GPUPI - SuperPI on the GPU

[_mat_]

 

In the last few weeks I've been working on a benchmark to calculate pi on graphics cards. It's an idea that's in my head for some years now, but I couldn't put it to life until now. Somehow the death of our beloved Turrican inspired me - I wanted to do something good, something that matters. So I tried hard and it finally worked out. I want to dedicate this benchmark to Turrican, like a virtual monument to never forget his legend. A benchmark in his honour.

 

Together with the community of overclockers.at I've tested the benchmark for the last week and put out some new versions, wrote a technical description and added some frequently asked questions. It's still a beta version, but it should be pretty stable with most of the graphics cards and processors around. But be sure to install the latest GPU/OpenCL drivers for the target device, old ones can be very bugged.

 

Here are a few examples for calculating the billionth digit of pi:

 

GPUPI 1B: AMD Radeon R9 290, NVIDIA GeForce GTX 980 and Intel Core i7-4960X@4 GHz

 

I think that there's a lot of potential in a benchmark like this, it's just fun to crunch pi on a gpu. If you like it, I would consider implementing multi gpu support and hwbot direct submission. Bugs and feature suggestions can be made here or in the international support thread at our forums.

 

Download, technical details & FAQ: GPUPI

Edited January 18, 2018 by _mat_

[ObscureParadox]

A graphics bench that is actually only using the GPU, about time really

[der8auer]

That's awesome! Maybe we can add it as Beta benchmark.

[Farjam]

awesome work

[TaPaKaH]

Does it also work in XP?

[_mat_]

Thanks!

 

I've already written Pieter on Facebook, but he seemed busy. He said that he will have a look tomorrow or so. The benchmark is already submitted to the HWBot application thingie.

[_mat_]

Does it also work in XP?

Yes. Target platform for compilation was Windows XP upwards. Please try it out, I didn't test it myself yet. But I wouldn't know of any advantage in doing that. Currently it seems like the newest drivers get the best results.

[Calathea]

Hahaha such an awesome idea! I will try it out a lot, thx!

[_mat_]

 

I've just found out that I can release the benchmark in a beta version myself. It's now listed in the Videocard benchmark list.

[GENiEBEN]

 

I've just found out that I can release the benchmark in a beta version myself. It's now listed in the Videocard benchmark list.

 

Please rename the benchmark to GPUPI 1Billion or whatever default setting you expect us to run.

And good job on this!

[Calathea]

Wow, my temporary pc with HD5770 lacks double precision lol. Maybe my quadro 4000 at work will do better.

[GENiEBEN]

Wow, my temporary pc with HD5770 lacks double precision lol. Maybe my quadro 4000 at work will do better.

 

Haha yea I remember that stunt from ATi, they kept it only for 58xx even though previous generation had it enabled on most models.

[_mat_]

I renamed the bench to "GPUPI - 1B". And thanks!

 

The first submissions are rolling in ... yeah!

 

My GTX 980: http://hwbot.org/submission/2672629_

Edited November 11, 2014 by _mat_

[Calathea]

Haha yea I remember that stunt from ATi, they kept it only for 58xx even though previous generation had it enabled on most models.

I barely even know what it means but was suprised when I saw HD4850 supported... HD5770 - NOPE.

 

I renamed the bench to "GPUPI - 1B". And thanks!

 

The first submissions are rolling in ... yeah!

 

My GTX 980: http://hwbot.org/submission/2672629_

How do you think performance between quadro and geforce are? If it's anything like oclhashcat then there's virtually no difference...?

[_mat_]

I barely even know what it means but was suprised when I saw HD4850 supported... HD5770 - NOPE.

 

 

How do you think performance between quadro and geforce are? If it's anything like oclhashcat then there's virtually no difference...?

I've seen no difference at all. High GPU core frequency is what counts, and of course choosing the best graphics card for the task.

 

It seems like the R9 290 is the best choice so far, because it has a very good integer performance.

[dinos22]

tried but couldnt get it to work even after downloading that software (which was already installed)

 

good work though, 1M seems super slow though, am i reading into this right?

[_mat_]

Dinos, have you read the FAQ? What's the problem? Which graphics card are you using?

 

No, it's not 1M, it's 1B. 1 BILLION!

1M is like 0.01 seconds or so.

[dinos22]

Faints

 

I read faq and had to download software but when I did it was already installed so stopped install. Get same error messagr missing xxxxx

 

I'll mess around again at some point

 

Dino

[_mat_]

I read faq and had to download software but when I did it was already installed so stopped install. Get same error messagr missing xxxxx

It should be easy. These DLLs are needed for every C++ application that's done in Visual Studio 2013. Just be sure to install the vcredist_x86.exe and run GPUPI.exe, the 32 bit version of the bench. Ignore the 64 bit version, even if you're on a 64 bit OS. It's not faster in any way.

[GENiEBEN]

Mat, is this done using Microsoft's AMP implementation of OpenCL? (haven't downloaded the benchie yet).

[_mat_]

No. It's pure OpenCL with C++. I used no libraries/APIs except pure WIN32, STL and of course OpenCL via the Intel SDK.

[skulstation]

i know its beta.

my gts 250 and Radeon HD 7650M not detected

[Schmuckley]

I like it! don't hate

I didn't have the right C-runtime libraries.so I had to dl dll though.

[_mat_]

Under 20 seconds with R9 290X.

 

http://hwbot.org/submission/2673063_nata_58_gpupi___1b_radeon_r9_290x_19sec_690ms

[_mat_]

i know its beta.

my gts 250 and Radeon HD 7650M not detected

Did you install the newest drivers? Are you sure they can handle OpenCL? If the device doesn't support double precision but can be detected on the system, it will get listed as ignored when starting the benchmark. I'm not sure about your cards, GTS 250 seems to have double support, I think.

 

Detection is mostly a driver issue and has not much to with the benchmark itself.