Performance Test V8.0 Build 1016 > Does not Complete > Dell Precision T5400

[BambiBoom]

RE: Performance Test V8.0 Build 1016 > Does not Complete > Dell Precision T5400

Hello Passmark,

In trying the latest version of Performance Test 8.0- build 1016, I have an error message > "Failed to switch to DirectX test to fullscreen mode", which then aborts the test. In checking the versions, this build is supposed to have addressed that error. The computer under test > Dell Precision T5400, Xeon quad core X5460 @ 3.16GHz, 16GB DDR2-667 ECC, Quadro FX 4800 (1.5GB), WD RE4 500GB, Seagate Baracuda 500GB, 875W PS, Win7 Ultimate 64-bit] The Quadro FX 4800 supports DirectX 10.

I had use the trial version of Performance Test 8.0 in December, 2012- sorry I don't have the Build number, but had no issues, and in fact was intended to purchase the current version as I'm preparing to upgrade the T5400 > install a 2nd CPU, install PCI-X SCSI controller, compare rendering and graphics performance between Quadro FX 4800 and Geforce GTX 285. Is it possible to use an earlier version or does that then not make baselines comparable?

Simultaneously, I am also writing an article, intended for Toms Hardware on the way in which Performance Test baseline searches might be used to help specify new system builds in three categories > gaming, workstation emphasizing 3D CAD, and GPU-based personal supercomputing. This article is currently 5,000 words but I plan to reduce that by about half. If Passmark is interested, I could submit it here as well.

Suggestions?

Cheers,

BambiBoom

[David (PassMark)]

Here is the download for V8.0 build 1015
http://www.passmark.com/ftp/petst801015.exe
Results are comparable between any V8 build.

We had a some users report crashes in the DirectX 10 test (prior to build 1016). Most of the time the crashes are due to bad video card device drivers. But in two cases we narrowed the problem down to the fact that the video card refused to go into full screen mode, instead leaving the benchmark test running in a window, which then crashed. Running in a window is bad as it means the test results might not be comparable between machines. Crashing is also bad for obvious reasons. To 'fix' both of these issues we now detect, in build 1016, if the video card was able to get into full screen mode, and 'gracefully' exit the benchmark test if it doesn't.

What we haven't been able to figure out is why a few systems refuse to go into DirectX full screen mode. (we estimate is might be around 1% of system). We haven't been able to reproduce the problem on in house machines.

We speculate that it might be some 3rd party software that runs on the affected machines. Maybe something like FRAPs that plays around with the 3D rendering or maybe something like a particular anti-virus package that pops up a desktop window when the 3D test is launched, forcing the operating system out of full screen mode and back to the desktop. Or maybe specific hardware. It happens on so few machines we haven't been able to establish what the machines have in common.

If possible can you save a baseline file (with build 1016, from the File menu) and EMail it to us. We can then investigate the full screen mode issue a bit more.

Article also sounds interesting.

[BambiBoom]

Attempts to complete PT 8.0 > Sequence of continued trials > Build 1016, Build 1015

David,

Thank you for your detailed reply and suggestions to configure Performance Test v8.0 such that the test will complete on my Dell Precision T5400.

Sequence of further attempts to complete>

1. Build 1016 > As Avast and Malwarebytes produce warnings and report pop ups so often > Disabled all warnings, notices, and reports in Avast v. 7.0.1474, enable "Silent / Gaming Mode"

2. Run> PT v 8.0 Build 1016 > Continued until error messages > 1. " Failed to switch to DirectX 10 test to fullscreen mode" Error message > 2. "The test was stopped by user before completing. The results for this test will not be used."

3. Results panel > Rating= 1157 CPU= 4796 2D=525 3D= not listed Mem= 676 Disk= 957. For reference, the last successful test of this computer (with 12GB RAM instead of the current 16GB) resulted in a system rating of 1623 with a 3D rating of (I think) 912.

4. Uninstall> Malwarebytes 1.70.0.1100

5. Defrag and reboot

6. Run > PT v 8.0 Build 1016 > Continued until error messages > 1. " Failed to switch to DirectX 10 tesrt to fullscreen mode" error message > 2. "The test was stopped by user before completing. the results for this test will not be used." Rating = 1150.6 with 3D results not displayed.

