Max PPD

[ochadd]

Seems Milkyway@Home delivers on ATI GPUs:

27k ppd for 3850
76k for 4870
151k for 4870X2

Wonder how long that will last.
Link

 

[Assimilator1]

Nice!

Got a 4830 here btw. Seems like a massive speed up!

 

[ochadd]

I'm kicking myself hard for selling a 3870 earlier this year.

 

[dajeepster]

hmmm... milkyway.. here I come

 

[Soulkeeper]

I purchased a gts 250 to run gpugrid/seti because milkyway has no linux support for the ati app. I heard there are shortages of wu's also

my 3870 has been collecting dust on the shelf now for a few months

 

[ochadd]

Originally posted by: Soulkeeper
I purchased a gts 250 to run gpugrid/seti because milkyway has no linux support for the ati app. I heard there are shortages of wu's also

my 3870 has been collecting dust on the shelf now for a few months

I imagine the lack of work units will become an increasing problem once people catch on to this. From a Youtube video I saw it seems a 4870 cranks through 8 work units every 90 seconds. Has this been around for awhile or is it just news to me?
Youtube

 

[Assimilator1]

8 every 90s???:shocked:, man that's gotta cause them problems!?

 

[ochadd]

I'm trading my card in for a 4870 crossfire setup. Giving up allot of SETI and GPUgrid points but this is too good to pass up. I haven't been able to think about much else since I saw those numbers.

 

[Philippart]

Personally I would think a lot about this decision:

with nvidia you have great clients for folding, seti, gpugrid, lattice and many projects to come, with ati you have milkyway and folding (with a miserable client) and as of now not even official boinc support. At this moment the ati gpgpu software language hasn't reached the maturity of nvidia's cuda (folding people notice it every day). I don't know about milkyway, but every other gpgpu ati software uses a whole cpu core to feed the gpu. (folding was able to prevent this by trading cpu usage with instability )

In addition I don't think this will last long since they are actively working on a milkyway gpu subproject with official clients and as far as I know credits in line with seti's cuda because so many people argued about milkyway's and aqua's credit inflation and seti is often used as guideline to the credit system.

That said it is indeed tempting provided that I'm interested in astronomics and it's a great way to speedup their research, but I wouldn't want to mess up my nice boincstats because I think the work done is not in line with the credit granted. Personally I crunch for the science and use the credits as a benchmark of the work done and incitation. Not the other way around.

Please consider this for your choice as I don't think it will last...

 

[ochadd]

You're right and from the hardware side I actually think Nvidia has bigger balls. However I'm still an AMD fanboy even though I haven't owned an AMD system in 2.5 years. I think this might be my first justifiable reason to go back to them

I'm not going to take a beating on my card and if I can't get a 4870x2 for my card +120-140 it's not going to happen.

 

[Assimilator1]

Philippart
(folding was able to prevent this by trading cpu usage with instability)

I gotta disagree with that, I've been able to reduce CPU useage by the ATI GPU client to about 5-10% with almost no problems. On the odd occasion I've found tale tale signs of a VPU recovery, but it is rare. That's with the latest core & beta dlls.
But yes the output sucks compared to Nvidia cards, but then I didn't buy this card for DC but for gaming, & a grx card should not be bought for a single DC project anyway.

 

[dajeepster]

has anyone gotten this to work?... i'm failing miserably on my 3850

 

[dajeepster]

Originally posted by: dajeepster
has anyone gotten this to work?... i'm failing miserably on my 3850

nm.... got it working.. had to rename a couple files cause i'm running vista and boinc 6.6.36