Recent content by Joe NYC

I need to check. But water cooling is kind of that - it can remove several X the amount of heat and at the same time, can keep the GPUs (and CPUs) cooler, reducing the stress on them, while also clocking them to the max. Air cooling will be gone in not too distant future at the Hyperscalers...

The goal seems to be to sell the entire rack as a self contained unit. So to go up to 2x GPUs, they came up with double rack, again as a unit. Maybe NVidia will announce something similar soon.

I don't think I will be able to resist the 12 core Zen 6 at 7 GHz with 2 layers of V-Cache. 😉

I think I read the limit for UA Link is 1024. I don't know if this was just Dylan's hallucination, or if Mi400 will be only 4 GPUs per node, rather than 8. So, to get to 256, they will need 64 nodes. Since they are going to be using double wide rack, maybe they are limiting themselves to 144...

NVidia has a networking system that lets as many as 72 GPUs share workload and share memory. This is important mainly in training but also may come more and more into play in complex inference tasks over time. AMD can only share 8 GPUs in the same system, currently. AMD is developing industry...

Wow, hard to believe. Moving on to Q3, Intel is guiding revenue up only $200m (12.9 to 13.1 midpoint) or +1.6% while AMD is guiding up $1 billion (7.7 to 8.7 billion) which is +13%

There was a lot of misunderstanding in how far behind AMD was in combination of GPU hardware, software and networking. And it is a combination of all 3 that allows you to have a competitive offering in the market place. Without all 3, AMD was just fighting for crumbs, offering Mi300/Mi325 at...

Additionally + $1 billion QoQ revenue guidance for Q3. And that's up from record quarterly revenue.

The difference is that in previous 3 investor calls, AMD gave either poor or no guidance on the datacenter GPU revenue, and this time, AMD gave strong guidance. It seems that this time, the immediate reaction is misinterpreting the results. I expect the Wall Street analysts to work the strong...

In client revenue: AMD is up 66% vs. a year ago Intel is down (undermined number) Hard to say how Intel client is performing since they now padded its totals by transferring into it NEX revenue.

Are we residing in the same universe? This was an excellent quarter, and old Mi300/Mi325, which were at lower profit margins, did not sell very well, are finally exiting the picture. There is nothing AMD could do about the (temporary) ban of Mi308 GPUs to China, and all of the consequences...

No, AMD implied this was silicon for upcoming Christmas shopping season (at lower than corporate margin) and also dGPU revenue increase. During the previous CCs, AMD said that there were some inventories of console chips, being worked down, but it appears, the inventories were gone by Q2 and...

Seems like some serious binning issues. Qualcomm also had similar binning issues, where only a small percentage of chips reached the desired clock speeds...

BTW, interesting that AMD is moving to N2 with Mi400 while NVidia will use N3 for Rubin. Maybe NVidia did not order enough capacity early enough

No, AMD is definitely aiming for training, as well as inference. Training with Mi400 GPUs in 2026. Inference currently, with Mi300, Mi325, Mi355. AMD is (or aims to) to outcompete: H100 with Mi300 H200 with Mi325 B200 with Mi355 The limitation is that the model has to fit within 1 or 8 GPUs...