7. Uninstall > PT v8.0 Build 1016

8. Download and install > PT v8.0 Build 1015

9. Reinstall > Malwarebytes 1.70.0.1100

10. Run > PT v8.0 Build 1015

11. Test completed through DirectX 10 sequences

12. Build 1015 initiated DirectX 11 sequences, running- well, crawling- at between .28 and .33fps (- OK rate for Solidworks, but not Crysis 3!)

13. Test halted , error messages > 1. Title > "Mandel" " ! DirectX11 is not supported on the video adapter." Error message > 2. "Unable to open the Directcompute benchmark's result file. Results for this test are unavailable." System Rating= 1150.6 CPU= 4844 2D=520 3D= not listed Mem= 678 Disk= 934

This was not very methodical, but at least Iperhaps suggests that neither Avast nor Malwarebytes had a serious effect. I was interested to see that PT 8 Build 1015 tried to get the Quadro FX 4800 to run DirectX 11, which it did - very poorly and for only a few seconds. One thing I should mention is that I am running the FX4800 on a Solidworks partnered driver with 128X anti-aliasing and anything resembling games oriented towards maximum frame rates seem to be "unhappy".

I don't have a DirectX 11 card, only another DirectX 10 card, a GTX 285 (1GB)- and which is more 3D and frame rate oriented. It may have been PT v.7, but I tried the T5400 with the GTX 285 and produced a rating of 2239 with the 3D score quite a bit healthier than the ol' stick in the mud FX 4800. Do you think there's any value in trying Build 1016 or 1015 with the GTX 285 running a standard NVIDIA GeForce driver? That would at least either eliminate or identify the FX 4800 and/or specialized driver as the culprit. Perhaps the other 1% of machines failing completing are also CAD workstations? Using DirectX 10 Quadros ? Finally, is there any possible effect that I am trying the test using an expired trial version?

This is a mystery as I had such good results with PT previously with the T5400, and it was fun, some time ago, testing my oldest of five, a 1998 Dell XPS T700R [Pentium III 750MHz, 768MB RAM, ATI Radeon 9200 AGP X8, Ultra 66 SCSI controller, 80GB, 30GB HD, Win XP Pro]. I don't remember the rating but it was not in the infamous "Bottom 100". Amazing to think what the T700R's $2,200 cost would buy today,..

If you're interested in the, "mining Performance Test baselines data to specify high performance systems" article, send me an email where to send it for your review- when I can successfully half or quarter the 5,000 words,... My motto should be, "I never use one word where twenty will do just as well."

Thanks again!

Cheers,

BambiBoom

[David (PassMark)]

The DirectX 11 frame rates you saw (0.3 fps) were probably not from the video card itself.

DirectX provides a 'software rendering mode'. Which means all of DirectX is run on the CPU and not the video card. When this happens results are normally abysmal.

A change was made in build 1016 to prevent the DirectX 11 test falling back to software rendering mode when the video card does't support DX11.

Yes it would be interesting to try the other card with build 1016, as at the very least it would eliminate a another possibility for the 'Failed to switch to full screen mode' message.

[BambiBoom]

Further Tests > PT 8 Builds 1015, 1016 > Dell Precison T5400

The DirectX 11 frame rates you saw (0.3 fps) were probably not from the video card itself.

DirectX provides a 'software rendering mode'. Which means all of DirectX is run on the CPU and not the video card. When this happens results are normally abysmal.

A change was made in build 1016 to prevent the DirectX 11 test falling back to software rendering mode when the video card does't support DX11.

Yes it would be interesting to try the other card with build 1016, as at the very least it would eliminate a another possibility for the 'Failed to switch to full screen mode' message.

____________________________________________

David,

Today's sequence of trials of Build 1015 and 1016 on the T5400>

1. Installed Autodesk Revit 2011 driver 265.79
Reinstalled> Build 1015
Ran > Build 1015 with Quadro FX 4800, did switch to DX10 fullscreen and complete then went into the 'software rendering mode' you mentioned, running at .78fps >
Error > " Direct X 11 is not supported on this video adapter"
Test Halted

2. Uninstalled HP Quadro FX 4800 >
Installed EVGA GeForce GTX 285 (1GB) >
Restarted
Oddly, there was no message for new hardware, wrong, or to update driver. I have heard that Quadros will run GeForce Drivers, but that GeForce cards will not run Quadro drivers. The GTX 285 does use the same GPU, is 512-bit, has 240 cores, and general configuration as the Quadro FX5800, but with 1GB instread of 4GB.
Ran > Build 1015 on GTX 285 using Autodesk /Quadro FX 4800 driver 265.70
Error > "Failed to switch to Direct X 10 fullscreen mode"
Test halted

