If this is the sort of thing that winds up in large custom supercomputers then it won't matter that it uses 375W. If there is demand, Cray or someone similar will build custom racks to house and power these wild hogs.
Check out this demo using an Exxact Computing Server. It houses 8 FirePro S10000 cards (16 GPUs) in one server for more than 8 TFLOPS of real world double precision computing performance
I'm not happy with both NVidia and AMD at the moment. AMD does not offer a GCN card at the 200$ price point and NVidia doesn't look to be announcing any Kepler counterparts to the Quadro 400 and 600.
These parts are important since they are relatively inexpensive and serve as a gateway card to demonstrate what these products can do for an employer.
On the other hand, low end workstation cards have never had the performance to justify their cost. When the majority of the cost is drivers and validation they have to cut too much hardware to get the price down. They probably aren't considered very cost effective by the market making them lower volume than a card like that needs to be. I only really seen low end workstation cards in mobile workstations anymore. Still, it sucks for individual and small businesses that don't worry so much about performance, but need the functionality.
I guess they are milking Fermi for all they can as AMD does not have much competition in that space yet.
As for the S9000/S10000, until AMD has a compelling language (CUDA has huge industry adoption so far) and recognized support, they will struggle in that market.
I know very little about using GPU's for HPC (other than it is awesome and powerful), but looking at numbers, I was wondering: The HD 7970 GHz claims 1.01 TFLOPS double precision, and the S10000 offers 1.48 TFLOPS double precision, but at about 7 times the cost. Is the difference of having ECC memory worth that much, and that critical to HPC?
Ask yourself, what would happen if I tried to use the HD7970 intensively 24x7x364? How long would it last? Now, suppose you had to do a calculation that took a few weeks to complete. How confident can you be that your results are correct if you use the HD7970? What if something happened to your HD7970, a tiny hardware hiccup or small power transient or even a software glitch, such that it causes a hang or crash in your program, and you had to restart your calculation all over again when it was just a couple days from being finished? In this scenario, how much would you be willing to pay for the assurance that your graphics card is going to perform as required? Changes to the design of the card itself, as well as the extra testing each one has to go through, cost money. Sure, there is a larger profit margin, but some of that is also needed to support the customer in case there are problems.
I'm not sure I get the real advantage of this card over the K20 and K20X. K20X delivers a total of 5.26 TFLOPS peak performance using only 225w, this thing can achieve a total of 7.39 TFLOPS BUT needs a high TDP of 375w which makes K20 and K20X more power efficient (single card comparison). For a large server, this might be a good choice given the huge memory bandwidth and lower overhead saved due to lower number of cards needed to meet the target performance.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
15 Comments
Back to Article
wwwcd - Monday, November 12, 2012 - link
nice AMDmckirkus - Monday, November 12, 2012 - link
If this is the sort of thing that winds up in large custom supercomputers then it won't matter that it uses 375W. If there is demand, Cray or someone similar will build custom racks to house and power these wild hogs.fireuser - Monday, November 12, 2012 - link
Check out this demo using an Exxact Computing Server. It houses 8 FirePro S10000 cards (16 GPUs) in one server for more than 8 TFLOPS of real world double precision computing performancehttp://fireuser.com/blog/8_amd_firepro_s10000s_16_...
MarsMSJ - Monday, November 12, 2012 - link
I'm not happy with both NVidia and AMD at the moment. AMD does not offer a GCN card at the 200$ price point and NVidia doesn't look to be announcing any Kepler counterparts to the Quadro 400 and 600.
These parts are important since they are relatively inexpensive and serve as a gateway card to demonstrate what these products can do for an employer.
JPForums - Tuesday, November 13, 2012 - link
A GCN card at $200?You mean like the Radeon HD 7850 currently selling for $200 on Newegg?
http://www.newegg.com/Product/Product.aspx?Item=N8...
Yeah, wish they had one of those (in workstation form).
On the other hand, low end workstation cards have never had the performance to justify their cost.
When the majority of the cost is drivers and validation they have to cut too much hardware to get the price down.
They probably aren't considered very cost effective by the market making them lower volume than a card like that needs to be.
I only really seen low end workstation cards in mobile workstations anymore.
Still, it sucks for individual and small businesses that don't worry so much about performance, but need the functionality.
frenchy_2001 - Tuesday, November 13, 2012 - link
not sure what you are talking about as nvidia has the quadro 410 based on GK107. That's the really low one:http://www.nvidia.com/object/quadro-410-graphics-c...
They are lagging a bit in Kepler release for workstation, as only the K5000 has been announced yet (based on GK104), aside from all the mobile chips.
http://www.nvidia.com/object/workstation-solutions...
I guess they are milking Fermi for all they can as AMD does not have much competition in that space yet.
As for the S9000/S10000, until AMD has a compelling language (CUDA has huge industry adoption so far) and recognized support, they will struggle in that market.
eanazag - Monday, November 12, 2012 - link
Did AMD cover their ECC bases?I haven't heard anything. I know this has been an issue in the past that was an advantage for nVidia.
fireuser - Monday, November 12, 2012 - link
Yes, the S10000 as well as the S9000, W9000 and W8000 cards all offer EDD memory.fireuser - Monday, November 12, 2012 - link
Oops - typo:Yes, the FirePro S10000 as well as the FirePro S9000, W9000 and W8000 cards all offer ECC memory.
SpeedMan88 - Monday, November 12, 2012 - link
will it play Crysis?frogger4 - Tuesday, November 13, 2012 - link
I know very little about using GPU's for HPC (other than it is awesome and powerful), but looking at numbers, I was wondering:The HD 7970 GHz claims 1.01 TFLOPS double precision, and the S10000 offers 1.48 TFLOPS double precision, but at about 7 times the cost. Is the difference of having ECC memory worth that much, and that critical to HPC?
Ktracho - Thursday, November 15, 2012 - link
Ask yourself, what would happen if I tried to use the HD7970 intensively 24x7x364? How long would it last? Now, suppose you had to do a calculation that took a few weeks to complete. How confident can you be that your results are correct if you use the HD7970? What if something happened to your HD7970, a tiny hardware hiccup or small power transient or even a software glitch, such that it causes a hang or crash in your program, and you had to restart your calculation all over again when it was just a couple days from being finished? In this scenario, how much would you be willing to pay for the assurance that your graphics card is going to perform as required? Changes to the design of the card itself, as well as the extra testing each one has to go through, cost money. Sure, there is a larger profit margin, but some of that is also needed to support the customer in case there are problems.texasti89 - Wednesday, November 14, 2012 - link
I'm not sure I get the real advantage of this card over the K20 and K20X. K20X delivers a total of 5.26 TFLOPS peak performance using only 225w, this thing can achieve a total of 7.39 TFLOPS BUT needs a high TDP of 375w which makes K20 and K20X more power efficient (single card comparison). For a large server, this might be a good choice given the huge memory bandwidth and lower overhead saved due to lower number of cards needed to meet the target performance.
Howlingmoon - Sunday, June 16, 2013 - link
Yes...but can I Crossfire 2 of these monsters and play Crysis 3 and Metro 2033 at Super, Ultra settings on my new 4k TV at 100+ FPS?Pessimism - Monday, June 24, 2013 - link
Its over 9000.