Subject: Graphics Cards | April 5, 2016 - 02:13 AM | Tim Verry
Tagged: HPC, hbm, gpgpu, firepro s9300x2, firepro, dual fiji, deep learning, big data, amd
Earlier this month AMD launched a dual Fiji powerhouse for VR gamers it is calling the Radeon Pro Duo. Now, AMD is bringing its latest GCN architecture and HBM memory to servers with the dual GPU FirePro S9300 x2.
The new server-bound professional graphics card packs an impressive amount of computing hardware into a dual-slot card with passive cooling. The FirePro S9300 x2 combines two full Fiji GPUs clocked at 850 MHz for a total of 8,192 cores, 512 TUs, and 128 ROPs. Each GPU is paired with 4GB of non-ECC HBM memory on package with 512GB/s of memory bandwidth which AMD combines to advertise this as the first professional graphics card with 1TB/s of memory bandwidth.
Due to lower clockspeeds the S9300 x2 has less peak single precision compute performance versus the consumer Radeon Pro Duo at 13.9 TFLOPS versus 16 TFLOPs on the desktop card. Businesses will be able to cram more cards into their rack mounted servers though since they do not need to worry about mounting locations for the sealed loop water cooling of the Radeon card.
|FirePro S9300 x2||Radeon Pro Duo||R9 Fury X||FirePro S9170|
|GPU||Dual Fiji||Dual Fiji||Fiji||Hawaii|
|GPU Cores||8192 (2 x 4096)||8192 (2 x 4096)||4096||2816|
|Rated Clock||850 MHz||1050 MHz||1050 MHz||930 MHz|
|Texture Units||2 x 256||2 x 256||256||176|
|ROP Units||2 x 64||2 x 64||64||64|
|Memory||8GB (2 x 4GB)||8GB (2 x 4GB)||4GB||32GB ECC|
|Memory Clock||500 MHz||500 MHz||500 MHz||5000 MHz|
|Memory Interface||4096-bit (HBM) per GPU||4096-bit (HBM) per GPU||4096-bit (HBM)||512-bit|
|Memory Bandwidth||1TB/s (2 x 512GB/s)||1TB/s (2 x 512GB/s)||512 GB/s||320 GB/s|
|TDP||300 watts||?||275 watts||275 watts|
|Peak Compute||13.9 TFLOPS||16 TFLOPS||8.60 TFLOPS||5.24 TFLOPS|
AMD is aiming this card at datacenter and HPC users working on "big data" tasks that do not require the accuracy of double precision floating point calculations. Deep learning tasks, seismic processing, and data analytics are all examples AMD says the dual GPU card will excel at. These are all tasks that can be greatly accelerated by the massive parallel nature of a GPU but do not need to be as precise as stricter mathematics, modeling, and simulation work that depend on FP64 performance. In that respect, the FirePro S9300 x2 has only 870 GLFOPS of double precision compute performance.
Further, this card supports a GPGPU optimized Linux driver stack called GPUOpen and developers can program for it using either OpenCL (it supports OpenCL 1.2) or C++. AMD PowerTune, and the return of FP16 support are also features. AMD claims that its new dual GPU card is twice as fast as the NVIDIA Tesla M40 (1.6x the K80) and 12 times as fast as the latest Intel Xeon E5 in peak single precision floating point performance.
The double slot card is powered by two PCI-E power connectors and is rated at 300 watts. This is a bit more palatable than the triple 8-pin needed for the Radeon Pro Duo!
The FirePro S9300 x2 comes with a 3 year warranty and will be available in the second half of this year for $6000 USD. You are definitely paying a premium for the professional certifications and support. Here's hoping developers come up with some cool uses for the dual 8.9 Billion transistor GPUs and their included HBM memory!
Subject: Graphics Cards | March 14, 2016 - 07:00 PM | Ryan Shrout
Tagged: VR, radeon pro duo, radeon, Fiji, dual fiji, capsaicin, amd
It’s finally here, and AMD is ready to ship it, the much discussed and often debated dual-Fiji graphics card that the company first showed with the launch of the Fury series of Radeon cards way back in June of last year. It was unnamed then, and I started calling it the AMD Fury X2, but it seems that AMD has other plans for this massive compute powerhouse, now with a price tag of $1,499.
