Very different products MI300 is high power Multi chip package solution for AI using 850W to achieve 2507 TFlops @ FP8. I would guess this will be in the ballpark of 400 FP8 TFLOPS, if so that is about 1/6 the performance of the MI300 but at a fraction of the power required(I would guess less the 75W if not less than 35W).
Tenstorrent never released those PCIe cards for retail, and I have yet to see them running inference/training loads in anything. Are conglomerates like their Korean partners just using them internally?
Who told you 4nm-class in 2024 is expensive? And why do you think SF4X is more expensive than previous? "tailored for high clocks and high voltages" means relaxed dimensions, that's CHEAPER to make. Samsung is simply tuning the same process for high voltage with RELAXED design rules and LOWER density. Just like N3E is CHEAPER than N3B.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
13 Comments
Back to Article
Dante Verizon - Tuesday, October 3, 2023 - link
Will this have a third of the mi300 performance?cyrusfox - Wednesday, October 4, 2023 - link
No I don't think it will be 1/3, my guess is 1/6Very different products MI300 is high power Multi chip package solution for AI using 850W to achieve 2507 TFlops @ FP8. I would guess this will be in the ballpark of 400 FP8 TFLOPS, if so that is about 1/6 the performance of the MI300 but at a fraction of the power required(I would guess less the 75W if not less than 35W).
Dante Verizon - Wednesday, October 4, 2023 - link
At least it will be very efficient if it reaches this TDP.Sudutooo - Wednesday, October 4, 2023 - link
Yes, very efficient, but many are looking for logical benefits, or has mathematics come into a strangle hold when it comes to computing?skaurus - Tuesday, October 3, 2023 - link
I find it strange that "low-cost, low-power" chip is going to be produced on 4nm (=expensive) node, "tailored for high clocks and high voltages".brucethemoose - Wednesday, October 4, 2023 - link
Has anyone seen/used these in the wild?Tenstorrent never released those PCIe cards for retail, and I have yet to see them running inference/training loads in anything. Are conglomerates like their Korean partners just using them internally?
Sudutooo - Wednesday, October 4, 2023 - link
One might think that with thinner strats, one could get more logical error due to electronic interference?brucethemoose - Wednesday, October 4, 2023 - link
Oops that wasn't supposed to be a reply.Dante Verizon - Wednesday, October 4, 2023 - link
It should be cheaper than TSMC's 4nm.Sudutooo - Wednesday, October 4, 2023 - link
May have a longer lifespan? No sense in having the law of obsolescence be too short? A test of 4 nm?dotjaz - Friday, October 20, 2023 - link
Who told you 4nm-class in 2024 is expensive? And why do you think SF4X is more expensive than previous? "tailored for high clocks and high voltages" means relaxed dimensions, that's CHEAPER to make. Samsung is simply tuning the same process for high voltage with RELAXED design rules and LOWER density. Just like N3E is CHEAPER than N3B.Sudutooo - Wednesday, October 4, 2023 - link
Yes, but what are the logical benefits of the chip otherthan lower power costs?Sudutooo - Wednesday, October 4, 2023 - link
In my opinion, a computer should last much longer than six to ten years; but AI may overtake us by then?