AMD Corporate Fellow Phil Rogers Leaves Company, Joins NVIDIA
by Ryan Smith on October 13, 2015 7:35 PM EST
With respect to both structure and personnel, the last few months have been busy – if not tumultuous – for AMD. The company recently reorganized itself so that their graphics division and its employees are once again a whole entity under Raja Koduri as the Radeon Technologies Group, as opposed to being dispersed among AMD’s various technical and organizational groups. Meanwhile on the personnel front, last month AMD ace CPU architect Jim Keller stepped away from the company after completing his work on Zen. As it turns out, Jim is not the only recent high-profile departure from AMD; as discovered by HardOCP today and since confirmed by AMD, AMD Corporate Fellow Phil Rogers has left the company as well.
As one of AMD’s high-ranking technology & engineering corporate fellows, Rogers’ held an important position at AMD. For the last several years, Rogers has been responsible for helping to develop the software ecosystem behind AMD’s heterogeneous computing products and the Heterogeneous System Architecture. As a result, Rogers has straddled the line as a public figure for AMD; in his position at AMD Rogers was very active on the software development and evangelism side, frequently presenting the latest HSA tech and announcements for AMD at keynotes and conferences.
Consequently, though by no means the only person working on the software side of HSA at AMD, Rogers’ role in its development is an important one. Along with serving as a corporate fellow at AMD, Rogers was also a major contributor to the HSA Foundation, helping to initially found it in 2012 and serving as the Foundation’s president until he left AMD earlier this quarter. So if there is any one person at AMD that could have been classified as the face of HSA at AMD, then Phil Rogers would have been it.
Given his position within AMD, both on HSA development and as one of a small number of technology fellows (the highest technical rank within AMD), Rogers’ departure comes as a bit of a surprise. Prior to leaving the company, Rogers’ had been with AMD (and ATI before it) for 21 years, serving as a fellow for the last 8 of those years. AMD for their part isn’t saying much on Rogers’ departure beyond confirming that he left earlier in the quarter, however it should be noted that the company is currently in its “quiet period” before their Q3 earnings release on the 15th, which typically prevents companies from discussing personnel changes such as these.
From an HSA development standpoint Rogers’ departure comes at an interesting time. On the one hand HSA is still in its infancy, with the software ecosystem still being built up and AMD just now shipping their first full HSA 1.0 capable APUs with Carrizo. On the other hand the HSA Foundation did finish the HSA 1.0 Final specification earlier this year, and some of the other foundation members have announced that they’ll have HSA-capable designs available in the near future, so the initial work on HSA is done. In which case similar to Jim Keller this may be an AMD employee leaving now that they’ve accomplished their key technical tasks.
Meanwhile of equal interest is where Rogers has landed: AMD’s arch-rival NVIDIA. According to his LinkedIn profile Phil Rogers is now NVIDIA’s “Chief Software Architect – Compute Server” a position that sounds very similar to what he was doing over at AMD. NVIDIA is not a member of the HSA Foundation, but they are currently gearing up for the launch of the Pascal GPU family, which has some features that overlap well with Phil Rogers’ expertise. Pascal’s NVLink CPU & GPU interconnect would allow tightly coupled heterogonous computing similar to what AMD has been working on, so for NVIDIA to bring over a heterogeneous compute specialist makes a great deal of sense for the company. And similarly for Rogers, in leaving AMD, NVIDIA is the most logical place for him to go.
Wrapping things up, we may yet hear a bit more about Phil Rogers’ departure from AMD on the earnings call on the 15th. Otherwise it looks like it will be AMD’s Gregory Stoner who will be stepping up to the plate to replace Rogers. Stoner is AMD’s current Senior Director of Compute Solutions Technology and long-time Vice President of the HSA Foundation, and with Rogers’ change in employment he is now the managing director of the Foundation as well.
Source: HardOCP
Facebook
Twitter
RSS
83 Comments
View All Comments
Michael Bay - Thursday, October 15, 2015 - link
Oh please, DX was and always will be primarily driven by MS, not specific GPU vendors. It`s MS who listened to developers` wishes and then went on to building a new architecture.asniper - Wednesday, October 14, 2015 - link
Not even GDDR5, AMD had hands in that as well.CiccioB - Wednesday, October 14, 2015 - link
AMD invents things that nvidia wants to copy???? This is the funniest thing I have ever read.AMD has a history of creating (copying) technology and offering it as a free gift to the market in order to try to limit its competitors gain on their own custom technology they have developed much earlier than AMD ever could thing that that could be done.
Physix vs Bullets Library (where is the latter, BTW?)
CUDA vs lots of failed language extension till OpenCL (Apple thought) came
GamesWork is much more than Hairworks or Tresshair
Games Experience vs 3rd party app to do the same
Gsync came a year before the HW on the market could support such technology (AMD just queue when other created the HW to support Adaptive Sync, which then AMD called FreeSync as their own implementation). Still Gsync works better.
New fast AA methods have always been copied by AMD.
nvidia has been useing memory compression since GDDR5 bandwidth was starting to be insufficient for high end cards. AMD has always been using bigger (and more expensive) memory controller instead. Now they have understood that memory compression helps.
