When AMD announced that it would launch its upcoming Vega GPU for professional markets before bringing it to the consumer space, would-be customers were somewhat less than thrilled. We now know that the GPU will launch for consumer products in July, with the professional Radeon Vega Frontier Edition available in June. Thanks to SabrePC, which briefly had price listings attached to the GPUs, we know the air-cooled variant will cost $ 1,199, while the water-cooled card will be $ 1,799.
That’s a $ 600 price difference for two cards that are supposedly the same in every other respect. This is assuming that SabrePC’s listing is accurate. But with the card scheduled to go on sale on June 27, and SabrePC being a primarily workstation-focused shop, the company should have some idea what the SKUs are going to look like and how they’ll be positioned.
Now, $ 1,200 and $ 1,800 might seem extremely expensive, and those prices would be for consumer cards. But AMD has been quite clear that Vega’s Frontier Edition is not aimed at consumers, even though it will run games and be fully supported in that mode. And when you consider the pricing on workstation and professional GPUs, $ ,1200 and $ 1,800 are relative steals.
Nvidia tends to command far higher prices. In the past, when AMD was chasing professional GPU market share more aggressively, the company often priced its cards well below the equivalent Nvidia counterparts. The company may be doing something similar here, banking on a Vega price/performance ratio that’s sexy enough to convince GPU programmers and professional users to take a chance on Team Red.
That said, there’s still a lot we don’t officially know about Vega. Estimated clock speed, based on its TFLOPS rating, is 1.6GHz, and AMD has promised that the consumer variant will be clocked even higher. 1.7GHz seems a reasonable target based on what AMD has said, but that’s strictly a guess on my part.
One issue I think is worth speaking to directly is whether Vega will be significantly faster than Fury X. This is a question several readers have raised in the recent past, apparently based on the idea that Vega is just another GCN tweak. While we don’t know how much faster Vega will be, it would genuinely surprise me if the card was just 10-15% percent more efficient than its predecessor. But even if it was (and I think that number is quite low), the additional clocks would still deliver an enormous boost. A theoretical 1.7GHz Vega with 10-percent-better IPC than Fury X would be 10 percent faster based on IPC and 1.62x faster based on clock. Fears that Vega won’t manage to be meaningfully faster than Fury X are overblown — we’re going to see a large gap between the two cards, even in a worst-case scenario.
Other factors, like the decrease in memory bandwidth, shouldn’t be an issue either. AMD has already said that it increased its memory bandwidth efficiency significantly with Vega, which again would be in line with how past memory bandwidth transitions have played out. It’s fairly common to see AMD and Nvidia take a brute force approach when they first introduce a new memory technology or standard, only to then walk back the bandwidth as future GPU generations become more efficient. This happened with the G80 – G92 shift for Nvidia, and we saw it in AMD’s HD 2000 family versus the later, more efficient HD 3000 series as well. All in all, the specs as listed are good — we’ll have to wait and see how they stack up next to Nvidia’s Pascal family, but the comparison should be fairly solid.