Hi

Could someone with ati radeon 6990 or some other beast post a video of a workunit crunching

Thanks

What's the point of going through the hassle of making a video of the card crunching?

MrS

---

Scanning for our furry friends since Jan 2002

Nothing,just curious

HD6970, 910 MHz core clock: 2 WUs every 100 s. 1.10 V for the chip, so power consumption may be around ~150 W. Cooled by a Thermalright Shaman - cool & quiet. Sorry, no video ;)

MrS

---

Scanning for our furry friends since Jan 2002

How bout a screenshot than? :)

Are you trying to learn something in particular with this?

No,like i said i`m just curious.Just wanted to see the speed difference between my gpu and a power beast like 7970 or 6990

You can get this data from Zydor as he is currently crunching in Donate@Home.

He has 2 AMD PCs online and one gets an RAC of @1.7 million from 4 x HD5800, and his 2 x HD7970s (on the other rig) get an RAC of @1.5 million ATM

---

Go away, I was asleep

No,like i said i`m just curious.Just wanted to see the speed difference between my gpu and a power beast like 7970 or 6990

Perhaps it is better not to see how bad is gtx560 for Milkyway@Home.
The good old 4850 is twice faster than your gtx560.
6950 is almost three times faster than 4850.
6990 is ....

I'm currently crunching with the following:
CPU: i7-3960x @ Stock Speeds
RAM: 32GB of Quad Channel 2133Mhz
GPU1: Radeon HD 7970 3GB @ Stock Speeds
GPU2: Radeon HD 7970 3GB @ Stock Speeds

On Average, each work unit is completing between 50-60 seconds.(GPU WU)

So if you want to see how these components perform with Milkyway@home you can look at my tasks.

Something I'm curious of, is if there's a way to have only my GPU's do the crunching since they take 55-60 seconds per WU, yet the 1.00 CPU WU's take around 6000 seconds(Granted, it can run 12 WU's at a time so about every 500 seconds I'm completing a CPU WU)

I am disappointed, I expected much more from 7970.
My 6950@900MHz complete workunit for 60 sec.

Actually, i just noticed that the GPU's werent getting enough assistance from the CPU's since I had 12 CPU WU's running and 2 GPU WU's running. I have since then changed max CPU's used to 90% and that allows 2 to stay free to assist the GPU more.
The GPU Usage was only 60-85% before, now it's 85-96%.
I'm going to test freeing up another CPU core so the GPU's can operate at 100% GPU usage and I'll let you know the results.

I have now set it to use at most 50% of cores, which is now keeping my overall CPU usage around 65-71% and the GU's are staying at 96-100% (Granted, they drop to 0 frequently since they're finishing work units so fast)

From initial observations, it looks like they're now finishing them in 44 seconds instead of 50-60.
The most recent 4 WU's finished in 44s, 44s, 44s, 44s.

Also I would like to note that these GPU's are NOT overclocked at all. The top card is only at 69c@75% fanspeed, and the bottom card is at 57c@70% fanspeed. From research I've done, they can easily be bumped up on air to 1125Mhz and beyond instead of 925Mhz. They have great overclocking potential, I just haven't done it yet. I want to wait a few months.

Sadly there's no official way to disable separation tasks on the CPU if you've got a GPU running. Some projects allow such settings in your account. However, you could use an app_info for the GPU tasks. This would allow to crunch 2 GPU WUs per card simulataneously, so you can get rid of the break between WUs. Improves throughput on fast GPUs significantly.

I could give you such an app_info. This would stop you from running CPU tasks for MW at all. It would probably be better to use the CPU for n-body tasks (which GPUs can't run). To do this something has to be added to the app_info, which I'm not familiar with.

Edit: on my rig I used to run 7 CPU tasks on my i7 2600K @ 4.0 GHz and 2 WUs on the GPU. CPU support was more than enough, GPU utilization 99-100%.

@Semtex: due to technical reasons (I'd say AMDs fault) the HD79x0 are running less optimized code, so there's there's still untapped potential left.

MrS

---

Scanning for our furry friends since Jan 2002

I am in the process of installing a Sapphire HD 6790 GPU on my #1 PC AMD Phenom II X6 1090T, running @ 3.8 GHz. I hope to begin crunching Milky Way tasks on Monday 26 March 2012. Can't wait.

Yea I was looking for the folder to put an app_info into but I didnt find it in my initial looking. If you can provide one to crunch multiple WU's per GPU, that would be awesome.. I noticed there's supposedly one in another thread but I wasnt sure if I should use it.

You want to unhide hidden folders and put the app_info file into your Program Data/BOINC/Projects/MilkyWay folder.

Here is he app_info file I use:


<app_info>
<app>
<name>milkyway</name>
</app>
<file_info>
<name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway</app_name>
<version_num>102</version_num>
<flops>1.0e11</flops>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>1</max_ncpus>
<plan_class>ati14ati</plan_class>
<coproc>
<type>ATI</type>
<count>0.5</count>
</coproc>
<cmdline>--gpu-target-frequency 10 --gpu-disable-checkpointing</cmdline>
<file_ref>
<file_name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>

Alright, I made a new file and named it app_info without an extension, and restarted BOINC. How do I get it to load? It still is only using 2 WU's for 2 GPu's.

C:\ProgramData\BOINC\projects\milkyway.cs.rpi.edu_milkyway is where I put it.

It needs to be called app_info.xml. Stop boinc, rename it, restart boinc.