HD 6950 2GB still working for Einstein@home?
23 Aug 2022 10:09:07 UTC
Topic 228011
(moderation:
I have the possibility to get one Radeon HD 6950 with 2GB memory on it for free to replace my even more ancient HD 3850, which is just able to do do Moo! Wrapper als the last project offering a CAL app, but which I don't consider very useful for the science (if at all).
I've seen many of those GPUs doing pretty well on Milkyway, even when feeded by a way slower CPU than my C2D. Well, not compared to Tahiti or VII chips, but the DP performance is still comparable or even higher than on many modern cards (the RX 5700 XT for example has same DP FLOPS according to this list).
Here at Einstein the top computers list is just 50 computers long, so I couldn't find any of those. I know however, that Einstein requires a lot more memory, hence the question.
Like I said, I can get the GPU for free (but I'll have to replace one of the fans, not an issue for me) and unlike a Radeon VII or some more modern high-end cards with 300+ Watts, my PSU can handle it easily. Just an easy temporary upgrade for the next 3 years until Win10 runs out of support, than I will decide what to do with this computer (if it survives until than), there's pretty much nothing in it, that's compatible with Win11. Main project for the card is going to be Milkyway, but I'd like to know where else I can go with it (except Moo! or IMHO complete nonsense like Collatz), before I start changing a perfectly running system.
WUProp@home does not have much useful information about the HD 6000 series, there seems to be however one HD 6570 doing Gamma-ray pulsar binary search #1 on GPUs, so I think HD 6950 should be even a lot better with it's DP capability?
.
I think I remember E@H having
)
I think I remember E@H having trouble with low ram gpu's. And I remember some kind of fix being added so the server would not send some applications or data? to low ram cards.
I am sorry. I can't remember any details.
I have trouble sitting on my hands so of course I would make the change.
How much trouble is it to make the change? And can you revert?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:I have trouble
)
Thanks for the reply.
Well, main issue is the broken fan, that I'd have to replace, the old one is good enough for "office stuff", but not for crunching (one blade missing resulting vibrations at higher rpms), swapping the card isn't much work and hopefully won't mess up the driver for the HD3850 which came with Windows.
There is one person before me in the queue for the free card, who hasn't decide yet and the card will be available in a week or two, so I thought I'll use the time for asking questions, so I know where to start when/if I get it and if it's worth it to fix the card. I think it should be at least for Milkyway, but they seem to have lots of server issues recently, so not sure how much work I can get from there.
.
i believe you could still use
)
i believe you could still use that GPU on the FGRPB1G application. it uses less than 1GB memory. it wont be fast though.
you could probably even use it on the new BRP7 tasks, but it would take absolutely forever to process lol.
DP doesnt matter much for Einstein. only a small portion of the computation uses that. the rest is SP/FP32
Milkyway is probably the best use for it.
_________________________________________________________________________
Link wrote: I have the
)
So how did this end up?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote: Link wrote: I
)
Didn't get it, the person who answered before me took it.
.
Link wrote: Tom M
)
Rats!
There is a p104-100 on eBay for $150
It doesn't have video ports so if you have enough slots...
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote:Rats!There is
)
the P104-100 is basically a GTX1070 with GDDR5X. and only 4GB of it.
(vs. the 8GB GDDR5 on the GTX1070)
(vs. the 8GB GDDR5X + 640 more cores on the GTX1080)
I wouldnt bother with a mining card at those prices. sub $100 would be more appealing given the compromises you have to make (less VRAM, no display, no support, no warranty, etc)
_________________________________________________________________________
I saw today on TEchSpot.com
)
I saw today on TEchSpot.com that "You can currently find an Arc A380 for about $140. As for the A-series, the cheapest RTX 3060 on Newegg is $349, while the lowest priced RTX 3060 Ti is $449, so Intel could be looking to price the A750 and A770 lower than their Nvidia competitors. We should find out for sure any day now."
[url]https://www.techspot.com/news/95920-intel-reveals-full-spec-list-four-arc-alchemist.html[/url[
Has anyone seen an Arc Intel gpu for sale at those, or any prices? And if so where? I think they would they work for crunching as the first person who commented wrote "Just for comparison, Intel's XMX Engines are equivalent to Nvidia's Tensor Cores (I.e. they're FP16/INT8,4,2 matrix FMA units). Each XMX Engine can do 128 FP16 ops, 256 INT8 ops, or 512 INT4,2 ops per clock; each Tensor Core (in the consumer Ampere architecture) does the same number of ops per clock as the XMX Engines, although they support sparse matrix ops for a higher throughput.
The Xe Cores are equivalent to Nvidia's SMs -- one Xe core has 16 Vector and 16 XMX Engines, with the former giving up to 128 FP32 ops per clock per Xe Core. Once again, that's the same as Nvidia's SMs. So the A770 has, for its given clock speed, a peak theoretical FP32 throughput of 17.2 TFLOPS and 34.4 for FP16. An RTX 3060 is 12.7 TFLOPS for both; a 3070 is 20.4 TFLOPS.
AMD's RDNA 2 Compute Units do 128 FP32, 256 FP16 vector ops per clock, and 512 INT8, 1024 INT4 opc for matrix stuff (no separate units -- the shader cores do everything). So a Radeon RX 6800, for example, peaks at 16.1 TFLOPS for FP32 and 32.3 TFLOPS for FP16.
With 16MB of L2 cache as well, the A770 looks great for compute work, even if it ends up sucking for gaming."
I have a couple a AMD 5870
)
I have a couple a AMD 5870 gpu's laying around, I think they only have 2gb of onboard memory on them...do you think they would still crunch tasks here? I'm pretty sure they will work at MilkyWay and maybe the TF tasks at SRBase and they used to work at Collatz but it's still down.
Tom M wrote: There is a
)
No more ports and definitely no power in the PSU for two cards. If it could deliver about 100W more, I might have bougt a Tahiti based card and crunch Milkyway with it, but I don't feel like I want to invest too much in this old C2D considering that it probably won't run Win11. Not to mention, what's going on here in Europe with electricity costs, that's another thing that might make crunching nearly impossible in the future.
This was just an idea in case I could get that card for free, if I should invest $150, I'd rather add another $150 or even more and replace the thing with something more future proof and energy efficient.
.