DirectCompute Benchmark for Vista SP2 with DX11 or Win7


Recommended Posts

That might be it. Anyway, I re-ran the benchmark:

DirectCompute%20Benchmark%20v0.35%20by%20Pat.png

EDIT: I installed the 64-bit ATI Stream SDK v2.0 and now, the DirectCompute benchmark won't even start. :|

Nevermind, I fixed it by renaming "OpenCL.dll" to "OpenCL_ATI.dll". There still isn't any OpenCL support though.

Mine is weird... nVidia GeForce 295 GTX, yet with such crappy scores...

CPU is fantastic though! :D

Couldn't run Open_CL, keeps crashing for me!

Yeah, it's gotta be dual-GPU video cards. This benchmark doesn't seem to work well with them. So far, I've beaten an HD 4870 X2 and GTX 295 with my HD 4870. That shouldn't be.

I found the latest beta version (v0.45b) here. OpenCL works with it, provided you have the ATI Stream SDK v2.0 installed.

Here are my results:

2cxahx.png

2njevte.png

OpenCL works with it, provided you have the ATI Stream SDK v2.0 installed.

Which is still stupid, the latest drivers should come with it as part of the package.

EDIT: Well, 0.45b refuses to even benchmark my GPU, keeps benchmarking the CPU which keeps jumping around in the score.

Which is still stupid, the latest drivers should come with it as part of the package.

EDIT: Well, 0.45b refuses to even benchmark my GPU, keeps benchmarking the CPU which keeps jumping around in the score.

Perhaps they'll include it in the next driver release. Anyway, try running the benchmark with the ATI Stream SDK v2.0 x64 installed. You can get it here. You need to register an account though.

Anyway, I'm pretty sure v0.35 isn't working properly. People are getting insanely high scores with single-GPU video cards. The latest version, v0.45b, seems to work fine for most people. You can get it here.

any reason why directx 9 is provided with this program? (can't run benchmark on xp with directx 9.0 :laugh:)

not sure if i missed it but does anyone know what this benchmark program is actually doing to test the cards?

In case you run into any compatibility issues, I assume. And it's quite obvious as to why the benchmark won't run on Windows XP. It doesn't support DX10, let alone DX11.

I did say:

Well, 0.45b refuses to even benchmark my GPU, keeps benchmarking the CPU which keeps jumping around in the score.

And, then you tell me where I can get 0.45b, so why would you tell me where to get it if I already have it & tried it, and it doesn't work:

The latest version, v0.45b, seems to work fine for most people. You can get it here.
I did say:

And, then you tell me where I can get 0.45b, so why would you tell me where to get it if I already have it & tried it, and it doesn't work?

Because I'm blind and the sky is falling down. :rolleyes: I know what you typed and the latter half of my post wasn't directed to you. It was directed to everyone else. Good job on acknowledging my attempt at helping you by pointing out something so ridiculous and trivial.

Off topic but I'm curious to how you get screenies of Windows like that? how you do that?

Back on topic...mine.

<Snipped>

http://localhostruploadr.com/

/offtopic

Also, use v0.45b of the benchmark. The version you used (0.35) seems to be buggy.

Because I'm blind and the sky is falling down. :rolleyes: I know what you typed and the latter half of my post wasn't directed to you. It was directed to everyone else. Good job on acknowledging my attempt at helping you by pointing out something so ridiculous and trivial.

http://localhostruploadr.com/

/offtopic

Also, use v0.45b of the benchmark. The version you used (0.35) seems to be buggy.

Off topic thanks, I know where they were uploaded too...what I mean was how where screenshots themselves made? I know when you capture a window in OSX it captures it like that but how you do it in Windows?

Off topic thanks, I know where they were uploaded too...what I mean was how where screenshots themselves made? I know when you capture a window in OSX it captures it like that but how you do it in Windows?

The program automatically grabs the Window and uploads it, with transparency and shadowing. It isn't a feature of Windows 7/Vista. There is a dedicated program for capturing Windows in much the same way as the localhostr uploadr; however, it isn't free. It's called Window Clippings.

Back on topic:

ATI video cards (HD 4000/5000-series) seem to get better scores than their NVIDIA counterparts. It seems as though OpenCL works well with ATI video cards right off the bat.

Off topic thanks, I know where they were uploaded too...what I mean was how where screenshots themselves made? I know when you capture a window in OSX it captures it like that but how you do it in Windows?

You take a screenshot of the window with a white background, then another one with a black background, and then you can reconstruct the alpha channel of the window by reading the green channel of both of the images, then you take the screenshot of the black window background and add the alpha channel to it.

