Print 40 comment(s) - last by MAIA.. on Jun 17 at 9:10 PM

FireStream 9250 delivers eight gigaflops-per-watt performance

AMD announced today that its latest stream processor -- the FireStream 9250 -- offers record setting performance. AMD says that the FireStream 9250 is optimized for high-performance computing, mainstream and consumer applications.

AMD says that its FireStream 9250 has broken the one teraflop barrier for single precision performance. The card itself is a single-slot design and consumes less than 150W of power. The 9250 has performance per watt of up to eight gigaflops per watt.

The FireStream 9250 promises much faster data processing for critical workloads like financial analysis or seismic processing than with the CPU alone. According to AMD, developers have reported up to a 55x performance increase on financial analysis as compared to processing on the CPU alone when using its products.

The 9250 has second-generation double-precision floating point hardware that delivers over 200 gigaflops. The 9250 has built on the capabilities of the FireStream 9170, which according to AMD was the industries first GP-GPU. Memory for the FireStream 9250 is 1GB of GDDR3. AMD also provides an AMD Stream SDK to help developers take advantage of the processing power of its FireStream products.

 “An open industry standard programming specification will help drive broad-based support for stream computing technology in mainstream applications," said Rick Bergman, senior vice president and general manager, Graphics Product Group, AMD. "We believe that OpenCL is a step in the right direction and we fully support this effort. AMD intends to ensure that the AMD Stream SDK rapidly evolves to comply with open industry standards as they emerge."

The AMD FireStream 9250 will be available in Q3 2008 for $999. The AMD FireStream 9170 retails for $1,999. AMD’s main rival, NVIDIA also has its own stream processing initiative with the product line called Tesla.

Comments     Threshold

This article is over a month old, voting and posting comments is disabled

By TechLuster on 6/16/2008 1:47:26 PM , Rating: 2
Based on the quoted flops figure, if this is a single-chip product then it's definitely RV770 based. Since it uses a single-slot cooler, this is quite likely. I'm guessing AMD's trying to steal a little of Nvidia's GT200 spotlight today, with a next-gen announcement of their own!

RE: RV770
By Warren21 on 6/16/2008 2:14:38 PM , Rating: 2
You are correct no doubt. This is just a rebadged HD 4850.

RE: RV770
By BruceLeet on 6/16/2008 4:49:38 PM , Rating: 4
Again AMD is the FIRST to do achieve something, first to 1 ghz, first with on die memory controller, first 64-bit processor, now another first. Well done AMD :)

Smiley face cause they are raising the bar, which is good for US!

RE: RV770
By ecktt on 6/16/08, Rating: 0
RE: RV770
By KernD on 6/16/08, Rating: -1
RE: RV770
By Warren21 on 6/16/2008 10:55:02 PM , Rating: 2
I am pretty sure the GTX 280 is 986 GFLOPS, if you want to be really anal.

RE: RV770
By Warren21 on 6/16/2008 10:55:37 PM , Rating: 2
Also, I think the HD 3870 X2 was 1TFLOP.

RE: RV770
By MAIA on 6/17/2008 10:35:55 AM , Rating: 2
Yeah the HD 3870 X2 was 1 teraflop. But it's not a GP-GPU, which is what i think we're discussing here ...

RE: RV770
By MAIA on 6/17/2008 10:22:53 AM , Rating: 2
AMD is just bragging instead of doing what it should, giving NVidia some competition at the top end of graphic.

This idea is rather pathetic. Quite the opposite: why employing resources to fight in a niche market in which revenues and sales are laughable ? It's ok to keep any company pride on the rise, but really nowadays it makes no sense anymore to invest in such a market like nvidia does.

ATI/AMD is going diverse with its production instead of creating monolithic products which have insane production costs. I mean, for what ? 10-15 of market share ? LOL They can get greater market share margins somewhere else much easier than in the enthusiast market.

I remember the NVidia architecture is better at being close to it's spec in real world usage.


I don't understand this sentence, but anyway. The specs say 945 (if I'm not mistaken) Gflops for the 280. That's not 1 teraflop, period.

All the rest is wishful thinking ...

RE: RV770
By Zoomer on 6/16/2008 7:17:54 PM , Rating: 2
The P3 1Ghz had a serious issue, leading to it being withdrawn from the market.

Your call on 1st to 1Ghz.

RE: RV770
By PB PM on 6/16/2008 8:26:38 PM , Rating: 3
They were 1st to 1Ghz, IBM/Motorola's PPC chips didn't reach 1Ghz until a year after Intel, let alone AMD.

