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

[Ado6]

The temp jumping from 26C to 40C during idle can be explained by the intel speedstep option. Your clock and vcore will increase as you use your pc. With my 2500K it sits at around 28C when im doing absolutely nothing. When Im listening to music or on the internet the temp will go up a little depending on what I'm doing since the CPU isn't sitting at idle anymore, it's clock and vcore fluctuate depending on the load that is put on the chip. If you want to heat the chip up a little more, Intel Burn Test or LinX will stress the chip a bit more. Again, idle temps shouldn't be anything to be concerned about though. The load temps is what you want to look at and keep under control.

 

[Blitz1776]

Could just be you have a couple of retarded sensors (the sensors are designed more to check for TJMax, and are more accurate the closer you get to it), I have a I7 970 (OC'd 4.0) using an H50 Push-Pull, on Idles I've got cores that read anywhere from 18-34 (even when I had all power settings disabled). At stressing stays around the 60-70 band.

 

[Termie]

:WORK IN PROGRESS:
Time for an update

The idle temp issue still has me scratching my head... Using Core Temp and Real Temp the idle is not consistent. It jumps around between 26C and 40C (similarly to how the Vcore jumps around inside the ASUS bios). I don't understand how at one second the core can be 26C and then a second later be 40C... regardless, under load the temps are consistent and well within tolerances.

With a quick overclock, 40x Multiplier, 1.2 Vcore i posted the following results in Prime 95.



It ran for an hour stable before i got bored and decided to start this post (and format hard-drives in the background).

100% LOAD - after 30 minutes of Prime95
Core 0: 46C
Core 1: 51C
Core 2: 51C
Core 3: 54C

Seems reasonable/good for an unoptimized 4Ghz OC

Despite jumping around on me, i have noticed the idle temps have lowered slightly over the last few hours. The lowest readings ive seen at idle have been.

IDLE
Core 0: 26C
Core 1: 29C
Core 2: 27C
Core 3: 30C

Wow, Boss, looks like you're good. Nice load temps, and pretty good idle temps too.

So, how's the system feel? Pretty sweet? How about firing up a game or two and letting it stretch its legs?

 

[DrBoss]

The temp jumping from 26C to 40C during idle can be explained by the intel speedstep option. Your clock and vcore will increase as you use your pc.

Understandable, but the vcore and idle temps jump when i am doing absolutely nothing (not moving mouse, or running any background applications)... though perhaps windows 7 is randomly doing BS in the background.

 

[DrBoss]

Wow, Boss, looks like you're good. Nice load temps, and pretty good idle temps too.

So, how's the system feel? Pretty sweet? How about firing up a game or two and letting it stretch its legs?

System feels awesome. I've still got days of tweaking and software installation ahead of me, but it feels good to get her up and running. Thanks again for pushing for the SSD (as well as all the random help), its definitely a welcome addition. I am also loving how quite the case/fans are. At idle, it's silent. Only under gaming can the GPU fans be heard.

Refer back to my last update, i finished the post and included some initial Crysis benchmarks (and updated my CPU benchmark image). Crossfire is acting a bit flaky... but on the plus side, it does not appear GPU temps will be an issue.

Idle temps are lowering as well... appears the Shin-Etsu has a bit of a break-in period. The jumping around appears to have gone from a 26C-40C range, to a 21C-34C range.

 

[DrBoss]

Great Rig overall DrBoss, I hate to ask you the obvious question, but are you sure your running a push pull setup and not a push push setup. If the mount & tim is good that would probably be the culprit, unless your case isn't moving air correctly (doubt it considering the fractal case is really good).

haha, i would have to punch myself in the face if this were the case. Fortunately it's not. Fans are running Push-Pull

 

[DrBoss]

Prime95 ran for 6 hours last night (4.0Gghz, 1.2Vcore).

Prior to running the test i was tinkering a bit with higher OC's. 4.4GHZ at 1.3Vcore was stable in prime for 10 minutes (i stopped the test to tinker elsewhere). The highest recorded temp during the test was 61C on core 3. The other cores did not break 60C.

