FM_Jarnis
-
Posts
193 -
Joined
-
Last visited
-
Days Won
2
Content Type
Profiles
Forums
Events
Blogs
Posts posted by FM_Jarnis
-
-
Combined test has always been a problem. The 6970 had the same Combined issue you're describing. As far as I can remember it was indeed because of a driver issue.
What happens if you disable SystemInfo and manually select the amount of cards to use for rendering? Just run combined.
If you do manual GPU count select to something other than the actual GPU count in system that is going to participate in rendering, you just make the combined test completely pointless. It is a debug option only for cases where you get somehow bogus results with the autodetect (most commonly, 2-4 GPUs in system but Crossfire not enabled on purpose would require manual setting to 1 to get correct single card score).
6xxx series Radeons have a driver issue of some sort with 3DMark 11 combined test - more than 3 cards combined test score suddenly tanks. We've told AMD. I actually reminded AMD person about this just today.
This doesn't happen on 7xxx series, by the way (nor on any NVIDIA setups). It is specific to HD 6xxx cards and specifically with 4GPUs.
-
That's not true I was on 13.1 when 3dmark came out and was they weren't approved but the 13.2 beta where lol
13.1 did go through approval process - it actually did so before the launch.
It was specifically NOT approved because it didn't render the Fire Strike test correctly (part of the particle effects were missing). 13.2 betas fixed this problem. Approval is not just a rubber stamp; the driver actually has to render things correctly.
-
Overclockers -love- beta drivers
//edit: how can you not see LOD as a major sticking point for the 3DMark benchmark? Tesselation "optimisations" are clearly invalid for 3DMark submissions, but the LOD adjustments are not? Why?
Tessellation adjustments are in default driver UI. You could do these without realizing that you are messing up the benchmark load -> high priority to recognize and flag.
LOD tweaks require third party tools mucking up with internal settings that the vast majority of people have no idea about -> not so high priority.
-
Not true. Non-beta drivers are generally approved within 48 hours now.
Also if the LOD thing is a major sticking point, I can bring it up as potentially something we could go and detect. Can't offer instant fix, but we can investigate what can be done related to it.
-
My experience with Titan and 3DMark so far: 3-Way-SLI with Geforce GTX Titan results in unusual behavior. With the exact same driver settings (used Geforce 314.09) the framerate drops from 14.8 Fps (2 x GPU) to around 5.5 Fps and the screen "twitches" already when loading the combined test. Also while Game test 1 scales from 2 to 3 GPUs (ca. +25 %), I noticed the framerate in Game test 2 is reduced by about 10 %. The GPUs were not overclocked and the same goes for the Core i7-3960X used fot the few runs.
2 GPUs vs. 3 GPUs (one dedicated for PhysX @ driver) improved the overall score by around 10 %. 3-Way-SLI results in lower scores because of the bugged combined test.
We know there is a 3- and 4-way scaling problem right now and are working with AMD and NVIDIA to resolve this. We hope to have more news next week and this may require a patch to 3DMark.
At the moment our recommendation is not to use any scores of Fire Strike from 3/4 GPU systems as things are clearly not working right.
Drawbacks of working on the very cutting edge of DirectX and DirectCompute, doing stuff no game really is doing yet... the fluid simulation of the smoke particles on the bridge is somewhat ambitious as far as the code goes and multi-GPU adds complexity considerably
-
3- and 4-way SLI/Crossfire problems are entirely down to drivers. Even 2-way sli didn't work right until the very latest betas. Guess have to give some time for AMD and NVIDIA get those things sorted. I believe they have some vested interest too - whichever gets their quad-GPU support working first on 3DMark gets to dominate hall of fame lists everywhere until the other guy catches up
Also FRAPS incompatibility is probably something that needs to be fixed - not sure if this is FRAPS or 3DMark issue but my hunch would be that FRAPS does something when it hooks to DirectX stack that isn't entirely kosher.
-
Whatever those UI fixes are, please include application version somewhere in the main window / score window.
Whenever the benchmark code is touched, the version number does change. 06 binary version number did not change because it was not recomplied.
Are there any specific benchmarks where the version number is not shown clearly enough?
..and only PCMark 7 scores should be changed by the patch.
-
We've just pushed out an update to 3DMark06 installer (see, we haven't abandoned it!). This installer is officially compatible with Windows 8. Binary itself has not changed (and is still v1.2.0) but installer has been rebuilt with latest SystemInfo 4.15 and with updated OpenAL installer. It also includes updated exporter.dll, ensuring that saving and loading results will work even with latest SystemInfo installed.