As part of the company’s Capsaicin event at GDC tonight, AMD showed the AMD Radeon Pro Duo, calling it the “most powerful platform for VR” among other things. The card itself is a dual-slot configuration with what appears to be a (very thick) 120mm self-contained liquid cooler, similar to the Fury X design. You’ll need three 8-pin power connectors for the Radeon Pro Duo as well, but assuming you are investing in this kind of hardware that should be no issue.
Even with the integration of HBM to help minimize the footprint of the GPU and memory system, the Radeon Pro Duo is a bit taller than the standard bracket and is more analogous the length of a standard graphics card.
AMD isn’t telling us much about performance in the early data provided, only mentioning again that the card provides 16 teraflops of compute performance. This is just about double that of the Fury X, single GPU variant released last year; clearly the benefit of water cooling the Pro Duo is that it can run at maximum clock speeds.
Probably the biggest change from what we learned about the dual-GPU card in June to today is its target market. AMD claims that the Radeon Pro Duo is “aimed at all aspects of the VR developer lifestyle: developing content more rapidly for tomorrow’s killer VR experiences while at work, and playing the latest DirectX® 12 experiences at maximum fidelity while off work.” Of course you can use this card for gaming – it will show up in your system just as any dual-GPU configuration would and can be taken advantage of at the same level owning two Fury X cards would.
The Radeon Pro Duo is cooled by a rear-mounted liquid cooler in this photo
That being said, with a price tag of $1,499, it makes very little sense for gamers to invest in this product for gaming alone. Just as we have said about the NVIDIA TITAN line of products, they are the best of the best but are priced to attract developers rather than gamers. In the past AMD had ridiculed NVIDIA for this kind of move but it seems that the math just works here – the dual-Fiji card is likely a high cost, low yield, low production part. Add to that the fact that it was originally promised in Q3 2015, and that AMD has publicly stated that its Polaris-based GPUs would be ready starting in June, and the window for a consumer variant of the Radeon Pro Duo is likely closed.
"The Radeon Pro Duo is AMD's evolution of their high-end graphics card strategy with them positioning the Radeon Pro Duo more towards a content creator audience rather than gamers. This helps justify the higher price and lower volumes as well as gives developers the frame of mind to develop for multi-GPU VR from the get go rather than as an afterthought." - Anshel Sag, Analyst at Moor Insights & Strategy
For engineers, developers, educational outlets and other professional landscapes though, the pure processing power compressed into a single board will be incredibly by useful. And of course, for those gamers crazy enough out there with the unlimited budget and the need to go against our recommendations.
Update: AMD's Capsaicin livestream included some interesting slides on the new Pro Duo GPU, including some of the capabilities and a look at the cooling system.
The industrial design has carried over from the Fury X
As we see from this slide, the Pro Duo offers 2x Fiji GPUs with 8GB of HBM, and boasts 16 TFLOPS of compute power (the AMD Nano offers 8.19 TFLOPS, so this is consistent with a dual-Nano setup).
The cooling system is again a Cooler Master design, with a separate block for each GPU. The hoses have a nice braided cover, and lead to a very thick looking radiator with a pre-attached fan.
From the look of the fan blades this looks like it's designed to move quite a bit of air, and it will need to considering a single (120 mm?) radiator is handling cooling for a pair of high-end GPUs. Temperatures and noise levels will be something to look for when we have hardware in hand.
Subject: Graphics Cards | January 25, 2016 - 11:51 AM | Ryan Shrout
Tagged: fury x2, Fiji, dual fiji, amd
Lo and behold! The dual-Fiji card that we have previous dubbed the AMD Radeon Fury X2 still lives! Based on a tweet from AMD PR dude Antal Tungler, a PC from Falcon Northwest at the VRLA convention was utilizing a dual-GPU Fiji graphics card to power some demos.
— Antal Tungler (@coloredrocks) January 23, 2016
This prototype Falcon Northwest Tiki system was housing the GPU beast but no images were shown of the interior of the system. Still, it's good to see AMD at least recognize that this piece of hardware still exists at all, since it was initially promised to the enthusiast market by "fall of 2015." Even in October we had hints that the card might be coming soon after seeing some shipping manifests leak out to the web.
Better late than never, right? One theory floating around inside the offices here is that AMD is going to release the Fury X2 along with the VR headsets coming out this spring, with hopes of making it THE VR graphics card of choice. The value of using multi-GPU for VR is interesting, with one GPU dedicated to each eye, though the pitfalls that could haunt both AMD and NVIDIA in this regard (latency, frame time consistency) make the technological capability a debate.