Seems the Archon is doing it's job well, but i do have a concern with the Vcore. While temps are within tolerances i am hesitant to raise beyond 1.3 Vcore. I've read in a few forums that 1.3V is really the highest you want to go for a 24/7 OC. Is this thinking generally accepted as true? I am tempted to keep raising the Vcore and Multiplier until Load temps reach 70C... but if that means i'm running a Vcore of 1.35 @ 3.8Ghz (hypothetical), am a likely harming the cpu?

also, is there a reason my 4th core is consistently hotter than the others (typically by 3C)?

 

[Termie]

This article may help you: http://www.xbitlabs.com/articles/cpu/display/core-i7-2600k-990x_8.html.

They hit 4.4Ghz at nominal voltage (1.29v). You did 4Ghz at 1.2v - is this undervolting? It sounds like your 4.4 at 1.3v is actually not a significant overvolt, nor is it causing any temp problems. Their hottest core at that setting was 69C. You were lower, but you do have a better cooler.

While the author raises vcore to 1.425 to reach 4.7Ghz, I don't think that's a long-term setup. Temps hit 85C. I don't care what they say - that's too high. People hitting 4.6+ are probably all doing so with serious voltage and high temps - I question the rationale behind that if you can hit 4.4 at near stock voltage.

By the way, look at the spread of temps on the four cores. Your 3C delta is insignificant in comparison.

 

[DrBoss]

You did 4Ghz at 1.2v - is this undervolting?

Termie, i am not familiar with undervolting, i simply set my multiplier to 40 and then set a "low" Vcore to start with. First attempt didn't post, slowly raised the voltage from their until i was able to get into Windows. At whatever that voltage was, it BSOD when i started Prime95, so i kept slightly upping the voltage until i was able to run Prime95 for an hour (which was 1.20v)

I assume undervolting is running the chip lower than it should be? probably helps to keep temps down, but is this bad for the CPU?

EDIT: Looking back on my screen capture on the last page, with the Vcore set to 1.2 in the bios, CPU-Z was recording the Vcore at 1.128... which seems like some pretty siginificant Vdroop. Thoughts?

 

[Blitz1776]

Undervolting is exactly that running the cpu on voltages lower than its default. It is used for one of three purposed generally. One is in order to achieve the same performance at a lower voltage. Two is to lower the voltage in order to reduce the amount of electricity required to run the unit. The last one is in order to reduce the amount of heat the processor produces to be dissipated (so that you can try to make a silent or near silent computer, with no fans or slow ones.).

 

[Termie]

Termie, i am not familiar with undervolting, i simply set my multiplier to 40 and then set a "low" Vcore to start with. First attempt didn't post, slowly raised the voltage from their until i was able to get into Windows. At whatever that voltage was, it BSOD when i started Prime95, so i kept slightly upping the voltage until i was able to run Prime95 for an hour (which was 1.20v)

I assume undervolting is running the chip lower than it should be? probably helps to keep temps down, but is this bad for the CPU?

EDIT: Looking back on my screen capture on the last page, with the Vcore set to 1.2 in the bios, CPU-Z was recording the Vcore at 1.128... which seems like some pretty siginificant Vdroop. Thoughts?

If I were you, I'd try overclocking at "auto" voltages first, just to see what the motherboard sets the CPU to on its own. For instance, clear out all custom settings on the bios and then try a 40x multiplier.

You were significantly undervolting your processor - this is fine if it can do it (my e8400 runs overclocked at the minimum voltage possible on my MB), but I also assume you in the process disabled dynamic voltage, which means you are at a very high idle voltage. This could explain your high idle temps. As it happens, that's fine with my e8400, since my manually set voltage is actually lower than the auto idle voltage, but that's not fine for you - your idle voltage should be 1v or less.

Plain and simple: 1.2v is extraordinarily low for 4.0Ghz and extraordinarily high for idle. See what your voltage is at idle currently and report back.