RE: RV770
By BruceLeet on 6/17/2008 2:14:15 PM , Rating: 2
I correct myself

March 8th 2000

Intel on Wednesday broke the gigahertz barrier with a 1GHz, or 1,000MHz, Pentium III chip. What a difference a couple of days makes -- AMD announced its 1GHz Athlon processor on Monday.

Further reading the article, I found some pretty interesting reports.

Dell will begin taking orders on Wednesday for its Special Edition Dell Dimension, a high-end PC that it will sell in limited quantities with the Intel 1GHz chip.

"We are going to have a (Dimension) product," a Dell spokesman said. "However, we don't have a date yet." While dates are not final, it is likely that Dell will have its 1GHz Pentium III-based Dimension PC ready by the end of the month as well.

Intel product rebates incentive? its possible...

"We're supporting the 1GHz Pentium III," said a Hewlett-Packard spokeswoman


The 1GHz Pentium III processor will cost $990 (£613) in 1,000-unit quantities, making it significantly less expensive than AMD's $1,299 (£805) 1GHz Athlon chip. The new Pentium III, which uses a 133MHz system bus, is available now in limited quantities, Intel officials said.

Since when has Intel ever had a lower priced product than AMD? Rebates? Also why would you sell a product you have limited quantities of for so little, unless they're ONLY buying YOUR product.

RE: RV770
By MAIA on 6/17/2008 10:04:14 AM , Rating: 3
Oh! ok ...
So you don't believe yet you throw doubt instead of searching.

Really, it's very easy. Just go to and insert the right query (that is if you're not affraid to write:) like : amd first to 1Ghz

RE: RV770
By MAIA on 6/17/2008 10:07:54 AM , Rating: 2
... and btw. Of course he's refferring to x86 and he's right about everything. That's why he's using the word "US" . I don't see that plenty of people using risc cpu's or powerpc for general purposes.

RE: RV770
By theapparition on 6/17/2008 2:21:40 PM , Rating: 2
He would still be wrong. Intel had a IMC on verions of the 486. It was not received well in the market and was discontinued.

RE: RV770
By MAIA on 6/17/2008 9:10:12 PM , Rating: 2
If you're talking about the DRAM controller, by no means it can be compared to the complexity and functionality of a real IMC like AMD's implementation. It's like saying a calculator is a computer.

Btw, Timna and Banias also share a similar controller.

RE: RV770
By winterspan on 6/16/08, Rating: 0
RE: RV770
By ChronoReverse on 6/17/2008 12:33:20 AM , Rating: 2
Which delivers 90GFLOPS at the moment. Maybe when their drivers mature? The previous G80 and G92 generation couldn't even do DP.

RE: RV770
By djkrypplephite on 6/17/2008 6:22:18 AM , Rating: 2
They actually were not the first with the on-die memory controller. Intel wasn't either, but Intel actually had it before AMD did in some experimental products back in 2000. I believe it was the Alpha chip perhaps, but then again that's all an Opteron is.

They also didn't have the first 64-bit processor, but perhaps you mean x86 only.

RE: RV770
By stryfe on 6/17/2008 5:20:23 PM , Rating: 2
The credit for first 64bit processor goes to Digital Equipment Corp. (DEC) for the Alpha, released Nov '92.

By vapore0n on 6/16/08, Rating: 0
RE: benchmarks!!
By amanojaku on 6/16/2008 1:54:24 PM , Rating: 4
No, because it's a GPU derivative designed for calculations. Sure, graphics require calculations, but this doesn't do the full range. Notice the lack of a video output.

RE: benchmarks!!
By Mitch101 on 6/16/2008 2:42:03 PM , Rating: 2
Would love to see some Anti-Virus companies start utilizing GPU's. Would have to find someone who makes a 1U card though.

RE: benchmarks!!
By tuteja1986 on 6/16/2008 2:47:33 PM , Rating: 2
RE: benchmarks!!
By Mitch101 on 6/16/2008 3:27:21 PM , Rating: 2
OK I would love to see MORE Anti-Virus companies start utilizing GPU's.

RE: benchmarks!!
By FITCamaro on 6/16/2008 3:41:58 PM , Rating: 1
Where's it say it uses GPUs?

RE: benchmarks!!
By 4wardtristan on 6/16/08, Rating: 0
RE: benchmarks!!
By MAIA on 6/17/2008 10:45:20 AM , Rating: 2
errrmmmm , this is rather an old one which some insist on using ... Many people are tired of this "joke"

Supercomputer on a budget?
By Klober on 6/16/2008 4:04:25 PM , Rating: 2
The researchers used PetaVision to set a processing record with Roadrunner, spinning up to an astonishing 1.144 petaflop/s.

