 24 1T 4 73 74 75 = TPF of 11:21 (averaged 2 frames)
8 9 8 24 10 67 5 5 5 24 1T 8 73 74 75 = TPF of 11:42 (averaged 2 frames)

Command rate 2T and tight:
7 8 7 20 8 59 4 4 4 20 2T 3 63 64 66 = FAIL stop error on boot
7 8 7 20 8 59 4 4 4 20 2T 4 63 64 66 = TPF of 11:22** (averaged 6 frames)
6 7 6 18 8 59 4 4 4 20 2T 4 63 64 66 = FAIL memtest86+ in 5 mins.

Command rate 1T and tight attempts:
7 8 7 20 8 59 4 4 4 20 1T 5 63 64 66 = FAIL boots with 20GB
7 8 7 20 8 59 4 4 4 20 1T 6 63 64 66 = FAIL boots with 20GB
7 8 7 20 8 59 4 4 4 20 1T 7 63 64 66 = FAIL boots with 20GB
7 8 7 20 8 59 4 4 4 20 1T 8 63 64 66 = FAIL to boot - watchdog timeout (??!)

Here I was able to set B2B to the minimum of 3 - which would not boot properly on the other system. This makes sense, as these are the settings that are much looser than they need to be to reliably boot with all 12 DIMMs - so there is slack there to cope with such a tight B2B setting. But even here none of the options with back to back cas delay set to anything other than auto (0) improved folding times.

Dropping from 1T to 2T allows you to tighten up everything else to compensate, so as you can (hopefully) see I ended up with 2 roughly equivalent speed settings (Marked ** in above list) So which do you chose if they give the same result? 1T with looser timings or 2T with tighter? Search your feelings. I prefer 1T and looser because it worked, I have done a lot of stability testing with it, and also it is easier to fine tune larger numbers (Going from 6 to 7 = 17%, 8 to 9 = 13%)
So if you are looking to tackle missing memory settings with one adjustment, perhaps this can be of use, but based on the above better results can be obtained by loosening all timings apart from 1T command rate by 10% or so.

I hope you have enjoyed another [H]ard Data™ post. Those of you with this setting set to auto kick back and enjoy another coldie, safe in the knowledge you are already doing it [H]ard :p
 
Last edited:
Wow, thanks for all of this. I want to try some of this when I drop this bigadv and I can play around some. For the moment I have all of my 6gb available. I have found that my problem was being caused by too much DIMM voltage, with NUMA off.
Running at 200blck with a 2:8 divider I have exactly 1600 on the ram. I was using 8-8-8-24 1T (which is the rated ram speed), but I wasn't able to get stable with it. After bumping up to 9-9-8-24 with 1.61v I have it running stable for about 5 days so far. It also seems that I have to have my VTT voltage higher than expected to stay stable. 1.42v had me worried, but the PWM temps are still low.

*edit* I fibbed, running 198bclk.
 
Great post/analysis, MIBW.
Looks like I might start farting around with B2B on auto, change my timings back, and re-try the whole disabling NUMA thing. If I lose ram, I'll just do what I did before and reboot and hope it pops back up. ;)
 
MIBW, you are the SR2 professor. Well done, sir.
 
Well, there is still a lot to learn about SR2 gremlins... after borking around with SR2#1 so much, I have had a nasty bug return.

Basically it will hang/slow frame times on folding. It happens on the second frame of my 2686 test unit - every time. I get a perfect first frame, then long on the second. (it is not a problem with the test unit - other units are messed up, just not as perfectly repeatable as my test unit.) Also at stock optimised defaults!

Last time I hit this I went nuts, and apparently solved it by the nuclear option: reflashing the bios, resetting cmos and manually reentering everything. Which fixed it and on I went for a month or so.

I have done this reflashing again (going to A50 while I am at it) but I am still stuck.

Ideas?
 
MIBW, damn good info, as always man. Hope you enjoyed a cold one after all that research. ;)
 
never mind, I think we are back and working again - after another cmos reset and manual enter of everything again..:rolleyes:

thanks all... yes, cold ones are in order :) the good news was I have slightly tightened my memory on SR2#2 during all that testing, and my second WU back folding again is an absolute scorcher of a 2692 - I haven't seen one faster. In fact the frame after I grabbed this was an 11:02... which cracks 168,000 ppd damn I want to break the 11 min barrier :cool:
original.jpg


Always nice when one machine is knocked out to have the other trying real hard to make it up.
 
I seriously need to ship you some good Canadian beer for all your hard work MIBW.