As for vdroop - yes, that seems like a lot, but perhaps other settings were affecting that. I'd like to see some voltage/temps at auto.

 

[DrBoss]

Plain and simple: 1.2v is extraordinarily low for 4.0Ghz and extraordinarily high for idle. See what your voltage is at idle currently and report back.

As for vdroop - yes, that seems like a lot, but perhaps other settings were affecting that. I'd like to see some voltage/temps at auto.

I will play around with this tonight and get some new numbers posted.

 

[DrBoss]

As requested, the following images represent AUTO overclocking through the Asus Sabertooth motherboard. The only thing i set to manual was the Multiplier. All load results captured while running Prime95

3.5 Ghz IDLE


3.5 Ghz LOAD


4.0 Ghz IDLE


4.0 Ghz LOAD


4.5 Ghz IDLE


4.5 Ghz LOAD


The 4.0 Ghz load temps are decently higher than my 1.2 Vcore results (previous page).

@ 4.5 Ghz, before running Prime95 (100% load kicked in VDroop), the Vcore was running around 1.38 doing menial tasks... which seems pretty high.

All results show that the CPU does a great job cutting the Vcore and multiplier during idle conditions.

Thoughts?



Time to start troubleshooting Crossfire scaling in Crysis.

 

[Termie]

This looks very good. Now you know what the baseline is. I'd recommend you find a way to overclock that doesn't leave you with a very high idle voltage - but that's something you'll need to figure out with your motherboard. It's tricky, because your chip can overclock at very low voltage, but you don't want to just plug that voltage in, because then you use that at idle too, which is too high. Maybe you can play around with a negative offset, which subtracts voltage from the auto voltages you found above, but that may set an idle voltage too low. Tricky, but hey, even the auto voltages/temps are pretty awesome.

 

[DrBoss]

Seems DX10 is to blame for shoddy crossfire performance in Crysis 1. DX9 using the "very high" tweaks basically doubled my framerate... time to move onto Crysis Warhead.

Any idea why installation of games would take forever on the SSD? Crysis loads extremely quickly, however; it took 30 minutes to install the game. Just started the Warhead installation and it seems to be more of the same.

 

[Termie]

Seems DX10 is to blame for shoddy crossfire performance in Crysis 1. DX9 using the "very high" tweaks basically doubled my framerate... time to move onto Crysis Warhead.

Any idea why installation of games would take forever on the SSD? Crysis loads extremely quickly, however; it took 30 minutes to install the game. Just started the Warhead installation and it seems to be more of the same.

Good to know on the crossfire front, although ideally it would work for Crysis DX10.

As for the long installations, you should probably run an SSD benchmark to make sure everything is within the right range. Start with Windows Experience Index - if the hard drive score is not above 7, you aren't in the right range. To determine what isn't working right, you'll need to run an SSD benchmark like AS-SSD : http://www.alex-is.de/PHP/fusion/downloads.php?cat_id=4. Your score should come out very similar to the one on the download page.

 

[DrBoss]

I spent the night troubleshooting Crysis and Crossfire. Turns out the problem is quite simple. Crossfire and DX10 do not play nice with Crysis in a 64bit environment.

BENCHMARKS
crysis | 1920x1200 | 8xAA | vsync off | all setting 'very high'

DX10 64bit
min:23.22 - max:75.74 - avg:52.15
DX10 32bit
min:49.73 - max:79.69 - avg:68.66
DX9 64bit
min:46.27 - max:102.43 - avg:76.67
DX9 32bit
min:47.07 - max:107.93 - avg:79.16

DX9 isn't as effected by running the 64bit version of the game, but DX10 obviously takes a big hit. In the 32bit environment, i am pretty damn happy with 68.66 average frames per second with all settings maxed. For reference, my DX10 32bit average frame rate using only 1 of the GPU's was 35.49fps. So in 32bit, Crossfire is scaling around 193% :]

The problem appears to be Crossfire GPU utilization in 64bit DX10. During that test, GPU 1 Load = 98%, and GPU 2 Load = 72%. Running in 32bit DX10, both GPU Loads were measured at 100%.