GDDR5 as any other memory adoption is NOT AMD's merit at all. It's just a sort of offer own products to first use the new standard technology in order to help the real creator of this one to have better and earlier testing devices. It just means that AMD gains few months of exclusivity.
nvida has never needed to use the latest memory technology to be superior to AMD. They jumped GDDR4 all together, for example. And HBM v1, which nvidia proved to equal with GDDR5 and better memory handling, keeping manufacturing cost lower. Not even needing it to lower its already low power consumption.
nvida has always worked with a scalar architecture, while AMD insisted for 6 years with VLIW, claiming it was a good architecture for GPGPU. Then at the end AMD had to create an architecture very similar to the one nvidia was developing, improving its GPGPU capabilities but loosing lots of efficiency in 3D work, both under the point of view of power consumption and die space. This was the opposite when nvidia was scalar vs VLIW.
It was AMD claiming "small is better" when using VLIW, while slowly coming to the dimension of nvidia professional GPU die area and at the end greatly surpassing them for the same performances, clearly indicating that they were trailing nvidia on technological choices, but failing to do better than them.
What are those great invention AMD put on the table in the last 8 years that nvidia copied?
I'm speaking about technology coming from AMD R&D laboratories, not free specifications that have been (partially) implemented by others with the aim to steal some of competition revenues with custom, closed, expensive, but working (which is fundamental for market adoption!).
I'm speaking about real technology that brought advantage to AMD that nvidia had to copy in order to not be left behind.
As far as I can think, I can't remember none. But maybe it's my fault. Waiting for a clear list.
Azix - Wednesday, October 14, 2015 - link
Delusional fanboy.Physx was bought.
CUDA is just another implementation of GPGPU. Nvidia cannot lay claim to it just because they put out a proprietary version.
Gameworks is not worth mentioning. It took existing effects and blackboxed them. It also runs like crap
G-sync was most likely nvidia cutting in before Adaptive sync got approved as a standard and even if not so the means already existed. brute forcing it with an FPGA is not innovation. The hardware already existed, the issue is certification for use with PC gpus etc. http://www.vesa.org/news/vesa-adds-adaptive-sync-t... - since 2009 its been possible.
There were already tons of AA methods. Nvidias own method just used a combination of things AMD had a decade before.
Nvidia chose to use higher clocked VRAM and AMD went for higher bandwidth. Both already used compression. eg. http://www.anandtech.com/show/2231/11 They just happen to improve the compression when they can. Ignorant people like you assume nvidia is the only one doing it because a lot of noise was made about the tech when maxwell 2 launched. probably because those cards had crap bandwidth and the compression was an excuse.
Not really understanding your point regarding GDDR5. It was developed by AMD. http://www.vrworld.com/2008/11/22/100th-story-gddr...
You're just being crazy. There is no reason to go out of your way to deny everything AMD and praise everything Nvidia. Its just crazy.
What nvidia did to lower their power consumption is partly why they are getting beaten in dx12. They made consessions in other places. AMD has been putting out chips that had quite a bit more power than nvidia, but were not taken advantage of due to dx11. I don't think we are quite at the point where memory bandwidth between gddr5 and hbm would make a significant difference. Its only 512GB/s vs ~336GB/s. The Fury often matching or beating the 980ti at 4K might be the most we see from that currently, as well as maybe xfire comparisons.
At this point AMD's hardware is more forward thinking than nvidia's. I am not sure about the history of moving to VLIW but I do know nvidia took a step back by removing their hardware scheduler with kepler and maxwell (for the sake of efficiency). AMD did maintain smaller die sizes after moving away from VLIW when comparing launch years. Hawaii was much smaller than kepler, Fiji is smaller than big maxwell 2.
So you want to cut out a lot of AMDs contributions just because they made it open? It only matters if its proprietary? How do you steal revenue using free open technology? Why does it HAVE TO bring advantage to AMD? Is innovation not innovation?
I doubt anyone will give you a list. You entire list was rubbish and not something that needs a countering list.
5150Joker - Wednesday, October 14, 2015 - link
AMD still loses in the latest DX 12 benchmark, I see 980 Ti at the top, what do you see? http://www.anandtech.com/show/9659/fable-legends-d...AMD has hot large dies that are inefficient, that is why they HAD to use HBM for their Fury line, just so they could cut down on the power draw and even then it sucks down more power than Maxwell which is just pathetic. What is more pathetic is Maxwell beats fiji in DX 11 and 12.
MobiusPizza - Thursday, October 22, 2015 - link
That's only because you are looking at 4K which is bottle necked by memory amount.. HBM is limited to 4GB 1st Gen and hence you are not comparing GPU architecture itself fairly. I can argue equally that FuryX beats 980Ti at 1080p. So what's your point? It's apple to oranges.bennyg - Thursday, October 15, 2015 - link
... AMD might well be more forward thinking... but immediate past history shows Nvidia has sold many more products for many years in a row. Warm fuzzy feelings about moving the world forwards don't contribute to earnings or PE multiples as much as superior products and sales do.StevoLincolnite - Wednesday, October 14, 2015 - link
AMD holds a few x86 patents as well, which Intel needs.medi03 - Thursday, October 15, 2015 - link
AMDs 390, is faster and consumes about 50w (total system) more than slower yet similarly priced 970.And that despite so much more modest R&D budget and all the wonderful "competitive" things that nVidia does with tesselation in GameWorks' libs.
nVidia spends gazillion to develop proprietary G-Sync that adds 200$ to Monitor costs, just to be countered in a couple of month by Display Port 1.2a/FreeSync, that comes FREE OF CHARGE with most (all?) scaler chips out there. (effectively only lazy manufacturers won't enable it in their monitors)
Totally outclass, yeah.
JoeMonco - Tuesday, October 13, 2015 - link
Why bother? They can buy AMD's patents on the cheap in a couple of years.