I saw the same things with the B2B CAS delay and abandoned it a few weeks back, but I didn't put it into pretty graphs and do serious testing. I just saw my TPFs went up even though my chips were running at a faster speed, so back to 3.9ghz I went.
 
wow 11.02 for a 2692 wu.. nice.

my last 2692 was 17.22 tpf with my dual quad core.
i guess six cores make a good difference.
 
What do you have the checkpoint interval set to (or default?).
It's possible the second frame is delayed due to IO or other processing associated with the checkpoint.

For testing purposes, trying setting checkpoint=30, which, for your frame times covers at least 2 full frames.

may i be worthy said:
Well, there is still a lot to learn about SR2 gremlins... after borking around with SR2#1 so much, I have had a nasty bug return.

Basically it will hang/slow frame times on folding. It happens on the second frame of my 2686 test unit - every time. I get a perfect first frame, then long on the second. (it is not a problem with the test unit - other units are messed up, just not as perfectly repeatable as my test unit.) Also at stock optimised defaults!

Last time I hit this I went nuts, and apparently solved it by the nuclear option: reflashing the bios, resetting cmos and manually reentering everything. Which fixed it and on I went for a month or so.

I have done this reflashing again (going to A50 while I am at it) but I am still stuck.

Ideas?
Click to expand...
 
Can you post a link to the work-units that you are doing performance testing and stability testing against? The raw data files from the work directory. It would be good if people did some testing against the exact same work units.

Also, has anybody generated test vectors for a known set of work units? eg: expected results for a specific 6701, 2684, 2685, etc. This would be a better test than just checking for hangs or faults during folding. Ideally, we should have test vectors for intermediate points as well.

If the above doesn't exist, I will look at saving off (and later analyzing) the results from work units above from a system that is stock and not overclocked, etc.


may i be worthy said:
Thanks for that - and sorry to hear you have still not had luck witht he NUMA disable trick. :(

I have been testing extensively this back to back setting, and will do a write up shortly - headline news: it can affect booting with/without all mem, but can have a large performance penalty.
Click to expand...
 
sfield said:
Can you post a link to the work-units that you are doing performance testing and stability testing against? The raw data files from the work directory. It would be good if people did some testing against the exact same work units.

Also, has anybody generated test vectors for a known set of work units? eg: expected results for a specific 6701, 2684, 2685, etc. This would be a better test than just checking for hangs or faults during folding. Ideally, we should have test vectors for intermediate points as well.

If the above doesn't exist, I will look at saving off (and later analyzing) the results from work units above from a system that is stock and not overclocked, etc.
Click to expand...

I have a half dozen work units saved off. I was working on a nicer front end to make a benchmark utility, but got frustrated with it over the weekend.
 
@10e - Canadian beer? Thanks, I think... I am sure you mean well.. ;P

@sfield - you have a PM with benchmark.

I think I have checkpoint set to min. I would love a current FAH test kit with known good outcomes. Right now I rely on dreaded 6701s to see if I have enough vcore to avoid BSODs

Anyhoo, I am off to pick up some more parts for SR2#3 :D

EDIT: ninja-ed Musky - I sent him the 2686. Sorry to hear the benchmark is a pita. A grateful [H]orde awaits. ;)
 
I have good news. I have completed a 2684 on my SR-2 running at 198bclk. Average tpf 24:32. Netting me just shy of 50k ppd.
It immediately picked up a 2685, I'm seeing 18:43 tpf right now, netting 76k ppd. This is where it's going to stay as long as it remains stable. These better overclocking and ppd results are making me enjoy my purchase that much more.

System is set to:
198blck (19x) 3762mhz
NUMA disabled
Memory 2:8 9-8-9-24
V-core 1.31 and 1.29
VTT 1.40 and 1.39
IOH 1.375
Temps are 62c and 66c cpu's
PWM 49c and 58c

Pulling 495watt from the wall, with gtx275 idle.
 
@nimbyfaygo - great to hear. I started out with crappy ram on SR2#1 and it's PPD was about two thirds of what it is today, and it was pretty miserable to have an underperforming expensive new machine.

Well, you know that fastest ever 2692 I got? Well, turns out I borked BOTH SR2's doing the B2B cas delay tests. I either restored the wrong saved settings or read my spreadsheet wrong, because I had the wrong vcore set, a bit too low...

Unstable machine at 94%. My first ever failed work unit. Lost almost 130,000 points.

images


Muck a Duck. Been that sort of a day :(

Moral of the story - never have so many SR2s you can't remember all the overclocks by heart :p
 
You must log in or register to reply here.
Back
Top