3. Downloaded from NVIDIA site, the latest GeForce Driver > 314.07
Unistalled > Driver 265.70
Installed > GeForce Driver > 314.07 [Clean Install]

4. Uninstall > Build 1015
Install > Build 1016
Ran > Build 1016 (GTX 285 Driver 314.07
Error > " Failed to switch to Direct X 10 fullscreen mode"
Test Halted

5. Uninstall > Build 1016
Install > Build 1015
Ran > build 1015 > went to DX 10 fullscreen and completed DX10
Ran DX 11 test at .78fps
Error > "Direct X 11 is not supported on this video adapter"
Test Halted


6. Disconnected second HD as this was the only variation in configuration from the last successful PT v8 test completion. That drive also has a clone of the system drive in case there are any system files that might be read on startup- though I think that doubtful.
Ran Build 1015 >went to DX 10 fullscreen and completed DX10
Error > "Direct X 11 is not supported on this video adapter"
Test Halted

7. Uninstall > Build 1015
Install > Build 1016
Ran > Build 1016 (GTX 285 Driver 314.07, 2nd HD disconnecteds)
Error > " Failed to switch to Direct X 10 fullscreen mode"
Test Halted

The interesting aspect to me is that Build 1015 would complete the DX 10 component but then try to run the DX 11 which halted the test, whereas Build 1016, which I believe was supposed to have solved the "failure to switch to fullscreen" error, was the Build that presented that error.

This was interesting as it seems these failures are not resident in the graphics card, specialized drivers, nor the generic NVIDIA "ODE" "business" driver for the Quadro. I have also completely uninstalled both anti-virus and anti-malware software. I had also updated the system BIOS to the latest (A11). Any thoughts as to possible BIOS/setup errors. Should I enable Enhanced Speedstep for example?

Is there any chance that the test is hobbled by my running it in expired trial mode? this occurred to me as there are i.e. two messages at test startup mentioning "some features are not available". If there is a chance, perhaps you could allow a second fully enabled trial to eliminate that potential? Or, I could try v.7 which I know ran on this current configuration with the GTX 285 (except with 12GB RAM instead of 16GB). If v.7 failed, that might help point to BIOS or setup problems as I have reloaded everything on the HD since then.

A diversion > I had a panic over the weekend as after all this time, I started monitoring system temperatures and was shocked when I saw a RAM peak of 93 C. This produced the thought that PT8 was stressful and possible failing due to hardware thermal throttling. Still, I've never had signs of heat throttling, warnings or thermal shut downs even on long 3D modeling or rendering slogs. The overclocking crowd seems to fuss when their DDR3 goes over about 55C. I then saw posts about hot DDR2-667 in Dell Precisions, including a fellow in a similar panic at the 89 C in his T5400. I've since added 4GB by subtracting 4 X 1GB and substituting 4x 2GB of Samsung that seems to have higher quality heat spreaders than the Hynix. I was unable to use software fan control to up the RPMs a bit. However, now the temperatures are peaking at 72 C- still frightening but apparently this was designed for that kind of high heat.

This has been one of the most mysterious software situations in a long while-and I hate to be in the 1% failure crowd!

Thanks again! You're very patient. I owe you one of those 1 gallon tins of Fosters!

Cheers,

BambiBoom

[David (PassMark)]

Can you actually have both builds installed at the same time in different folders if you want. This can save an uninstall and re-install.

> ...whereas Build 1016, which I believe was supposed
> to have solved the "failure to switch to fullscreen" error,

No. This is not the case. We ADDED this error message in build 1016 to prevent a crash on some systems and avoid having benchmark results that were run in different video modes.

So I think we have eliminated the possibility of specific video hardware or specific device driver versions causing the problem.

If possible can you save a baseline file (with build 1016, from the File menu) and EMail it to us. We can then investigate the full screen mode issue a bit more. We don't think it will be anything to do with running in trial mode. But we can send you a key if you can send in the baseline file.

As noted above, we speculate that it might be some 3rd party software that runs on the affected machines. Getting the baseline file will help in checking this. Besides antivirus software other possible applications that might cause a problem are tools that skin the desktop, allow remote desktop access, desktop recording software or other stuff that might hook into the video display functions.

The couple of similar documented cases we have both had a very large number of 3rd party software applications running on their machines.