The code's pretty simple, and pretty fast when written well (Mozilla use it for windowed plugins and GTK widgets and other things without issue, but they're the ones who wrote it originally), my implementation sucked but it was enough to get Mike's working :p

Edit: Actually, a DirectCompute/OpenCL/CUDA version of it would be quite interesting, and very fast as well.

This topic is now closed to further replies.
  • Recently Browsing   0 members

    • No registered users viewing this page.
  • Posts

    • Researchers claim Microsoft's quantum breakthrough is flawed by basic Python errors by Karthik Mudaliar Microsoft's aggressive roadmap to deliver a commercial quantum supercomputer by 2029 has now hit a bit of a snag, and it's not because of a complex sub-zero dilution refrigerator, but rather because of a few lines of basic Python code. A new critique published in the scientific journal Nature argues that simple software errors effectively manufactured the breakthrough that Microsoft's foundational research claimed back in 2025 into Majorana-based topological qubits. Topological quantum computing, the path that Microsoft chose for its research, relies on creating and controlling "Majorana zero modes." These are exotic quasiparticles that theoretically offer vastly superior error resistance compared to the highly sensitive superconducting qubits currently being championed by rivals like Google and IBM. However, physically proving you have created these particles requires sifting through massive amounts of complex electrical conductance data to isolate a specific "topological gap." Because of the sheer volume of data, physicists rely heavily on custom software pipelines to process the results. This is where the Python scripts come in. Now, according to the critique, Microsoft’s data processing software contained fundamental programming errors that ultimately skewed the published results. By mishandling data arrays or deploying incorrect logic within the Python script, the software supposedly discarded "noisy" or contradictory data. Which is why it only highlighted the specific electrical measurements that supported the topological-gap claim. The researchers behind the critique argued that this makes the findings invalid, suggesting the heralded "quantum leap" was actually a false positive generated by bad code and not a product of groundbreaking physics. However, Microsoft is pushing back hard against these allegations. The Redmond giant has formally rejected the criticism, saying that it's just a minor anomaly rather than a fatal flaw. According to the company, while there may have been a minor oversight in the data parsing scripts, it does not alter the fundamental reality of their physical experiment. Just weeks ago, Microsoft unveiled the Majorana 2 quantum processor, a milestone so significant that the company boldly accelerated its timeline for a commercial quantum supercomputer from 2035 down to 2029. But the new software allegations reopen an old wound. Microsoft's quantum division faced a remarkably similar crisis when a landmark 2018 paper on Majorana particles was famously retracted in 2021 after independent physicists discovered the data had been inappropriately cropped. That historical baggage makes the current Python-related allegations particularly sensitive. If the foundational math and data processing for the 2025 breakthrough are genuinely flawed, the highly anticipated 2029 commercial timeline could easily be delayed or, worse, cancelled.
    • Because of what they have done to VMware I will never buy anything Broadcom again.
    • AMD releases hotfix for driver install issues on Windows 10 PCs by Taras Buria Earlier this week, AMD released an important graphics driver update. Version 26.6.2 brought AMD FSR 4.1 support to the previous-gen Radeon lineup, the RX 7000 series, giving users better upscaling tech that was previously locked to the newest GPUs. However, the driver turned out to be a little buggy, with users reporting installation issues on systems still running Windows 10. AMD quickly acknowledged the bug and today released a hotfix to resolve the problem. The AMD 26.6.3 Hotfix update is now available for download from the official website. Given that it is a hotfix release, it has only one change in its release notes: AMD announced the update on its official X account and added that a WHQL driver update with the necessary fixes would be released next week. Meanwhile, users can apply the hotfix or roll back to the previous driver using the official AMD Cleanup Utility. You can download AMD Software: Adrenalin Edition 26.6.3 Hotfix Preview Driver from the official website here. It is compatible with all currently supported graphics cards and 64-bit Windows 10 and 11. Full release notes are available on the same page.
    • With Microsoft now listening to its core audience and acting upon received feedback, fans can finally expect a much better version of Windows 11 than what was available five years ago. Here is to five more years, Windows 11! I guess we all need a good laugh now and again...
  • Recent Achievements

    • Dedicated
      Scoobystu earned a badge
      Dedicated
    • First Post
      Tom Schmidt earned a badge
      First Post
    • One Month Later
      D0nn13 earned a badge
      One Month Later
    • Rookie
      +ChiefOfNeo went up a rank
      Rookie
    • One Year In
      Tom Schmidt earned a badge
      One Year In
  • Popular Contributors

    1. 1
      +primortal
      466
    2. 2
      +Edouard
      177
    3. 3
      PsYcHoKiLLa
      123
    4. 4
      Michael Scrip
      82
    5. 5
      Xenon
      76
  • Tell a friend

    Love Neowin? Tell a friend!