GV100 is pretty much the only option that matches all the specs and capabilities, but there’s no free lunch and they’re all still like $1000+ each
Agreed- we watch Ebay for used GV100 GPUs that go up and they are never cheap. Even the GP100 is still relatively expensive.
Tom,
If you can cool it, the P100 can be a good value (for fp64 work) because you can find them for really, really cheap if you watch for them on Ebay. Ian&Steve C. is right though, you will not have all of the performance (especially with fp32- the P100 is really, really bad). If you plan on using it for work here, the (G)P100 is not a good option, in my opinion.
the biggest reason the GV100 is holding value is the 32GB vram for AI stuff.
Tom recently bought a server chassis that’s designed for these passive data center GPUs, and even has the proper connector for them. (Data center GPUs up until very recently all used the EPS12V/CPU 8-pin connector, and that’s very standard in the enterprise/server space, but different than the VGA/PCIe connectors we use on consumer cards)
but he’s trying to go against the grain and not use the loud server fans because he wants it to be quieter. IMO, it’s not going to work well without the loud fans either way.
Hi all. Re message that GeForce MX550 not working back on 3 July. Thanks to GWGeorge//Tom M//SanFernandoValley//Klik//IanStevec/// and all. The MX550 is working. So sorry for failure to respond to all of you. I had to deal with severe medical problems and my five ponies in here were basically on auto-pilot. I tried to set up new acct to test and mangled that....so I tried processing suggestion from Tom M and KliK to try BRP7 only and it worked! So to anyone else with an MX550 not being used, there is hope. Unfortunately, production on this machine does not warrant using this one for for Ein and MW at the same time so I will alternate between them as needed.
Hi all. Re message that GeForce MX550 not working back on 3 July. Thanks to GWGeorge//Tom M//SanFernandoValley//Klik//IanStevec/// and all. The MX550 is working. So sorry for failure to respond to all of you. I had to deal with severe medical problems and my five ponies in here were basically on auto-pilot. I tried to set up new acct to test and mangled that....so I tried processing suggestion from Tom M and KliK to try BRP7 only and it worked! So to anyone else with an MX550 not being used, there is hope. Unfortunately, production on this machine does not warrant using this one for for Ein and MW at the same time so I will alternate between them as needed.
MW no longer has any gpu tasks so you can keep it here on Einstein full time.
Thanks Mikey. Understand, but I am running only two projects-Milky and Ein. I have tried to balance out all five hosts running both--CPU on all for Milky and GPU on all for Ein and I end up killing my production for both Mil and Ein. Am going back to Ein only on four hosts and Milky on one host That should keep my head above water for Ein. I think in my case, I just need to save my pennies and get a host with a bunch of CPUs/cores dedicated for Milky...
Mikey, more specific to your comment, yes, I can leave the MX550 host on EIN only, so that is a good suggestion. I have had a problem balancing out usage of my hosts with Milky going CPU only and Ein doing GPU. Perhaps my problem involves that Ein GPU tasks also use at least some CPU time?? My overall performance took a big hit with Milky/CPU and Ein/GPU conflicts.
Ian&Steve C. wrote: GV100 is
)
Agreed- we watch Ebay for used GV100 GPUs that go up and they are never cheap. Even the GP100 is still relatively expensive.
Tom,
If you can cool it, the P100 can be a good value (for fp64 work) because you can find them for really, really cheap if you watch for them on Ebay. Ian&Steve C. is right though, you will not have all of the performance (especially with fp32- the P100 is really, really bad). If you plan on using it for work here, the (G)P100 is not a good option, in my opinion.
the biggest reason the GV100
)
the biggest reason the GV100 is holding value is the 32GB vram for AI stuff.
Tom recently bought a server chassis that’s designed for these passive data center GPUs, and even has the proper connector for them. (Data center GPUs up until very recently all used the EPS12V/CPU 8-pin connector, and that’s very standard in the enterprise/server space, but different than the VGA/PCIe connectors we use on consumer cards)
but he’s trying to go against the grain and not use the loud server fans because he wants it to be quieter. IMO, it’s not going to work well without the loud fans either way.
_________________________________________________________________________
Tom M wrote: I am very
)
I guess the EVGA Powerlink adapters won't work because they are made for PCIe power pinouts,
Don't know what your definition of 'thin' is though. The Powerlink adapter is only about 3/4" high.
based on the pics of the
)
based on the pics of the chassis and how the GPUs mount, there's almost no room for top mounted cables. certainly not 3/4". maybe 1/8-1/4" if that.
Tom could verify this himself by just installing one of the titan Vs into the chassis and looking at the clearance without cables.
_________________________________________________________________________
Ian&Steve C. wrote: based on
)
What! And ruin my anticipation of failure?
No clearance. Sigh.
I will start packing it up tomorrow for a return on Monday or something.
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor)
Tom M wrote: I am very
)
I know K5000 & K6000 has that, but also M5000 has it (as I do own it). Now, I do work with RTX 4000 card currently, which also has it!
Also, note: almost all TESLA has it & almost all Quadro / RTX has it (not GeForce RTX)!
Had a same problem with C30 workstation, but now work with D30 so plenty of room here! So that is why I have that experiance. ????
non-profit org. Play4Life in Zagreb, Croatia, EU
Hi all. Re message that
)
Hi all. Re message that GeForce MX550 not working back on 3 July. Thanks to GWGeorge//Tom M//SanFernandoValley//Klik//IanStevec/// and all. The MX550 is working. So sorry for failure to respond to all of you. I had to deal with severe medical problems and my five ponies in here were basically on auto-pilot. I tried to set up new acct to test and mangled that....so I tried processing suggestion from Tom M and KliK to try BRP7 only and it worked! So to anyone else with an MX550 not being used, there is hope. Unfortunately, production on this machine does not warrant using this one for for Ein and MW at the same time so I will alternate between them as needed.
James Bradshaw wrote: Hi
)
MW no longer has any gpu tasks so you can keep it here on Einstein full time.
Thanks Mikey. Understand,
)
Thanks Mikey. Understand, but I am running only two projects-Milky and Ein. I have tried to balance out all five hosts running both--CPU on all for Milky and GPU on all for Ein and I end up killing my production for both Mil and Ein. Am going back to Ein only on four hosts and Milky on one host That should keep my head above water for Ein. I think in my case, I just need to save my pennies and get a host with a bunch of CPUs/cores dedicated for Milky...
Mikey, more specific to your
)
Mikey, more specific to your comment, yes, I can leave the MX550 host on EIN only, so that is a good suggestion. I have had a problem balancing out usage of my hosts with Milky going CPU only and Ein doing GPU. Perhaps my problem involves that Ein GPU tasks also use at least some CPU time?? My overall performance took a big hit with Milky/CPU and Ein/GPU conflicts.