Monday, February 10th 2020
AMD Radeon Instinct MI100 "Arcturus" Hits the Radar, We Have its BIOS
AMD's upcoming large post-Navi graphics chip, codenamed "Arcturus," will debut as "Radeon Instinct MI100", which is an AI-ML accelerator under the Radeon Instinct brand, which AMD calls "Server Accelerators." TechPowerUp accessed its BIOS, which is now up on our VGA BIOS database. The card goes with the device ID "0x1002 0x738C," which confirms "AMD" and "Arcturus,". The BIOS also confirms that memory size is at a massive 32 GB HBM2, clocked at 1000 MHz real (possibly 1 TB/s bandwidth, if memory bus width is 4096-bit).
Both Samsung (KHA884901X) and Hynix memory (H5VR64ESA8H) is supported, which is an important capability for AMD's supply chain. From the ID string "MI100 D34303 A1 XL 200W 32GB 1000m" we can derive that the TDP limit is set to a surprisingly low 200 W, especially considering this is a 128 CU / 8,192 shader count design. Vega 64 and Radeon Instinct MI60 for comparison have around 300 W power budget with 4,096 shaders, 5700 XT has 225 W with 2560 shaders, so either AMD achieved some monumental efficiency improvements with Arcturus or the whole design is intentionally running constrained, so that AMD doesn't reveal their hand to these partners, doing early testing of the card.
-- images removed --
Looking through the BIOS I also found what looks like several clock tables that top out at 1334 MHz, 1091 MHz, 1000 MHz. AMD's engineers typically list clocks in the following order: GPU clock, SOC clock, memory clock. This suggests that the GPU will tick at up to 1334 MHz, way lower than what Navi and Vega were able to achieve — maybe they do that to operate the chip in a more power-efficient way. The memory clock at 1000 MHz, matches the BIOS id string's "1000m", and falls in range with the 2.0 - 2.4 Gbps that Samsung is speccing their HBM2 memory chips at.
Arcturus' debut as a Radeon Instinct product follows the pattern of AMD debuting new big GPUs as low-volume/high-margin AI-ML accelerators first, followed by Radeon Pro and finally Radeon client graphics products. Arcturus is not "big Navi," rather it seems to be much closer to Vega than to Navi, which makes perfect sense given its target market. AMD's Linux sources mention "It's because Arcturus has not 3D engine", which could hint at what AMD did with this chip: take Vega and remove all 3D raster graphics ability, which shaves a few billion transistors off the silicon, freeing up space for more CUs. For gamers, AMD is planning a new line of Navi 20-series chips leveraging 7 nm EUV for launch throughout 2020. Various higher-ups at AMD, including its CEO, publicly hinted that a big client-segment GPU is in the works, and that the company is very much interested at taking another swing at premium 4K UHD gaming.
Sources:
Arcturus Linux Patches, Arcturus Linux Patches
Both Samsung (KHA884901X) and Hynix memory (H5VR64ESA8H) is supported, which is an important capability for AMD's supply chain. From the ID string "MI100 D34303 A1 XL 200W 32GB 1000m" we can derive that the TDP limit is set to a surprisingly low 200 W, especially considering this is a 128 CU / 8,192 shader count design. Vega 64 and Radeon Instinct MI60 for comparison have around 300 W power budget with 4,096 shaders, 5700 XT has 225 W with 2560 shaders, so either AMD achieved some monumental efficiency improvements with Arcturus or the whole design is intentionally running constrained, so that AMD doesn't reveal their hand to these partners, doing early testing of the card.
-- images removed --
Looking through the BIOS I also found what looks like several clock tables that top out at 1334 MHz, 1091 MHz, 1000 MHz. AMD's engineers typically list clocks in the following order: GPU clock, SOC clock, memory clock. This suggests that the GPU will tick at up to 1334 MHz, way lower than what Navi and Vega were able to achieve — maybe they do that to operate the chip in a more power-efficient way. The memory clock at 1000 MHz, matches the BIOS id string's "1000m", and falls in range with the 2.0 - 2.4 Gbps that Samsung is speccing their HBM2 memory chips at.
Arcturus' debut as a Radeon Instinct product follows the pattern of AMD debuting new big GPUs as low-volume/high-margin AI-ML accelerators first, followed by Radeon Pro and finally Radeon client graphics products. Arcturus is not "big Navi," rather it seems to be much closer to Vega than to Navi, which makes perfect sense given its target market. AMD's Linux sources mention "It's because Arcturus has not 3D engine", which could hint at what AMD did with this chip: take Vega and remove all 3D raster graphics ability, which shaves a few billion transistors off the silicon, freeing up space for more CUs. For gamers, AMD is planning a new line of Navi 20-series chips leveraging 7 nm EUV for launch throughout 2020. Various higher-ups at AMD, including its CEO, publicly hinted that a big client-segment GPU is in the works, and that the company is very much interested at taking another swing at premium 4K UHD gaming.
