Thursday, May 7th 2015
AMD Fiji XT Reference PCB as Short as GTX 970 Reference, R9 295X2 Performance
AMD's upcoming Radeon R9 390X graphics cards will ship in two SKUs - an air-cooled one, with a moderately long reference design board (though not as long as the R9 290X), and a new Water-Cooled Edition (WCE) SKU, which will feature a very compact PCB - one that could be no bigger than that of the GeForce GTX 970 reference. This is possible because of AMD's HBM implementation. The 8 GB of memory on this card is present on the GPU package, as bare 3D-stacked DRAM dies, surrounding the GPU die, with an IHS covering everything; rather than the GPU package being surrounded by memory chips. Below is a mock-up of the card by ChipHell. It's not a picture. The radiator is off-proportions, the Radeon logo is misaligned, and the PCIe I/O is misaligned, etc. It should still give you a good idea of what the card looks like, particularly its length. Other specs on hand so far, include 4,096 GCN 1.2 stream processors, 256 TMUs, 128 ROPs, and a 4096-bit wide HBM interface, which at 1.25 GHz memory clock, will offer memory bandwidth of 640 GB/s.
While Fiji package will be bigger than that of, say, "Hawaii," overall the setup is more space-efficient, and conserves PCB real-estate. The PCB hence only has the GPU package and the VRM. AMD is doing away with the DVI connector on its reference PCB. It will only feature three DisplayPort 1.2a and one HDMI 2.0a. The WCE variant will feature a pump+block covering the GPU package, which will come factory-fitted to a 120 x 120 mm radiator. The air-cooled R9 390X will be longer, but only to house a heatsink and lateral blower. The single-GPU card could offer performance comparable to the dual-GPU R9 295X2, which is faster than the GeForce GTX TITAN-X. AMD CEO Lisa Su, speaking at the Investor Day event, in New York, on 6th May, hinted that the product could launch on the sidelines of either Computex 2015 (early-June) or E3 (mid-June).
Image Courtesy: ChipHell. Many Thanks to GhostRyder for the tip.
While Fiji package will be bigger than that of, say, "Hawaii," overall the setup is more space-efficient, and conserves PCB real-estate. The PCB hence only has the GPU package and the VRM. AMD is doing away with the DVI connector on its reference PCB. It will only feature three DisplayPort 1.2a and one HDMI 2.0a. The WCE variant will feature a pump+block covering the GPU package, which will come factory-fitted to a 120 x 120 mm radiator. The air-cooled R9 390X will be longer, but only to house a heatsink and lateral blower. The single-GPU card could offer performance comparable to the dual-GPU R9 295X2, which is faster than the GeForce GTX TITAN-X. AMD CEO Lisa Su, speaking at the Investor Day event, in New York, on 6th May, hinted that the product could launch on the sidelines of either Computex 2015 (early-June) or E3 (mid-June).
103 Comments on AMD Fiji XT Reference PCB as Short as GTX 970 Reference, R9 295X2 Performance
That being said, I'm pretty sure we see the same things, just from different viewing angles.
nVidia's pitch with the GTX Titans is an interesting one: they're trying to find a way to get people to spend more on their high-end stuff, at least for a while (until they consider it's time to roll out the GTX vanilla "equivalent"). It's a widely known fact that the high-end stuff rakes in less cash than the mainstream parts (they may get more per card sold, but they sell a lot less in the grand scheme of things). Things are quite different when it comes to professional-class cards, like the Quadros or Teslas, where they make a healthy (to say the least) profit and sell in large(r) numbers. AMD, on the other hand, stuck with the old way of doing things, including (for the most part, at least) pricing. Now, to the best of my knowledge, Radeon cards, especially the high-end stuff, have just about always been able to do the DPFP part quite well (I'd say just as well as their FirePro siblings), the difference was mainly related to software. nVidia's approach was different, they crippled the cards' DPFP crunching power to some extent, mainly via firmware, so their pro stuff wouldn't be threatened. The GTX Titans are somewhere in between.
As for custom coolers for dual-GPU cards... I'm not really sure what to say. nVidia does indeed frown upon custom stuff for their dual-GPUs, which are sometimes designed to make custom air cooling impossible by any practical means (see: GTX 295). On the other hand, there were SOME custom built models, such as the ASUS very limited editions (the Mars cards). AMD had some ASUS custom-built stuff as well (Ares) and the very brief collaboration with Palit/Gainward that resulted in a custom-built HD 4870X2. It could very well be that both manufacturers (AMD and nVidia) are actively discouraging OEMs from pumping out custom dual-GPUs, or it may be that said OEMs simply consider the investment not worth it for a low-volume (high-margin though it may be) product. Frankly, it could be both.
As for stock cards, frankly, I do tend to favor AMD's reference PCB+VRM design. They're usually top-notch, especially on the high-end (as opposed to nVidia's designs, sadly). Some of the best and most satisfying cards I had were AMD reference. And this is also why I generally look out for reference PCB+VRM when I want to buy a new AMD card, even though I do generally opt for custom coolers nowadays. With nVidia cards, I do tend to look for models with a redesigned PCB and VRM (especially the VRM). Besides, I don't mind waiting a couple of months or even longer, if need be.
So, if you would like to continue this conversation, I suggest we do it over PM. I'm not sure how much people mind all the off-topic here, and I don't want to find out.
Cheers!