Xbox next likely to be zen 5 (+ RDNA 5)Semi OT:
The next console generation both excites me and worries me. If we presume Z6 that will be a very welcome IPC increase, I'm not really worried about that. Hopefully more than 8c/16t as well, I'm counting on 12c/24t minimum. On the GPU side I feel a lot more iffy, whether tech currently quite far in the pipeline will be able to meet the AI/ML/RT needs of 2027-2034, can those GPU cores be beefy enough given AMD's current situation (or NV for that matter, just iteration since Ada)? I also fear for the (V)RAM capacity, it should be able to hold future AI models in memory along with improved textures, and there will probably be a lot of AI breakthroughs in that time period, and we just can't imagine all the potential use cases right now, so I fear being hobbled by 16GB or 24GB, hopefully they go for 32, 48, or even 64GB. History tells us that memory has been one of the most vulnerable specs.
Semi OT:
Hopefully more than 8c/16t as well, I'm counting on 12c/24t minimum.
Next-Gen consoles will take the learnings of this current-gen. They'll be priced to recover margins quickly and not suffer from price raises as console gamers are sensitive to price increases. Xbox will also be a PC, so that will result in a very different console that will be sold in fewer countries and for profit.Semi OT:
The next console generation both excites me and worries me. If we presume Z6 that will be a very welcome IPC increase, I'm not really worried about that. Hopefully more than 8c/16t as well, I'm counting on 12c/24t minimum. On the GPU side I feel a lot more iffy, whether tech currently quite far in the pipeline will be able to meet the AI/ML/RT needs of 2027-2034, can those GPU cores be beefy enough given AMD's current situation (or NV for that matter, just iteration since Ada)? I also fear for the (V)RAM capacity, it should be able to hold future AI models in memory along with improved textures, and there will probably be a lot of AI breakthroughs in that time period, and we just can't imagine all the potential use cases right now, so I fear being hobbled by 16GB or 24GB, hopefully they go for 32, 48, or even 64GB. History tells us that memory has been one of the most vulnerable specs.
I anticipate Zen 6 to over-perform for what it is. Zen 5 was the new architecture but being forced to stay on 5nm didn’t let it shine the way it would have. Zen 6 is the new packaging, probably new IOD and builds on the new core. If they jump straight to 2nm that would be exceptionally ballsy and deliver two node jumps over Zen 5.If Zen 6 is 12c/24t you won't think it's a big deal because that was your minimum expectation? Would it require 16/32 to get you excited about Zen 6?
What is your minimum for an average IPC increase?
I anticipate Zen 6 to over-perform for what it is. Zen 5 was the new architecture but being forced to stay on 5nm didn’t let it shine the way it would have. Zen 6 is the new packaging, probably new IOD and builds on the new core. If they jump straight to 2nm that would be exceptionally ballsy and deliver two node jumps over Zen 5.
I think it may be another Zen 4 moment. Not a massive IPC increase but everything else together should make for nice ST and MT gains. Zen 4 delivered 25% better ST and half of that was just from clockspeeds. Very impressive job by everyone involved.
They really don't need that. 8xZ6c as a single CCX is already an absolutely massive upgrade over PS5, and I frankly think that the additional power and die area above that would be better spent on the GPU. It's okay for console CPUs to not be exactly top of the line, there is less random stuff the machine needs to do in the background.Semi OT:
The next console generation both excites me and worries me. If we presume Z6 that will be a very welcome IPC increase, I'm not really worried about that. Hopefully more than 8c/16t as well, I'm counting on 12c/24t minimum.
I think the most likely config is going to be 256-bit GDDR7 with 24Gbit chips, so 24GB of RAM. Possibly a single LPDDR6 module or something for the system host chip, Sony likes the model where they don't have to wake up the main CPU/RAM for downloads, it helps them meet EU energy standards about idle power and also helps them keep the secret security stuff more isolated from the things they have to give access to people outside the company.I also fear for the (V)RAM capacity, it should be able to hold future AI models in memory along with improved textures, and there will probably be a lot of AI breakthroughs in that time period, and we just can't imagine all the potential use cases right now, so I fear being hobbled by 16GB or 24GB, hopefully they go for 32, 48, or even 64GB. History tells us that memory has been one of the most vulnerable specs.
How has the 5nm node held back Zen 5 performance?I anticipate Zen 6 to over-perform for what it is. Zen 5 was the new architecture but being forced to stay on 5nm didn’t let it shine the way it would have. Zen 6 is the new packaging, probably new IOD and builds on the new core. If they jump straight to 2nm that would be exceptionally ballsy and deliver two node jumps over Zen 5.
I think it may be another Zen 4 moment. Not a massive IPC increase but everything else together should make for nice ST and MT gains. Zen 4 delivered 25% better ST and half of that was just from clockspeeds. Very impressive job by everyone involved.
it's a two node shrink.I hope you are right. I would be shocked if Zen 6 hit 6GHz under any scenario until ambient non custom cooling solution. Shocked in a good way of course!
If it's actually using N2, that seems pretty likely. Conservative, even.I hope you are right. I would be shocked if Zen 6 hit 6GHz under any scenario until ambient non custom cooling solution. Shocked in a good way of course!
3D cache benefits low resolution gaming the most since it’s CPU bound. Given consoles are framerate locked to 60 FPS on a good day, what would be the point? Especially if most are playing at 4K or 4K upscaled.What I'm more interested in is if the next Gen game consoles get 3dCache? If their primary use case is exactly what that is targeted at, why not use it?
Doubt it will include the cache stacked because that process takes both time and there is not the worldwide capacity for it. Any extra capacity will be taken my higher margin products. Maybe the gen after that.What I'm more interested in is if the next Gen game consoles get 3dCache? If their primary use case is exactly what that is targeted at, why not use it?
X3D is more expensive and time-consuming. Maybe for a Pro on Sony side. Xbox is whatever as they're doing a for profit device, so they can push as high as they want.What I'm more interested in is if the next Gen game consoles get 3dCache? If their primary use case is exactly what that is targeted at, why not use it?
Found this.
Seems like AMD are already talking about a Strix Point refresh next year...
View attachment 120639
View attachment 120640
View attachment 120641
Can't find the name AI 400 on the slides, but it's on the original image filenames.
From what I've noticed, it's pretty normal for graphical demands to increase a lot in the first half a console generation, and then plateau in the second half (there are some notable exceptions of course). Games will get much more demanding after the PS6 comes out.
It's not raw perf but perf at iso TDP. Which doesn't mean it makes sense.??? gorgon point looks nonsense
how will it get +95% CB MT perf over strix point when it uses the same core type and count lol
what is the "refresh" even about
You are comparing the wrong datapoints. The 100% is Strix Point @15W. So when using more watts, both Strix and Gorgon will perform better in nT. For comparing Strix and Gorgon @iso power, you need to check the two close columns at same wattage, i.e. @45W Strix Point will perform +81% respect to Strix Point @15W while Gorgon Point will perform +95% respect to the baseline at the same power. That is, the performance increase in nT @45W going from Strix to Gorgon is 195/181=7% approximately. This refresh will be the same chip, maybe new stepping, slightly better efficiency allowing slightly higher clocks and a +10-15% on the NPU performance. That's all.??? gorgon point looks nonsense
how will it get +95% CB MT perf over strix point when it uses the same core type and count lol
what is the "refresh" even about
That s 95% better perf at 45W comparatively to a 15W Strix Point and 5% at same 15W power.??? gorgon point looks nonsense
how will it get +95% CB MT perf over strix point when it uses the same core type and count lol
CASEY: You will often hear “people on the internet” say that ARM as an ISA is better for low power than x64. People like me who study ISAs tend to be skeptical of this claim. As a hardware designer, are there any specific things about the x64 ISA that you find difficult to deal with for low-power designs?
MIKE: Having spent my career working on x86, I might have a bias here! I do think each ISA has its own quirks that influence some of the microarchitecture. But at the base level, we can build low-power x86 designs as well as ARM can, and ARM can build high frequency, high performance designs as well as x86 can. None of the quirks are really limiting you on the microarchitecture. The reality is that the markets we've been targeting have been different, so they've driven the architectures to optimize for different design points. ARM is in much lower power markets where x86 hasn't had the market share to chase.
On the x86 side, the higher performance / higher frequency devices are the established market that our devices have to compete in, so that's where our design focus is. We could build the same Zen microarchitecture with an ARM ISA on top instead. We could deliver the same performance per watt. We don't view the ISA as a fundamental input to the design as far as power or performance.
The rest of the interview is worth reading and covers topics like 4 KB vs 2 MB pages, CPU vs GPU design targets, why CPUs still use 64B cache lines, and more…
I think you are reading the graph wrong.??? gorgon point looks nonsense
how will it get +95% CB MT perf over strix point when it uses the same core type and count lol
what is the "refresh" even about
That s 95% better perf at 45W comparatively to a 15W Strix Point and 5% at same 15W power.
Most of which is probably a more optimised stepping for a very new µArch, more clock, less errata, and more tuning in the platform firmware/BIOS/microcode relative to Strix (Zen1 was still tuning at the platform level well over a year after release IIRC).