Subject: General Tech, Graphics Cards | June 16, 2015 - 01:34 PM | Tim Verry
Tagged: water cooling, SFF, Fiji, E3 2015, E3, dual fiji, amd
AMD revealed a new liquid cooled small form factor PC called Project Quantum during an E3 livestream today.
On the outside, an angled dual compartment aluminum case with rounded edges houses the processing hardware in the bottom and all the cooling components in the top part. AMD is using liquid cooling for the processor and graphics with the tubing running up the center column joining the two pieces together to a radiator or radiators. Red LEDs light up the center column while Radeon R9 branding sits in the bottom left corner.
While at first glance that Radeon R9 branding might be unassuming, it is actually referring to AMD's latest Fiji architecture. That's correct, Project Quantum is part of the Fiji product family and is, in fact, powered by two AMD Fiji-based graphics procesors!
Update: AMD has posted a behind-the-scenes video on the development of Project Quantum which you can watch below.
In the video, AMD reveals that they are using a modified ASRock Z97E-ITX/ac motherboard (thanks to djotter in the comments for pointing that out) which means that Project Quantum is using an Intel Haswell processor in addition to the two Fiji-based GPUs. AMD has removed all of the rear IO connectors save two USB 3.0 ports and an Ethernet jack. They have also moved the 8-pin CPU power connector to the back panel of the board next to the USB ports. My guess is that they did this for cable management and height restriction reasons within the bottom compartment. Specifically, from the CAD render shown in the video, it appears that the AMD graphics card sits horizontally on top of the motherboard which meant that at least some of the rear IO ports had to be removed or relocated.
Another bit of information from that AMD video is that Project Quantum is using what looks like an external power supply. The power brick connects to the system over a single cable to an internal board. This board provides power to a Pico PSU that is plugged into the ATX 24-pin connector on the motherboard and provides power to the AMD branded Solid State Drive (SSD) as well as the motherboard and CPU 8-pin connectors (which have both been modified to right angles for height and cable management reasons). The internal power board that connects to the socket at the back likely also powers the Radeon graphics card via PCI-E connectors, but it is difficult to tell from the photo (it is that red PCB towards the top of the photo).
Interestingly AMD has switched out the power and USB 3.0 headers with right angle models and removed the blue ASRock heatsinks covering the VRMs and PCH. AMD is instead using two large waterblocks to cool the components on the motherboard and graphics card. A large radiator and pump sit in the top compartment cooled by an 180mm Enermax Apollish fan. The 180mm radiator should result in quieter, or at least less annoying, fan noise since the large fan can spin slower while moving similar amounts of air as smaller fans paired with 120 or 140mm radiators. Using a single large radiator for both the CPU and GPU is an interesting choice here, and I think a correct one.
A rendering of the water loop layout on Project Quantum. Image from AMD with annotations by Aibohphobia.
It was actually djotter and Aibohphobia in the comments who spotted the Pico PSU and provided an example. (I did not notice that in the video initially, so thanks for pointing that out!) This power brick and tiny Pico PSU setup would certainly help to explain how AMD was able to make Project Quantum so thin (though an external PSU isn't necessarily a bad thing). The Pico PSU does suggest that the dual Fiji GPUs may be closer to lower end R9 Nanos than two high end Fury Xs (heh) or maybe some other yet unannounced cut-down Fiji chip entirely.
(End of update)
During the PC Gamer E3 Twitch stream, AMD CEO Lisa Su showed off Project Quantum, and Ken was able to snap a photo of the back panel.
Project Quantum has, from left to right, a single power input (see above), two analog audio jacks, two USB 3.0 ports, an Intel-powered Gigabit Ethernet jack, four USB 2.0 ports, and a single horizontal PCI slot. A Radeon R9 graphics card is installed in this slot and features three DisplayPort and one HDMI 1.4 video outputs. We still do not know all the specs of this card, but is is Fiji-based and supports LiquidVR along with AMD's other features including FreeSync and Frame Rate Target Control.
(End Update 9:30PM)
Beyond that, we do not know many details on Project Quantum. From the other announcements around Fiji today, particularly the R9 Nano and R9 Fury X, this little machine is going to be a powerhouse with impressive power efficiency and performance per watt – especially for its size!
Of course, pricing and availability were not discussed at the event. Stay tuned to PC Perspective as we get more details closer to its official release!