76 Comments on AMD Radeon Instinct MI100 "Arcturus" Hits the Radar, We Have its BIOS
edit:oh yeah,that's what it is.
v100s was 1600mhz on 12nm at massive 815nm.ampere on 7nm euv ? probably dies not need remotely close to 8000sp to beat this given the clocks and asic hardware
I know it's comming and I know that consoles probably will get RDNA2 for RT, but AMD must have at least an enthussiast card and if we have luck a performance/high performace card at the launch of the next gens to leverage some of the new graphics of the next-gen multiplatform titles. Perhaps Cyberpunk 2077?
that's actually very good they're developing specific architectures for hpc and gaming.
vega 14nm/7nm was a jack of all trades,master of none.people tend to want a gaming card or a hpc card,not hpc and gaming.high performance is a vague term.
I wouldn't call my 2070 super a high performance card,it's okay,kinda upper mid range and that's all.It's fine for 1440p.For now.
meanwhile 5700xt is high performance for amd despite it's 2070 non super.
There's also likely a version with actual display out. V20 powers the MI50 and MI60 that have their display functionality cut. I also really hope that AMD decides to turn on the degenerated rendering hardware finally.
What'll be interesting is if AMD decided to add in AI cores to boost the same specific work loads that Tensor cores do.The Radeon VII is the people's champion of video and image processing right now.
Also, the Radeon VII was NOT open to corporate pricing (e.g. bulk orders, support, etc.) and you would be recommended to go with the MI50 or MI60 at the time. It really was a niche halo card aimed at the consumer-market while AMD was getting Navi ready.
I can buy a lot of Radeon VIIs for the price of a single V100.
Guess which will do Premiere faster? How many 16x slots do I get to fill 6, 7, 8?
I'll watercool the Radeons, and absolutely destroy the V100 and likely work on multiple projects at the same time.
Not all of us are named Steve and get NV corporate gear at solid prices or basically free. *snicker*
The V100 is nowhere near a consumer product. I don't know why you keep comparing the Radeon VII to the Tesla V100. It's practically $680 versus ~$13,000 (NVLINK). Both of these aren't even workstation cards, where one is a normal consumer card and the other is a data center GPU not meant to be used in a workstation.
And where are you seeing that Premiere is faster on a Radeon VII? Unless you're running a NVIDIA card in OpenCL on Windows (which is counter-productive and not recommended), the Radeon VII is not the champion you think it is:
It is damn good in DaVinci Resolve and when using certain plugins.
Also, if you're testing the Radeon VII on a Mac, you should be using Metal in Premiere Pro as OpenCL has been deprecated in the new OSX. This is where I think you would see a big performance increase (matching the 2080 Ti).
GCN6 is just speculation coined by me.... so its just speculation (unofficial).. but i know it makes sense either way.
AMD is not exactly organized...
The codename is probably MI100 tbh
The prototype chip says MI100 XL which is usually in the place of where it would say Polaris 10 for example.
Edit: Also Raven is GFX9 and not Vega architecture
It's like Polaris with early Navi encoder and decoder engine
VegaM (Polaris 22) is GFX804 CLRX 1.2
Polaris with HBM
AMD is confusing trust me I work on a lot of their stuff (for gpudb) So GCN6 is coined.. pure speculation but I have reason to think it's not GCN5.
also I wont be surprised if it is GCN5 either, this is just my speculation from the chaos that is AMD... that is all.
Where CUDA shines the most are smaller projects, ad-hocs, daily computation problems - especially when code is written not by full-time developers but scientists, engineers or analysts.
So as long as these clusters are used for weather models and so on - nothing bad happens.
But if this was a cluster for end users (i.e. private cloud), lack of CUDA would surely hurt.CUDA is easy to use and very popular. This is not a matter of what's needed, but what's the most beneficial.
this is "ARCTURUS" SERVER/COMPUTE ACCELERATOR..
****THIS DOESN'T EVEN HAVE 3D CAPABILITIES****
BIG NAVI IS COMING!!!!!! but WILL NOT BE AS LARGE AS THIS "ARCTURUS" cGPU... aka(COMPUTE GPU) or ML/AI/COMPUTEACTUALLY....TBASA...lol
Arcturus is a red giant star in the Northern Hemisphere of Earth's sky and the brightest star in the constellation Boötes (the herdsman). Arcturus is also among the brightest stars that can be seen from Earth. Astronomers say Arcturus will end up as a white dwarf at the end of its life
this may be the last of GCN in other words.... At the end of its life, is AMD throwing a Riddle at us??? using this as the name of the silicon>>>
idk
NAVI 14 shader engine has 24 CU.
NAVI 10 shader engine has 20 CU, but NAVI 10 has two shader engines.