AMD to open source Micro Engine Scheduler firmware for Radeon GPUs
(2024/04/05)
- Reference: 1712336412
- News link: https://www.theregister.co.uk/2024/04/05/amd_mes_open_source/
- Source link:
AMD plans to document and open source its Micro Engine Scheduler (MES) firmware for GPUs, giving users more control over Radeon graphics cards.
It's part of a larger effort [1]AMD confirmed earlier this week about making its GPUs more open source at both a software level in respect to the ROCm stack for GPU programming and a hardware level. Details were scarce with this initial announcement, and the only concrete thing it introduced was a GitHub tracker.
However, yesterday [2]AMD divulged more details , specifying that one of the things it would be making open source was the MES firmware for Radeon GPUs. AMD says it will be publishing documentation for MES around the end of May, and will then release the source code some time afterward.
[3]
For one George Hotz and his startup, Tiny Corp, this is great news. Throughout March, Hotz had agitated for AMD to make MES open source in order to fix issues he was experiencing with his [4]RX 7900 XTX-powered AI server box . He had talked several times to AMD representatives, and even the company's CEO, Lisa Su.
[5]Standardization could open door to third-party chiplets in AMD designs
[6]Kaby Lake-G chip back from the grave, now on modest firewall-router-NAS mobo
[7]TSMC's 3nm node powers up, setting stage for tech giants' next-gen chips
[8]Tiny Corp launches Nvidia-powered AI computer because 'it just works'
While Hotz was initially "70 percent confident" that MES would be made open source, after his conversation with Su in late March he figured it wasn't going to happen after all. This would have killed the AMD-powered TinyBox, which uses six RX 7900 XTXs, but the project was saved thanks to the discovery of a single "umr" repository that made the box work well enough.
Still, Tiny Corp had reluctantly decided to launch an RTX 4090-powered TinyBox anyway since it offered a solution that [9]"just works." An Intel version might also be in the pipeline, though the performance of Intel's best GPU, the Arc A770, would be a significant concern since it comes nowhere near either the RX 7900 XTX and RTX 4090.
[10]
Yet Hotz's 70 percent confidence level was right in the end as AMD is seemingly doing exactly what he was asking for. Though making the MES firmware open source is a strange conclusion to the saga, and it's unclear why AMD seems to have changed course now.
"Spoke with them for months and months and got nothing, despite the public engagement you saw. I think it was treated as a PR problem, not a GPU problem," [11]Hotz speculated on X before AMD announced its open source efforts.
[12]
Hotz has not yet responded to AMD's announcement, and we've asked him and AMD to comment. ®
Get our [13]Tech Resources
[1] https://twitter.com/amdradeon/status/1775261152987271614
[2] https://twitter.com/amdradeon/status/1775999856420536532
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[5] https://www.theregister.com/2024/03/27/amd_chiplets_future/
[6] https://www.theregister.com/2024/03/27/failed_intel_processor_nas/
[7] https://www.theregister.com/2024/03/27/tsmc_3nm_revenue_growth/
[8] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[9] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[11] https://twitter.com/realGeorgeHotz/status/1774533299559629033
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
It's part of a larger effort [1]AMD confirmed earlier this week about making its GPUs more open source at both a software level in respect to the ROCm stack for GPU programming and a hardware level. Details were scarce with this initial announcement, and the only concrete thing it introduced was a GitHub tracker.
However, yesterday [2]AMD divulged more details , specifying that one of the things it would be making open source was the MES firmware for Radeon GPUs. AMD says it will be publishing documentation for MES around the end of May, and will then release the source code some time afterward.
[3]
For one George Hotz and his startup, Tiny Corp, this is great news. Throughout March, Hotz had agitated for AMD to make MES open source in order to fix issues he was experiencing with his [4]RX 7900 XTX-powered AI server box . He had talked several times to AMD representatives, and even the company's CEO, Lisa Su.
[5]Standardization could open door to third-party chiplets in AMD designs
[6]Kaby Lake-G chip back from the grave, now on modest firewall-router-NAS mobo
[7]TSMC's 3nm node powers up, setting stage for tech giants' next-gen chips
[8]Tiny Corp launches Nvidia-powered AI computer because 'it just works'
While Hotz was initially "70 percent confident" that MES would be made open source, after his conversation with Su in late March he figured it wasn't going to happen after all. This would have killed the AMD-powered TinyBox, which uses six RX 7900 XTXs, but the project was saved thanks to the discovery of a single "umr" repository that made the box work well enough.
Still, Tiny Corp had reluctantly decided to launch an RTX 4090-powered TinyBox anyway since it offered a solution that [9]"just works." An Intel version might also be in the pipeline, though the performance of Intel's best GPU, the Arc A770, would be a significant concern since it comes nowhere near either the RX 7900 XTX and RTX 4090.
[10]
Yet Hotz's 70 percent confidence level was right in the end as AMD is seemingly doing exactly what he was asking for. Though making the MES firmware open source is a strange conclusion to the saga, and it's unclear why AMD seems to have changed course now.
"Spoke with them for months and months and got nothing, despite the public engagement you saw. I think it was treated as a PR problem, not a GPU problem," [11]Hotz speculated on X before AMD announced its open source efforts.
[12]
Hotz has not yet responded to AMD's announcement, and we've asked him and AMD to comment. ®
Get our [13]Tech Resources
[1] https://twitter.com/amdradeon/status/1775261152987271614
[2] https://twitter.com/amdradeon/status/1775999856420536532
[3] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[4] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[5] https://www.theregister.com/2024/03/27/amd_chiplets_future/
[6] https://www.theregister.com/2024/03/27/failed_intel_processor_nas/
[7] https://www.theregister.com/2024/03/27/tsmc_3nm_revenue_growth/
[8] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[9] https://www.theregister.com/2024/03/25/tiny_corp_amd_nvidia/
[10] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[11] https://twitter.com/realGeorgeHotz/status/1774533299559629033
[12] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44ZhB0eR@phGuNz-et-5ZtZQAAANY&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[13] https://whitepapers.theregister.com/
Navi coolz
HuBo
Sounds like great news to me ... the more open source the firmware, the better the potential for innovation! And the [1]RX 7900 XTX -based [2]tinybox looks quite neat, especially if it can run FP64 HPC loads at 11.5 TF/s (and less than 3 microwaves of power, 3.2 KW -- for those of us not that interested in AI/ML).
I'm not 100% clear though on how the [3]RTX 4090 version can give 991 TF/s at FP16 unless the cards are doubled-up (and then power should exceed what the PS provides). Maybe Holz's team meant FP8 in its specs table for this one (green team)?
[1] https://www.techpowerup.com/gpu-specs/radeon-rx-7900-xtx.c3941
[2] https://tinygrad.org/#tinybox
[3] https://www.techpowerup.com/gpu-specs/geforce-rtx-4090.c3889
Is this all to do with self driving cars, or has George Hotz realised that's actually a very difficult problem to solve?