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
Add your own comment

73 Comments on AMD Radeon Instinct MI100 "Arcturus" Hits the Radar, We Have its BIOS

#1
DeathtoGnomes
something must be getting in the way of higher clocks,maybe?
Posted on Reply
#2
mak1skav
HBM2 ??? NOooooo, price will be too high once again :(
Posted on Reply
#3
FordGT90Concept
"I go fast!1!11!1!"
DeathtoGnomes
something must be getting the way of higher clocks,maybe?
They're probably designed for maximum performance per watt which is what server farms want. Spending 50% more power for 10% performance uplift doesn't fly in an environment running 1000 cards.


I'm kind of disappointed Arcturus is aiming for ultra-high end. I hope "big Navi" is still coming.
Posted on Reply
#4
londiste
Does BIOS show it is a 80CU design?
Memory specs are exact match for what was used in MI60, so not much newsworthy there.

FordGT90Concept
They're probably designed for maximum performance per watt which is what server farms want. Spending 50% more power for 10% performance uplift doesn't fly in an environment running 1000 cards.
This. 100MHz lower base clock (1091 vs 1200) and considerably lower boost clock (1400 vs 1800) does help a lot with power efficiency. Assuming MI100 is 80CU, AMD still has managed a huge efficiency boost though.
Posted on Reply
#5
T4C Fantasy
CPU & GPU DB Maintainer
londiste
Does BIOS show it is a 80CU design?
Memory specs are exact match for what was used in MI60, so not much nesworthy there.
It's a 128cu design and essentially gcn 6.0
Posted on Reply
#6
delshay
Well at lease the power limit is not that far away for a Nano card.
Posted on Reply
#7
HwGeek
So it confirms that TSMC's N7 is already good enough even for large 650+ mm² silicone?
Posted on Reply
#12
ARF
Wow, this chip is impressive nonetheless!

AMD is finally very brave to design such a monster of a chip! :eek:

I'd like to see Navi 21-based consumer Radeon as soon as possible, too!
Posted on Reply
#13
T1beriu
DeathtoGnomes
something must be getting the way of higher clocks,maybe?
It's the doubling of CUs. If you want 2GHz from this beast then it would use 6-700W. It's simple frequency/cores/power scaling.

T4C Fantasy
It's a 128cu design and essentially gcn 6.0
Since Komachi discovered that Arcturus is GFX908 then that's GCN 5 (Vega) or GCN 5.2 (if Vega 20 is GCN 5.1)


ARF
It's actually N7+ according to the AMD MI100 Specs page at https://www.techpowerup.com/gpu-specs/amd-mi100.g927
That's not AMD's spec page. It's Techpowerup's speculated spec list.

ARF
I wonder which one will be better for gaming: MI100 or Navi 21 https://www.techpowerup.com/gpu-specs/amd-navi-21.g923 ? ?
We'll never know because Arcturus seems to have no video out ports.
Posted on Reply
#14
T4C Fantasy
CPU & GPU DB Maintainer
T1beriu
It's the doubling of CUs. If you want 2GHz from this beast then it would use 6-700W. It's simple frequency/cores/power scaling.



Since Komachi discovered that Arcturus is GFX908 then that's GCN 5 (Vega) or GCN 5.2 (if Vega 20 is GCN 5.1)




That's not AMD's spec page. It's Techpowerup's speculated spec list.



We'll never know because Arcturus seems to have no video out ports.
GCN 3 and 4 were GFX8 so I don't see why it can't be GCN6
Posted on Reply
#15
Imsochobo
mak1skav
HBM2 ??? NOooooo, price will be too high once again :(
Why should you care? it's not for you.

As in, it's not for gamers, datacenter only :P
Posted on Reply
#16
notb
Imsochobo
Why should you care? it's not for you.

As in, it's not for gamers, datacenter only :p
IMO the idea here is that some would like this card rebranded for gaming.
Posted on Reply
#17
HwGeek
But it wont. it's just for compute , it's cannot become a desktop GPU like Mi60->Radeon VII.
Posted on Reply
#18
ShurikN
700mm2 on 7nm EUV... damn.
Imagine this thing on 14nm, you'd have 3 working dies per wafer :laugh:
Posted on Reply
#19
HwGeek
I smell $1000 Gaming GPU's this year by AMD...
Posted on Reply
#20
HenrySomeone
notb
IMO the idea here is that some would like this card rebranded for gaming.
That would be another massive failure just like the Crap-eon 7, but then again, AMD just doesn't seem to tire of failures... :D
Posted on Reply
#21
Vayra86
There is no way AMD is going to present an 8000+ shader consumer GPU and if they do, it won't be economically viable. Can't go this low on clocks and this large on die and still sell with any sort of margin. Especially not with HBM on top. I also struggle to see the relation to Big Navi here.

Also... if AMD is really going to shoot for another Vega repeat with HBM and once more get eclipsed by a simple x80 Ampere... they can close up shop.

So, forget it. Nice proof of concept, not happening for us. Still its nice to see them do a large die like this.
Posted on Reply
#22
A.Stables
Waiting for the die immage :) this is impressive stuff AMD
Posted on Reply
#23
john_
I looked at NVIDIA Quadro RTX 8000 specs and it says over 200 TOPS at INT8. Is there really so huge difference between Radeon and RTX, or does Nvidia counts differently?
Posted on Reply
#24
notb
john_
I looked at NVIDIA Quadro RTX 8000 specs and it says over 200 TOPS at INT8. Is there really so huge difference between Radeon and RTX, or does Nvidia counts differently?
In general use, running dumb, unoptimized code, they're pretty close (Mi60 vs V100).
But Nvidia adds Tensor cores and they make all the difference - Nvidia ends up 3-4 times faster in some tasks.

In real life, Nvidia dominated GPU computing before they added Tensors - all thanks to a better ecosystem and support. And nothing changed here.
Even if Mi100 temporarily pulls ahead in performance, it won't be enough.
Posted on Reply
#25
ppn
ShurikN
700mm2 on 7nm EUV... damn.
Imagine this thing on 14nm, you'd have 3 working dies per wafer :laugh:
ASML
"...new maximum die size of 429 mm². Say goodbye to the massive dies we got used to from Intel and Nvidia. ...",

Vega 20 Transistors 13,230 million Density 40.0M / mm² 7 nm Die Size 331 mm²

MI100 Transistors29,000 million Density 41.4M / mm² 7 nm+ Die Size 700 mm²

Same Density as first gen 7nm. The real EUV should be 18% denser or 505 mm² NAVi21 gets shrinked to exactly 429mm²
Posted on Reply
Add your own comment