Also of note, the highest measured GPU temp during my benchmarks was 68.0C, but i assume that will rise with extended play.


Bedtime. Hope this information helps others.
Does it run Crysis? Yes

Will check the SSD tomorrow.

 

[DrBoss]

Have yet to check the SSD.
I turned 29 yesterday so there was cause to celebrate my almost being dead.

I did, however; check CPU scaling in Crysis. Running at 4.5Ghz (as compared to 4.0Ghz), i only gained 3fps on average. This was somewhat expected though considering at 4.0Ghz both GPU's were running at 100% - They are the bottleneck.

After running the benchmark i ended up playing Crysis (at 4.5Ghz) for at least 30 minutes. I had CPU-Z, Core Temp, and GPU-Z running in the background.

CPU Temps maxed at 68C - The Archon heatsink is doing an amazing job.
CPU Voltage under load (VDroop) was 1.306 - Need to move on from ASUS Auto OC, Voltage is too high.

GPU Temps maxed at 76C (the upper card runs about 5C hotter than the lower). Which must be an acceptable temperature considering fan speed never exceeded 64%. (which is barely audible). After a few days of testing, i feel pretty confident saying that the Twin Frozr III cards work extremely well so long as you provide 1 free pci slot between them. This thread, on HardOCP Forums indicates the cards do not perform well when stacked without a spare slot.

 

[rangda]

After a few days of testing, i feel pretty confident saying that the Twin Frozr III cards were extremely well so long as you provide 1 free pci slot between them. This thread, on HardOCP Forums indicates the cards do not perform well when stacked without a spare slot.

I'm the one that started the HardOCP thread and I'll echo DrBoss' sentiments, the frozr's need some room to breathe to run cool. However there is a caveat here and that is our different cases.

DrBoss' case is a conventional high airflow case. Mine is a Silverstone FT02 which rotates the motherboard 90 degrees and runs in a stack configuration. It pushes less air volume relying on thermodynamics (sucking in cold air at the bottom, exhausting hot air at the top) to do part of its work for it. The design of the frozr is such that the fans pull air directly into the heat sink and then the air scatters and flys around everywhere. (Which if you crank the frozr's fan to 100% you can easily feel with your hand, as a matter of fact most of the air gets dumped out on the power connector side rather than the port side of the card.) This design ends up fighting with my case, the card is ejecting most of its hot air right into my intake fans. This prevents the card from getting cold air from my intake and limits the effectiveness of the case cooling as the case loses its chimney effect. The temp difference was huge, I was seeing 90c peaks with the frozr (and 100% fan) vs 63c peak with a crossfiired 5870. But I really think at least some of this was due to my case configuration.

So while I agree that you really want that free slot or two between the frozrs, it very well may not be as dire as it was for me if you have a more conventional high airflow case. (If I was smart I'd have thought about this before actually purchasing the frozrs, but live & learn.) I'd also say that if you have one of Silverstone's stack cases (raven/fortress/TJ11) then I'd recommend you avoid the frozrs altogether as even when it works well, it will be fighting your case cooling.

 

[DrBoss]

rangda,
what cards do you plan to swap the frozrs for?

i was considering this Asus before coming across the MSI Frozr III. Reference cooling design, slight OC, and ASUS as you probably know is a solid manufacturer.

 

[DrBoss]

unrelated, but its time to cannibalize my old pc,

anyone interested in an HT Omega Claro+
http://forums.anandtech.com/showthread.php?p=31523430#post31523430

 

[rangda]

what cards do you plan to swap the frozrs for?

I replaced them with 2x MSI reference cards. I agree that ASUS is a solid company; I've used their motherboards for over 10 years. I picked the MSI's purely based on price and the assumption that all the reference boards are pretty identical and I can use afterburner for a slight overclock if temps allow it.

