- Mar 3, 2017
- 1,684
- 6,227
- 136
But can they really? M$ is being disingenuous in a way and not really announcing anything about how the TOPS requirement can be met with a discrete card. This FOMO tactic on their part could make quite a few people jump to ARL due to its NPU (regardless of whether it meets the requirement or not and Intel could strike up a deal with M$ to enable support for its NPU so the requirement can be met using the entire SoC instead of just NPU). I just hate the uncertainty M$ has created and I'm curious to see how Lisa will counter that.not big box pc's that can do AI on a discrete card.
What he said is correct, yes.But can they really? M$ is being disingenuous in a way and not really announcing anything about how the TOPS requirement can be met with a discrete card. This FOMO tactic on their part could make quite a few people jump to ARL due to its NPU (regardless of whether it meets the requirement or not and Intel could strike up a deal with M$ to enable support for its NPU so the requirement can be met using the entire SoC instead of just NPU). I just hate the uncertainty M$ has created and I'm curious to see how Lisa will counter that.
Allow me to summarise the general reaction of the entire pres:What are the chances that Nvidia's keynote spoils the party? Announcing something that out-halos Halo.
Aren't upcoming (or even existing, can't remember) AMD chips claiming 75 TOPS?You guys are overlooking the elephant in the room, namely the M$ 40 TOPS requirement. How are they gonna put a positive spin on the fact that they don't meet that requirement? Yes, they can remain silent and ignore the topic altogether but it's such a huge indicator of the straining relationship between M$ and AMD/Intel.
Not the desktop parts unless AMD has had absolute perfect opsec.Aren't upcoming (or even existing, can't remember) AMD chips claiming 75 TOPS?
I am saving that image. Btw, what precision is that? INT8?View attachment 100231
RTX 40 series TOPS
Not the desktop parts unless AMD has had absolute perfect opsec.
My mind would be blown if the NPU part is somehow hidden on Zen 5 and then one fine day, AMD releases a new AGESA to enable it.Not the desktop parts unless AMD has had absolute perfect opsec.
Calm down or you may get indigestionWhat a waste of my 1pm breakfast.
Likely, I went to find the numbers and it popped up on twitter from someone I follow.I am saving that image. Btw, what precision is that? INT8?
I'm not the only one concerned:That is also why desktop chips having an NPU isn’t as important.
It is a requirement for a sticker, not an OS requirement.I'm not the only one concerned:
just not one that is as ridiculous as this one.
When Microsoft revealed the first Copilot+ laptops, all running Qualcomm Snapdragon X Elite processors, it wasn't just a slap in the face to AMD and Intel — Nvidia also got snubbed. Because Nvidia is the biggest name in the AI space right now, at least from the hardware side of things, many wondered why Copilot wasn't also able to run on powerful GPUs. Wonder no longer, as Nvidia announced at Computex 2024 that it's collaborating with Microsoft to bring the Copilot runtime to GPUs sometime later this year.
Nobody gives a flying feather about what MS will ask in terms of tech specs for the next version of Windows of for whatever killer new features. Believe it or not, it's MS who's more worried about what we ask of them to prop their adoption rates.I'm not the only one concerned:
OH Now I understand what you are getting at. Yeah, that likely is just a ploy to sell surface devices, but they didn’t want to limit it to just surface devices because it would look incredibly bad (and make OEMs unhappy) and they wouldn’t be able to sell as many subscriptions.Windows Copilot will add GPU support in a future release — Nvidia details the advantages of high performance GPUs for AI workloads and more
Firing up a GPU versus an NPU will still create a massive drain on your laptop battery.www.tomshardware.com
Somewhat better but I hope to hear the same good news from AMD too.
People think we are in a recession right now (we aren’t), but when the AI bubble pops we’ll beNobody gives a flying feather about what MS will ask in terms of tech specs for the next version of Windows. Believe it or not, it's MS who's more worried about what we ask of them to prop their adoption rates.
Microsoft's Windows 11 upgrade pleas get even more desperate
Windows 11 hasn't been as successful as Microsoft had hoped—and they still really want to convince Windows 10 users to switch.www.pcworld.comWindows 10 returns to 70% market share as Windows 11 continues to sink
Windows 11 has entered its second month of market share decline on Statcounter's global chart. The OS sank to around 25 percent adoption among Windows users while...www.techspot.com
Imagine not being able to convince your users to upgrade for free.
The four machines highlighted here are all based on the upcoming AMD Strix Point platform and include a discrete Nvidia GeForce GPU.
We got a long term roadmap and nothing else of value.They have little to show, and whatever new they could show isn't out until next year.
Sure they could tease client BW but that really doesn't change anything outside of being seen as desperate.
All this is is their introduction to doing PC SoC's and faffing off about software as per usual.
It is very obviously designed to take anything away from AMD's actual Computex opening keynote. I guarantee they mention both Qualcomm and Intel and avoid AMD.
A very petty off premises unofficial 'keynote' plus the event they are doing a few days after is all to try to fill the newscycle and obfuscate AMD.
Because guess what, AMD are the only actual competitor and if you need proof, well how NV deals with them is the proof.
The 2025 part is clearly a premium ultrathin part, they want dGPU for the higher TDP designs.
AMD will show what comes out in Q3 and even then RDNA4 is likely a teaser with launch at probably Gamescom.
I am hoping for a one more thing, whether a Halo tease or MI400 tease, but AMD probably wants to keep max opsec on 2025 parts.
Yeah NV's desperation was mostly visible at the seams (that MI300 assassination article that got destroyed within 2 business hours), but right now, it's visible at the core. Jensen is seriously under stress. Which raises my expectation/certainty of MI350X announced within a few hours within the slew of AI talks and Granite Ridge.They have little to show, and whatever new they could show isn't out until next year.
Sure they could tease client BW but that really doesn't change anything outside of being seen as desperate.
All this is is their introduction to doing PC SoC's and faffing off about software as per usual.
It is very obviously designed to take anything away from AMD's actual Computex opening keynote. I guarantee they mention both Qualcomm and Intel and avoid AMD.
A very petty off premises unofficial 'keynote' plus the event they are doing a few days after is all to try to fill the newscycle and obfuscate AMD.
Because guess what, AMD are the only actual competitor and if you need proof, well how NV deals with them is the proof.
if half the pres is AI AI AI, you very quickly reach 90 minutes.
Not speaking for him obviously, but the fundamental trait of Nvidia is that they play alone and share nothing. They always wanted total control.You seem to quite optimistic about AMD's chances.