So, pick up ~1,500-2,000 of these things, network them together (with the required hardware for the PCs to put them in of course), and have almost equivalent performance to Roadrunner? Or would scaling deficiencies push this back to a much higher number of FireStream cards? Could be an interesting experiment once we've gotten more efficient and experienced at programming for OpenCL.

RE: Supercomputer on a budget?
By Zoomer on 6/16/2008 7:22:55 PM , Rating: 4
In other news, AMD achieves Q2 profitability with Klober's $4 M purchase.

Good luck finding a bank to, erm, bankroll you on that.

RE: Supercomputer on a budget?
By Goty on 6/16/2008 8:39:41 PM , Rating: 2
$4 Million isn't a bad price for a supercomputer. Hell, one of my old professors put in a $1 Million grant proposal to build a cluster out of NVIDIA Tesla D870s just this last year to run galactic dynamics simulations.

RE: Supercomputer on a budget?
By emboss on 6/16/2008 8:42:11 PM , Rating: 2
Not quite ... Roadrunner does 1.4 pflops DOUBLE precision. This card does 1 tflops SINGLE precision. It's double precision throughput is about 10% of its single precision throughput (it combines multiple single-precision units together to act as a single double precision unit, which NV has finally done with the G200) so you'd need probably 10,000-plus cards. Then you need to take into account that GPU-based clusters don't scale well because of latency: GPUs have about a 300 us latency just to get out of the GPU. In Roadrunner, it's 1.5 us, and that's already hurting them in some situations.

GPUs are great for HPC, as long as your computation doesn't have to leave the GPU (or is VERY latency tolerant). They're still very limited in which problems they can solve.

I'm in the process of setting up a home cluster, and at the moment I have four nodes, each with a Phenom 9550, hooked together with a massively overkill Quadrics (QSNet-1) switch. So that's about 70 gflops theoretical single-precision for the cluster. I've also got my main development machine, a Q6600 with an 8800 GTX. So 500 gflops and change single precision there. Many of the things I write run BETTER on the cluster with only a 7th of the theoretical grunt, and not for lack of effort in trying to get the GPU to perform.

I'm all for doing HPC on a GPU, but the reality is the required interconnects are quite different. Graphics doesn't care about latency, and it makes sense to trade off latency to increase throughput. HPC is the opposite. And graphics performance is much more important design target for GPUS than HPC ...

Yes, we need it. Renderfarm on a Chip.
By greylica on 6/16/2008 4:01:17 PM , Rating: 2
Yes, the 3D artists are awaiting for this, but...
Will CL Language become open so we can purchase it and use the ENTIRE CHIP as a render farm for Blender or any other free software as well ?
Will finnaly Radeon XXXX became driver opensourced ?

We are still waiting for your promises AMD...
Still waiting and Using MacIntel to render.

RE: Yes, we need it. Renderfarm on a Chip.
By chaos386 on 6/17/2008 3:24:05 AM , Rating: 3
AMD already come through on opening their hardware specs months ago:

The Stream SDK is fairly open as well:

By chaos386 on 6/17/2008 3:25:06 AM , Rating: 3
*came* through, *came* through. >_<

Aw man...
By Shadow Conception on 6/16/2008 2:07:29 PM , Rating: 5
I saw "1 teraflop" and "9250" in the same title and thought I might see some use out of my old beloved PCI video card... and then I see "Firestream".

Damn you AMD, damn you to hell!

RE: Aw man...
By lagitup on 6/16/2008 8:11:24 PM , Rating: 2
I'm sure you could *try* softmodding your old radeon 9250 into one of these...

By Sulphademus on 6/16/2008 2:25:09 PM , Rating: 5
"The 9250 has performance per watt of up to eight gigaflops per watt."

There is no such thing as a "Teraflop"
By MC17 on 6/16/2008 6:28:25 PM , Rating: 2
However, the 1 teraflops barrier has now been broken.

By Kougar on 6/17/2008 3:20:33 PM , Rating: 2
So does this actually surprise anyone?

GTX 280

Single Precision: 933MFLOPS
Double Precision: 116MFLOPS (Not sure on this number, corrections appreciated)


Single Precision: 1TFLOP
Double Precision: >200MFLOPs

I was expecting RV770 to underperform against the GTX 280 considering it costs 53% less MSRP. Anyone that has seen GTX 280 cards Folding@Home will know what I am getting at here, if RV770 can outfold GTX 280 it will be quite the coup d'état.

"We’re Apple. We don’t wear suits. We don’t even own suits." -- Apple CEO Steve Jobs
Related Articles

Copyright 2016 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki