I can finally make this thread.
Phoenix is indeed RDNA3. My advice to everyone: treat the old APU rumours as being out of date.
Phoenix is indeed RDNA3. My advice to everyone: treat the old APU rumours as being out of date.
On the plus side, exascale APU from DOE Fastforward programs is real!Do you remember all those super interesting accelerated things we were exposed to at those Fusion Dev Summits?
I mean half the issue is there being no unified accelerator programming model on Windows ever since MS killed C++ AMP off.Drivers regularly breaking browser HW acceleration, OpenCL kernels being rather disabled due instability, etc.
Yawn, scalable APUs are such an antique concept. After all those years of M-SPACE, Zeppelin big-APU, Exascale HP, etc. it is somehow economically viable. Yet, a good ol' monolithic GPU rules 'em all.On the plus side, exascale APU from DOE Fastforward programs is real!
Nah.Yet, a good ol' monolithic GPU rules 'em all.
That's a newer development/toy.
It's kind of funny how much AI has blown up recently. AI has been part of mobile phones for ages and nobody really cared (at least compared to today's extend).
Aside frequency the Qualcomm SoCs Microsoft uses aren't even anything special in the context of high end ARM SoCs. Even lagging behind contemporary SoCs in high end phones when they originally launched.Microsoft is pushing ARM so hard because Qualcomm is willing to give them what Intel and AMD have been unable to.
Considering ChatGPT is already perfectly mimicking humans' ability of empty sweet talking without regards to facts the chance AI will ever be able to detect actual human typing has become nil.I wonder how much Artificial Intelligence it takes to distinguish between humans typing and butt dialing / texting. But Apple with all the AI in the iPhones is still not capable of doing it.
Nuvia should help that substantially. But regardless, you need to think of this from Microsoft's perspective. Which means asking, "What will fundamentally change how someone uses their computer?".Aside frequency the Qualcomm SoCs Microsoft uses aren't even anything special in the context of high end ARM SoCs. Even lagging behind contemporary SoCs in high end phones when they originally launched.
The focus of 8cx g4 is literally CPU performance (and perf in general).10% better CPU performance
Ugh, no.Nuvia should help that substantially
Ugh, no.But true all day battery life
Modem is separate since 8cx g3 and that ain't changin'.pervasive cellular connectivity
I too love free dark silicon!and ubiquitous AI
They literally aren't.Microsoft is willing to put up with Qualcomm's shortcomings
The focus of 8cx g4 is literally CPU performance (and perf in general).
Like lol.
Ugh, no.
It's good but not that good.
Ugh, no.
Still very efficient.
Modem is separate since 8cx g3 and that ain't changin'.
I too love free dark silicon!
They literally aren't.
We've spent a decade seeing more and more of our workflow move into badly written Electron apps, with web tech running on top of an entire JavaScript engine. And now we all need top end single thread CPU performance just to run a basic chat app smoothly (I'm looking at you, Slack).The current edge/notebook AI hype brings back AMD memories.
Do you remember all those super interesting accelerated things we were exposed to at those Fusion Dev Summits? So much Microsoft, so much heterogenous, so much APU, so much faster, so much notebooks, so much wow.
A decade later we are pretty much struggling to accelerate like anything. Drivers regularly breaking browser HW acceleration, OpenCL kernels being rather disabled due instability, etc.
It sounds like you have not been corporately forced to use Microsoft Teams... It's way worse than Slack - a combo of utterly slow rendering of even the tiniest element combined with terrible backend response times, broken search functionality, broken formatting, terrible Office365 webhooks, laughable notification rate limits, etc. Slack is great, trust me.I'm looking at you, Slack.
Oh, suddenly people are seeing the light on this topic?I think there's some truth to that, but they're heavily pushing AI on the client side as well. No doubt it will be deeply integrated into Windows and Office in the not-so-distant future. There's a reason Intel and AMD are suddenly so concerned with it. This first gen isn't good for much more than Windows Studio effects, but I think we'll see big jumps with Strix Point. It's quite possible we see a future where the AI accelerator is as big and important as the GPU.
They're both bad software. I miss when desktop developers cared about performance.It sounds like you have not been corporately forced to use Microsoft Teams... It's way worse than Slack - a combo of utterly slow rendering of even the tiniest element combined with terrible backend response times, broken search functionality, broken formatting, terrible Office365 webhooks, laughable notification rate limits, etc. Slack is great, trust me.
Zoom's developers are from China - the C in that initialism - and it's pretty good with using resources. Maybe not good software but at least it uses a reasonable amount of memory.Why do we allow developers from BRIC countries that have every incentive to create mediocre results? There once was a time such behavior was supporting the enemy. Cheap is cheap, never breeds good results.
I wonder how much Artificial Intelligence it takes to distinguish between humans typing and butt dialing / texting. But Apple with all the AI in the iPhones is still not capable of doing it.
Generally you'd want hardware to handle the simpler/common cases, and ucode for the rarer and more complex ones. To quote Jim Keller:I wonder how much statistical analysis by AMD goes into CPU design and microcode instruction design. If 25% of your microcode is used by 95% of programs, do you target performance across recognized bottlenecks, at that 25%, at the 5% of programs using less common microcode, or spread efforts evenly across all instructions? My intuition would guess effort is directed at the biggest bottlenecks. I would imagine there is always low hanging fruit to tackle as processes shrink. But I do not know.
80% of core execution is only six instructions - you know, load, store, add, subtract, compare and branch. With those you have pretty much covered it.
AI is broad, but for what it's worth: in the near future, for stuff on a 128-bit bus with basic LPDDR5, running generalized LLM's locally won't be as much of a hit as people think nor as plausible in the sense of a good experience.Nuvia should help that substantially. But regardless, you need to think of this from Microsoft's perspective. Which means asking, "What will fundamentally change how someone uses their computer?".
10% better CPU performance, while nice, isn't going to change how you use your device. But true all day battery life, pervasive cellular connectivity, and ubiquitous AI - those will. Microsoft is willing to put up with Qualcomm's shortcomings in order to have a partner willing to push those experiences will them.
A lot. And accurate simulation and monitoring are the most important aspect in all that. There a lot changed in the last decades.I wonder how much statistical analysis by AMD goes into CPU design and microcode instruction design.
Yeah, discord is the only chat app, that has a decent native PC client And by far (memory consumption, perfomance etc...). It used to be golang and progressively switched to rust. A far cry from the electron apps nost other chats use.They're both bad software. I miss when desktop developers cared about performance.
Last time I checked the official "Discord Desktop app for Linux" it was Electron based. Is there an alt client?Yeah, discord is the only chat app, that has a decent native PC client And by far (memory consumption, perfomance etc...). It used to be golang and progressively switched to rust. A far cry from the electron apps nost other chats use.
Too bad it has the reputation of being a quirky gamer-only software.
Damn, it seems I was mistaken. The serverside uses Rust, Elixir, Python and C++ It seems that the client side is still either Electron or React Native (mobile clients).Last time I checked the official "Discord Desktop app for Linux" it was Electron based. Is there an alt client?