AMD's Latest Server Compute GPU Packs In 32GB of Memory
Deathspawner writes: Following-up on the release of 12GB and 16GB FirePro compute cards last fall, AMD has just announced a brand-new top-end: the 32GB FirePro S9170. Targeted at DGEMM computation, the S9170 sets a new record for GPU memory on a single card, and does so without a dual-GPU design. Architecturally, the S9170 is similar to the S9150, but is clocked a bit faster, and is set to cost about the same as well, at between $3,000~$4,000. While AMD's recent desktop Radeon launch might have left a bit to be desired, the company has proven with its S9170 that it's still able to push boundaries.
This will cut my rendering time of Hentai down dramatically.
1) You're old... We get it!
2) 2013 called, they want their fad back
3) "Render" Hentai? Unless you count encoding DVD rips in some lossy codec to be "rendering": it really won't. Do you have some fucked up 3D animation hentai running on OpenGL or DirectX?
Starting at about the time that the "read the X comments" link was removed, I noticed a big drop in the number and quality of comments and moderation. Am I the only one?
Also, what happened to polls? I've only seen one since they removed the sidebar. Have those been killed off?
640kili-Terabytes should be good. That way we can buffer over a full year of 8k video.
If something is so important that you feel the need to post it on the internet... It probably isn't that important.
Excuse me. For what are these used?
So how fast does it run SETI@home?
D = Double precision (as opposed to S = Single precision, C - Complex single precision, or Z - Complex Double precision)
GE = GEneral, as opposed to HErmetian for example.
M = Matrix
M = Multiplication
-- Each tock of the Planck clock is a new world and here we are still life. --
Putting 32GB of ram on a card with a 512bit memory bus is impressive (that's a lot of traces!) regardless of the fact that it isn't an APU. It is interesting to see this dual direction from AMD however. While they're releasing HBM and the associated "Fury" products the hard 4GB limit is obviously impacting competition and makes the newly released product non-competitive. It does seem like once they can improve the memory limit and get the cost down, that type of product would be a great addition to an APU, finally granting near-competition to the discrete options. Question is, do we wont to spend several hundred dollars for a part that likely won't be compatible with a new motherboard when the next improvement comes along.
I'm confused. How would a GPU with 32GB of "integrated memory" be news, but a GPU with 32GB of [non-integrated*] memory is not news? I'm not sure what you mean when you say "integrated memory". This is not the graphics half of an APU, it is a discrete card, and nowhere do the summary or the press release state otherwise. The term "compute GPU" just means it's targeted at computing workloads, not graphics workloads.
What exactly is your complaint?
* Not even sure what this means, but you seem to be contrasting "integrated memory" and "memory".
My suspicion is the GP meant "on-die" not "integrated" Still a silly post though, for one thing I dont think, as you pointed out, that they know what a compute GPU is
"goodbye and hello, as always" ~Prince Corwin, from Zelazny's Amber series
Like every other Hawaii based professional card, it does.
This is a discrete card with 1 GPU and non-integrated memory. Not an APU. This is designed for rendering and other parallel compute tasks. Go google.. do you see any single GPU card that addresses this much RAM? I do not.
Silence is a state of mime.
Is it still a "Graphics Processing Unit", if it does not even offer any way to connect a display to it?
Well, maybe, it just means "Ginormous" now...
In Soviet Washington the swamp drains you.
The real question here is: when did "compute" go from being a verb to an adjective?