Futuremark's own page is not yet fully updated (will be updated on Monday) but the file is already available from various mirrors, for example here;
http://www.gamershell.com/news_148511.html
We're also going to publish updates to 3DMark 11, 3DMark Vantage, PCMark 7 and PCMark Vantage in the near future. In addition to updated SystemInfo, these will also contain actual changes to benchmark code - mostly UI fixes and exporter.dll updates, tho in the case of PCMark 7 there will be changes that actually affect scores. More information on them as the updates become available.
-
By the way, Futuremark Hall of Fame now features Fire Strike, Fire Strike Extreme and Fire Strike physics score lists.
http://www.3dmark.com/hall-of-fame/
We didn't want to wait with this over the weekend just for QA so it was pushed out with very limited testing. Let us know if you see something hilariously broken or otherwise suspicious.
-
Only tested a few graphics cards
Only 52? I am disappoint.
(kidding, that's insane number...)
-
We tried to keep the actual test parts as short as possible while still getting reliable numbers. We know some run them over and over again (heck, we did in our QA testing) and after certain point you just want them to do their thing quickly and give numbers
-
If someone has a result file from sub-zero run, send one to info [at] futuremark.com and we can test our fix
-
sadly with my MSI 7970 and 680 Lightning the temp's did not drop past 0 on the chart.
There are a few new things to learn, 3DMark 11 is fine and runs no issues. Fire strike with the same settings would crash
:D Okay. Could be the chart drawing code... Will report this as a bug, will try to fix in future updates.
Fire strike is way harsher than anything in 3DMark 11 so it crashing on heavy overclocks where 3DM11 gets through is not unexpected. The sheer number of complex compute shader calls per frame in Fire Strike is... umm... we did hear from some vendors (no names) that there were too much compute shaders
But we like to torture GPUs to the limit. Only way to see what the're made of.
-
Is beta5 approved?
Yup they are
We actually haven't tested but I think they will be.
Looks like AMD is not changing the version number while updating the betas
-
Heh, I was like "uuuuh, that is some wicked Fire Strike score. Perhaps one zero too much"
-
Also a small note that many review seem to get wrong - tho I must say we also haven't communicated this well. 3DMark actually does use DX11.1. In a very small and technical way.
From the FAQ (which has also lots of other good info);
Does 3DMark use DirectX 11.1?
Yes, but only in a minor way and with a fallback for DirectX 11 to ensure compatibility with the widest range of hardware and to ensure that all tests work with Windows Vista and Windows 7 as well as Windows 8.
DirectX 11.1 API features were evaluated and those that could be utilized to accelerate the rendering techniques in the tests designed to run on DirectX 11.0 were used.
Discard resources and resource views
In cases where subsequent Direct3D draw calls will overwrite the entire resource or resource view and the application knows this, but it is not possible for the display driver to deduce it, a discard call is made to help the driver in optimizing resource usage. If DirectX 11.1 is not supported, a clear call or no call at all is made instead, depending on the exact situation. This DX11.1 optimization may have a performance effect with multi-GPU setups or with hardware featuring tile based rendering (sometimes found in tablets and entry level notebooks).
16 bpp texture formats
The 16 bpp texture formats supported by DirectX 11.1 are used on Ice Storm game tests to store intermediate rendering results during post processing steps. If support for those formats is not found, 32 bpp formats are used instead. This optimization gives a noticeable performance effect on hardware for which the Ice Storm tests provide a suitable benchmark workload (tablets, entry level notebooks).
There are no visual differences between the tests when using DX11 or DX11.1 in 3DMark and the practical performance difference from these optimizations is generally limited to Ice Storm on very low end Windows hardware, and on Windows RT.
So.. in theory Win8 vs Win7 is a small difference due to these DX11.1 optimizations but mostly they matter with tile-based renderers (ie. PowerVR based tablets and netbooks, some WinRT tablets) and in some cases multi-GPU scenarios (not sure how well AMD/NVIDIA drivers handle such scenarios yet)
But no, no special eye candy. Because, frankly, DX11.1 doesn't offer anything much in that department.
