An interview taken with the head of AMD’s radeon Technologies Group. Raja koduri, in which it was revealed that the impressive 4K/60FPS+DOOM demo, company showcase was merely running on the company`s brand new high end Vega 10 GPU, rather than the smaller mid range 11 chip. Leaving that, the Vega 10 powered Radeon graphic card engineering sample unit was then used in the demo later. It looked like a high end part. It was a bit long, comparing RX 480 and a boosted bold LED-lit Random symbol.
The head of technology was later pressed to leak whether the demo unit was reducing variant of Vega 10 or gloating the full Vega 10, but he would have not shifted. Ryzen and AMD Vega 10 leak, PCWorld`s Gordon Mah Ung and Brad Chacos sat down with Koduri and chatted for about 40 minutes. During the interview one very interesting question came in debate,which is this, whether Vega 10 would still “beat the pants off the GTX 1080” if the demo of the DOOM ran on OpenGL rather than of more efficient Vulkan.
In an OpenGL, AMD Vega Should Still “Beat the Pants oFF the GTX 1080”
Senior Editor,PCWorld Brad Chacos
The main demo you’ve given individuals a chance to play so far is DOOM running at 4K Ultra. I’ve played it, it runs wonderful it’s an incredible diversion. It can hit up to 85 FPS, however regularly it’s running 60-70 FPS at 4K Ultra everything turned. Which beats the heck out of the GTX 1080, yet that is running Vulkan. Would despite everything it beat the jeans of the HTX1080 in the event that it was running DX11?
Raja Koduri – Chief Architect Radeon Technologies Group, AMD
DOOM doesn’t have a DX11 path…
Brad Chacos – Senior Editor,PCWorld
Yeah, OpenGL, you got me. (laughs)
Raja Koduri – Chief Architect Radeon Technologies Group, AMD
I think it would. Since at that determination things are more GPU bound, less CPU bound. That is one element. In any case, let me not state that too unquestionably on the grounds that the DOOM engineer had done some reassure advancements like line gathering shaders and shader intrinsics and I don’t think they are accessible in OpenGL. That is not a major OpenGL issue, it’s quite recently that we didn’t put into putting every one of the augmentations again into OpenGL which we could have done. I just have a limited number of architects so they’re recently centered around Vulkan.
This is said to be an important question because of the low level APIs like the DirectX 12 and the Vulkan have shown a clear tendency to favor the AMD’s GCN structure over NVIDIA`s Maxwell and Pascal. While there are others who are highly distracted APIs such as the OpenGL and DirectX 11 which continue to favor NVIDIA’s GPUs. if Vega 10 can constant its lead over the GTX 1080 in OpenGL this would augur well for the GPU’s performance in DirectX 11 as well. The answer to the question is YES, Vega 10 would still perform better than the GTX 1080.
There was a critical admonition to Raja’s answer notwithstanding. He said that Vega 10 ought to in actuality beat the GTX 1080 running at the same 4K determination. Despite the fact that, at lower resolutions where the diversion is more CPU bound there would be to a lesser degree an execution differential between the two. Talking about low level APIs, Raja affirmed that 2017 will be a much greater year than 2016 was for DirectX 12 and Vulkan titles. The quantity of DirectX 12 and Vulkan amusement discharges this year is relied upon to be exponentially higher than a year ago. Something that Raja was exceptionally amped up for.
Potential of Vega 10 Overclocking and Clock Speed
While we go through overclocking on Vega 10 we should first take a look on the Clock speed regularly looks like in Vega 10 as we know that Vega 10 has been designed for a higher clock speed and thanks to the earlier leak it is now been confirmed that the Vega 10 has a total of 4096 GCN stream processors and 64 compute units. Adding to it, thanks to AMD`s MI25 announcement we know that the GPU is rated 12.5 FP32 TFLOPS, in its cooled configuration. 4096 GCN stream processors would need to be timed at well more than 1.5GHz to accomplish that 12.5 TFLOPS figure. The correct recurrence would need to be 1528 MHz. We know this in light of the fact that every Vega stream processor can execute two FP32 operations for each cycle.
This is the means by which the math goes : 4096 x 2 x 1.528 = 12517 GFLOPS = 12.5 TFLOPS.
Vega 10’s 1528 MHz clock speed is as of now significantly higher than the RX 480’s 1266 MHz, 20% higher to be correct. So there’s as of now a tremendous differential here in what Vega can accomplish contrasted with Polaris as far as clock paces. At the point when Raja was gotten some information about what AMD expects out of Vega 10 as far as overclocking he said that RTG designers are as yet making sense of that themselves and ought to have more data in a few weeks time.
Overclocking potential of Vega 10
When Scott wasson sat down with Keith may the technical marketing manager at RTG, he debated about Vega 1o high clock speed potential with a huge smile on his face, but neglects to give any details on the topic. Overclocking potential is one of those things that is amazingly hard to predict early on. So as we go closer to the Vega launch expect more. Since December, Vega GPU`s have been used in the public demos and AMD already has working physical graphic card samples that is shown to the public. All of which augur well for the Vega`s maturity.
Confirmed specification of Vega 10
What has been affirmed so far is that AMD is chipping away at two Vega GPUs. The main chip is the top of the line ~530mm² Vega 10 that has been utilized as a part of the majority of AMD’s demos to date, including the Star Wars Battlefront demo that was showcased back in December at the “New Horizon” squeeze occasion. The second is Vega 11, a littler mid-go chip that is required to opening in the middle of the RX 480 and Vega 10. We know next to no about Vega 11 other than it will probably highlight HBM2 innovation like it’s more seasoned sibling and dispatch at some point amidst the year. We know a lot about Vega 10 on the other hand. It is confirmed to have two HBM2 stacks with a total capacity of 16 GP in its professional MI25 configure and 8GB in its Radeon gaming configuration.
Both setups have the same 512GB/s memory data transmission figure. Regarding register the GPU has 64 NCUs, Next Generation Compute Units, comprising of an aggregate of 4096 stream processors. As far as power and cooling the gaming variation of Vega 10 is effectively cooled while the expert MI25 variation is indolent cooled. So you ought to expect gaming-centered Radeon Vega design cards to highlight much higher clock speeds than 1.5GHz. By how much precisely stays to be seen. At last, the GPU itself is appraised at around 225W in its full 64 NCU arrangement, while a whole Vega 10 fueled representation card is evaluated at <300W.