Originally Posted by whiz882
GTX590 Quad SLI Reviews
☆3DMARK VANTAGE / CPU 4.8GHz
Crysis Benchmark Results Thread ↓
☆Crysis DX10 Results : 2560x1600, 8xAA , Very High
Avg. FPS :72.815
☆Crysis DX9 Results : 1280x1024, No AA, High.
Avg. FPS :127.16
Good job doing benchies Wiz, but why is your QPI so freaking high for a 990x?
Look at your GPU usage under the Crysis benchmark, its looks to be about 65~70% as well. I anticipate you might have an unstable overclock, probably CPU territory.
On the Crysis 4GHz run, you look to be undervolted quite a bit, have you run stability testing? If you are getting core errors, its going to greatly reduce you score. Over voltage on the QPI can increase likelihood of this as well.
You maybe able to increase your scores greatly without increasing your speed, just adjusting the voltages.
I personally run (for 24/7) 4.100 GHz @ 1.270v (full load), so about 1.31875v in BIOS with VDroop on. My QPI is only +75mV, so its stock of 1.200v + 75 = 1.275v, with a 30 mV LLC, making it 1.305~1.310 at full load. You are entitled to run things however you want, but the Gulftowns are more sensitive from voltages when it comes to error, too high or too low can cause errors. Those voltages are as low as I can get without getting either a Core failure / BSOD. IBT 100 passes Max Memory and Prime95 Custom Blend with 21000 MB of RAM stable for 16 hours straight.
The max safe vCore & QPI voltages I have read is approximately 1.40v. However, most enthusiasts agree going over 1.400v on QPI causes errors, and does not aid overclock. Now on the vCore, error town (without sub-zero) begins appearing around 1.59~1.60v, thus making your max OC limited to vCore.
Also what are your Uncore/Northbridge & QPI frequencies? Having a stock Uncore can hold back your benchmarking. QPI should be set to around 3.5 GHz and if you don't have direct control over that, have to adjust Bclk and lower CPU multiplier. Its more complicated I know but worth it. If you scroll down the EVGA's hall of fame
, you can see I am in 4th place on 3DMark11 for 2x GPUs, but its primarily due to a boost from my Combo score at the end, nearly a 1000 higher, just from raising my Uncore to a stable 3.6 GHz as well. It increases system intercommunication greatly. In addition to that, I have added a small boost to my IOH Core voltage, which has been shown to increase the performance of multiple card / multi GPU setups. I run that at approximately from 1.100v to 1.225v, not sure what the live voltages are but I assume it doesn't go much higher or lower. 0.060v for every GPU
Originally Posted by rush2049
Okay so question..... was playing some l4d2 earlier and it BSOD on me..... I had reverted everything in my system to stock, stock cpu, stock gpu, stock mobo, stock ram......
I didn't even have a custom fan profile for the 590 going and it BSOD'ed after ~1 hour of L4D2..... now I don't know what would the cause be, but I am guessing heat.
I tried to start up immediately and got screens looking like the following (frozen at this point):
See the pixels of color across the top..... now what is peoples opinion on this?
After 20-30 min of waiting with the power off / cord pulled out, it booted up fine and I checked all the temps, and they are all < 40 C...... I don't know what is going on.
Where it freezes (not just how) can be many things, and where you freeze really opens up a ton of things. It could be because of unstable clocks (which you said are running stock), GPU Heat, GPU Drivers, System files corrupting.
Every BSOD has the chance to break Windows, we are just living in an age where there is tons of redundancy and repairability. I doubt the card has damage, due to the Windows loading logo has to produce virtually no load on the GPU, thus I would say its a software based artifacting as opposed to physical problems, UNLESS, you continue to have issues in games.
That length of time you reported would definitely have me concern because an hour in game should already be at its maximum temp for whatever game that is, 70c, 80c, 85c, 90c <-entering the unsafe zone there, etc. So try it again using on SCREEN temp monitor such as Afterburners OSD which can monitor everything (it can be somewhat annoying to really be playing but just keep an eye on temps while playing for an hour.) If the same thing occurs and its 85c and below on each GPU, I would anticipate either driver OR another piece of hardware such as RAM. At which point you might want to post more data for us to diagnose better like the memory dump/stopcode etc.
Your best bet is to see what caused it (I suspect drivers) and then just re-install them to see if it helps. Could just be a driver bug with L4D, nVidia can have them, RARE I know
To find out some info on the Crash (for anybody who wants to know)
Start -> Control Panel -> Administrator Tools -> Computer Management -> Event Viewer
Display all sub-options after selecting Event Viewer, then go Windows Logs, and expand it.
There are two spots I check, Applications and then Systems. Looking primarily for a red circle with a ! (exclamation point) in it. This will give you basic info (without having to tear through a memorydump if you did infact get one.)
Once you find it you open it up into its own window or just click it on it and see the summary below the list of what kind of error. You can even post in here if you are unsure, it has a .XML format which is easy to follow if you are familiar with XML.
Look for values or .dll or .dvkm that start with nV <nVidia (or ATI which I don't know why you would be here lol.) nvkdlm.dll is a big one. Also DX errors can appear under System OR Applications. They can be sneaky, and in Applications, you may only get the game executable, which I have seen this being still related to drivers, but its less definitive as it may be a game bug.
Edited by RagingCain - 4/13/11 at 9:15am