Today was an interesting day in AMDland, first the company announced their latest GPU, the FirePro W8100 and then later in the day they announced a program where you could buy any of their latest GPUs for a whopping 50% as long as its the first one, every subsequent one will be full price. But first, you have to go through an ‘approval process’. Now, let’s get back to the new GPU AMD just announced, what is it exactly? Well, the FirePro W8100 is part of AMD’s professional line of graphics cards branded as FirePro.
So, looking at the rough specs we can see that the W8100 delivers over 2 TFLOPs of double precision, which is actually less than what Intel’s new Knight’s Landing is capable of delivering, which was announced today. It does, however, also do over 4 TFLOPs of single precision which is quite impressive since its double Nvidia’s K5000’s 2.1 TFLOPs. This GPU is effectively a professional version of AMD’s R9 290 GPU which we reviewed and found overall to be a very impressive GPU for the money, and it still is. What makes this GPU different, however is that it can drive four 4K displays simultaneously and has 8 GB of GDDR5 memory as opposed to 4 GB, making better use of the 512-bit memory bus on the Hawaii Pro GPU inside. This is, however, less than what the W9100 supports which is six 4K displays. But realistically you won’t be doing any gaming on these 4K displays so it doesn’t seem outrageous to think someone could be using 32 million pixels. AMD accomplishes this through putting four DisplayPort 1.2 connectors on the back of the card as you can see above and below.
As you can see from the above specs, AMD has decided to change the GPU’s name to an engine and say its clocked at 824 MHz, a solid 123 MHz less than the R9 290 gaming graphics card that it mimics. It does, however have double the memory of the R9 290 which is why it is capable of driving up to four 4K displays. AMD also powers it with two 6-pin power connectors, drawing 220W and supporting PCIe 3.0, everything pretty standard here. It also supports OpenCL 1.2 and already has OpenCL 2.0 support baked-in, which is good to know for anyone planning to buy a ‘future-proof’ GPU. It also supports OpenGL 4.3 and will support OpenGL 4.4, which isn’t that much of a feat as most of that support will be accomplished though a driver update. What is interesting, though, is that it supports DirectX 11.2, but AMD is making no mention of future compatibility with DirectX 12 at all, which seems a bit missing. It isn’t anything shocking since this graphics card is based on a GPU that was announced in 2013, but it is still interesting that AMD has nothing to mention there.
AMD also couldn’t help but compare themselves to Nvidia’s Quadro K5000, Nvidia’s older professional workstation GPU (as they’re currently on the K6000) so naturally, here in AMD’s comparison they basically spank Nvidia. Yes, the W8100 is $2499, which makes it more price comparable with the K5000 as opposed to the K6000 which sells for a whopping $4,999 and is more comparable with AMD’s W9100.
AMD also draws a comparison against Nvidia’s Maximus 2 development platform, which we also reviewed, as that solution is absolutely bulletproof but also incredibly expensive. Here AMD is claiming that they deliver more performance and doing it with fewer GPUs and with comparable memory. However, AMD doesn’t talk about the development scenarios that it enables or how good their professional drivers are compared to Nvidia’s. The Maximus 2 platform (and subsequent versions) are all about stability and reliability and not necessarily about performance as we learned in our review. So, until AMD can put these GPUs in our hands and show us that their GPUs and platforms are as stable as Nvidia’s in the same applications, then we’re not entirely sure that AMD can draw these comparisons. Yes, fewer GPUs will consume less power, but sometimes power isn’t as much of a concern when in professional graphics scenarios.
Last but not least, AMD’s W8100 was benchmarked in a ton of AMD-favorable benchmarks and applications (mostly OpenCL heavy) and they obviously won pretty well. However, the most interesting benchmark to me that isn’t cherry picked by AMD was their DaVinci Resolve performance benchmark showing scaling in Resolve using W8100’s. In that benchmark they show almost 100% scaling with DaVinci Resolve, which may be incredibly attractive to professionals that do lots of heavy post-processing.
Also, in regards to AMD’s 50% off promotion, there are actually only specific GPUs eligible for the promotion, including the W9100. And frankly, if you’re going to use the 50% off promotion, you might as well use it on their fastest and most expensive (and capable) professional graphics card. Other options include $800 off the MSRP of the W8000, $450 off the MSRP of the W7000, $1250 off the S9000’s MSRP and $715 off the S7000 at MSRP price. So, obviously it isn’t 50% off all professional graphic cards, but rather up to 50% off some of them.
I’m not sure why AMD is doing this, maybe to introduce people to their GPUs by getting to buy one cheaply, which isn’t a bad sales strategy. However, it may also be that they’re desperate to sell these GPUs and are cherry picking specific models and prices in order to make sure that they’re still making a profit on them.