-
So, how are ya liking it? Fried any GPUs yet? Any info if the temp data actually works with sub-zero temps? (we don't have liquid nitrogen setups to test... code says it should work)
Specifically interested in feedback on the "view run details" and how the result validation works (note that in Pro you need to enable online validation or manually click the "?" and hit revalidate. In Advanced it is always on as long as you have network connection)
Also there is now an "anti-cheat" validation link on top of the "view run details" window - it can say once and for all if the result is photoshopped or not
(the only way you won't get the result from that link is if the user hides it on purpose and wants to block you from viewing it)Also just a heads up; you really want to use betas. 13.1 drivers do not render everything correctly (so they are not approved) - some particles are missing from Fire Strike demo and GT2. Use 13.2 beta3/4 to get correctly rendered scenes (well, except for some minor flickering in crossfire)
310.90 NVIDIAs on the other hand do not have SLI profile and perform considerably less well than 313.95/96 betas... Okay, even these are not absolutely 100% perfect in SLI - again, some flickering there. Complex particle effects and smoke simulation seems to be hard .
13.2 betas and 313.95/96 betas are approved. Approval also shows in validation and in the future we aim to have less than 48h gap from "new driver" to "approved".
-
Isn't Nvidia part of the BDP ?
Yes. http://www.futuremark.com/business/benchmark-development-program
-
Thanks for your answer, I guess I'll find out soon.
EDIT: LOD is not a problem but the easy fix will be to reset it to default value before render starts
Problem is making it in a way that doesn't break every time a new driver ships.
-
Im more interested to see how much secure it is seeing that Vantage/11 can be manipulated (+ORB).
Depends what you mean by that. I know two (theoretical) loopholes that were closed (and they will be closed in 3DMark 11 update as well) but I have no idea if they are what you are referring to.
And in any case, our philosophy is that casual hacking and tampering is prevented (ie. you can't just randomly edit files or modify executables to get better scores) but beyond that there is really very little we can do without truly draconian Punkbuster-style active monitoring of whole system state (which would unavoidably degrade system performance).
Also as to "hidden" / "undocumented" driver setting tweaks using nvinspector and the like, as long as the settings are not publicly available on normal driver control panel, we currently do not look at those. The whole LOD thing is extremely lame and obviously makes the result bogus as you are not rendering what you are supposed to be rendering. Claiming such a score to be legit is... well... I think it is pointless.
Only real way to "fix" that would be to include several rendered frames in the result file (making it considerably larger) and then do frame comparison server-side against DirectX reference rasterizer image. Problem with that is that the video drivers these days are not always deterministic - there is always very minor inaccuracies so it is far from trivial to do image comparison that allows such nearly-invisible minor inaccuracies that are considered normal.
Ultimately hardcore competition benchmarking is *your* hobby. We make some tools, you decide how you use them. If you collectively decide it is okay to render whatever on the screen and cheer for big numbers, well...
-
Are those temperature numbers going to be able to cope with negative numbers or possible -#.INF coming from the sensors?
We honestly haven't tested them with extreme temperatures. As far as reading the code, in theory it should.
So I guess the answer to that is "we think so - if it breaks for you, let us know"
-
Also the screenshots of the result view and detailed result view are now public (I know some of you saw a draft of this earlier...)
http://www.futuremark.com/images/screenshots/3dmark-results-ui-screenshot.jpg
http://www.futuremark.com/images/screenshots/3dmark-details-ui-screenshot.jpg
Also, while not shown in this image, the 2nd view (hardware details) will also show, on top edge of the screen, a link to the result validation if the result is valid. So a screenshot of 3DMark run is always going to offer a simple link which anyone can use to see if the result is photoshopped or not. No need to do forensics on the image any more
We also plan on iterating on the detailed view more based on your feedback. The ultimate goal is that the detailed view would show everything you need for a result screenshot without having to have separate CPU-Z/GPU-Z running. I'm fully aware that it may not be perfect yet (it was bit of a last minute addition) but we'll improve it for the inevitable first update.
And yes, both views are in windows that are fully scalable so you can easily fit them side-by-side at least on 1080p monitor.
-
FYI: Monday, February 4th. 18:00 UTC
Better stock up on needed liquids for some benching (both for cooling use and for personal use).
-
Well, the upcoming 3DMark will probably then be interesting in this regard...
3DMark 11 patch 1.0.4
in Benchmark software
Posted
Just put out a new update for 3DMark 11
http://www.futuremark.com/pressreleases/3dmark-11-version-104-released
Fixes some minor bugs, bundles the latest SystemInfo. Scores unaffected.
Hall of Fame will start requiring this version (and SI 4.15) in 1-2 weeks time. Suggest updating to this when doing record attempts etc.