Replacement cards arrived before the frozr's even left (go go Amazon prime), just ran them through 1 hour or so of BF2. Peak temps were 66/64c which is quite reasonable IMO with them right next to each other. (I do find it interesting that there is no real gap in temps between the two cards.) I did have to use the zip tie trick to separate the cards a bit, the shrouds were touching when I installed them.

I'm really starting to think that the cards working along with the airflow design of the case instead of against is making a big difference.

 

[DrBoss]

I started playing around with the offset voltage function and have ended up with some pretty encouraging temps.

4.4Ghz
offset: -.06

LOAD
voltage moved between 1.200 and 1.218


IDLE
voltage moved between 0.918 and 0.936


When not under 100% load, the 4.4 OC runs a voltage around 1.27. So the Vdroop is in the neighborhood of 0.06.

I only ran Prime95 for 15 minutes, but in that time temperatures never rose over 60c.
I am going to run the stress test overnight to ensure stability.

 

[Termie]

I started playing around with the offset voltage function and have ended up with some pretty encouraging temps.

4.4Ghz
offset: -.06

LOAD
voltage moved between 1.200 and 1.218


IDLE
voltage moved between 0.918 and 0.936


When not under 100% load, the 4.4 OC runs a voltage around 1.27. So the Vdroop is in the neighborhood of 0.06.

I only ran Prime95 for 15 minutes, but in that time temperatures never rose over 60c.
I am going to run the stress test overnight to ensure stability.

Amazing temps and awesomely low voltage for 4.4. Seems most guys are over 1.3 at that OC. But I'm not sure you're analyzing the "vdroop" correctly, assuming you still have turbo on. The processor needs a different voltage to run at 4.4 full load than it does to run at 4.4 low load. When turbo is engaged, the processor will actually require a higher load. Thus, you see 1.27 at 4.4 low load (which would be at turbo). Of course, you'd actually see the speed ramp up too on one more more cores, so it would be pretty obvious what was going on. Get Tmonitor to help you figure out what's going on with the cores. Actually, not many people mention it, but this Intel utility looks pretty cool as well: http://software.intel.com/en-us/art...intel-core-processors-codenamed-sandy-bridge/. Only works on Sandy Bridge so I can't use it.

 

[DrBoss]

Amazing temps and awesomely low voltage for 4.4. Seems most guys are over 1.3 at that OC. But I'm not sure you're analyzing the "vdroop" correctly, assuming you still have turbo on. The processor needs a different voltage to run at 4.4 full load than it does to run at 4.4 low load. When turbo is engaged, the processor will actually require a higher load. Thus, you see 1.27 at 4.4 low load (which would be at turbo). Of course, you'd actually see the speed ramp up too on one more more cores, so it would be pretty obvious what was going on. Get Tmonitor to help you figure out what's going on with the cores. Actually, not many people mention it, but this Intel utility looks pretty cool as well: http://software.intel.com/en-us/art...intel-core-processors-codenamed-sandy-bridge/. Only works on Sandy Bridge so I can't use it.

Unfortunately, the stress test didn't make it through the night. @ -.06 offset Prime95 encountered faults around an hour and ten minutes. It did not BSOD, but simply canceled the test. Not sure if that is a sign my numbers are close, but past failed OC's have BSOD.

This morning i changed the offset to -.05 and started running it again. It's got a good 10 hours to run before i get home from work, we shall see if it makes it.

I will check out that utility tonight. When the computer initially boots, it sits at 4.4Ghz with a voltage around 1.27 (with low load), after a minute or two it scales back to 1.6Ghz. While Prime95 is running, the voltage fluctuates between 1.200 and 1.218. Is there anyway to stress turbo mode (1core)?

I am also going to install MSI Afterburner and start playing around with my gpu clocks. During extended play my gpu temps have yet to creep out of the low 70's (with fans never exceeding 65&#37, and most of the time they stay in the upper 60's. Figure i've got some headroom to play with.