• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

*** AMD "Zen 4" thread (inc AM5/APU discussion) ***

The FX-8350 had 8 Integer Units, the Integer Unit is what you would traditionally call the "core"

It had a programmable, or configurable front end, it had two 128Bit FPU's per Integer which could be reconfigured to one 256Bit FPU giving it the ability to run either wide FPU's for low threaded workloads or narrow FPU's for high threaded workloads. Or even narrow FPU's in idle loads to conserve power..

Actually a very clever and innovative design, the problem is it didn't work, the Windows Scheduler only ever saw eight 128Bit FPU's no mater what the load on it and Microsoft had no intention of, and never did fix that.
 
humbug said:
This, i think that will come when AMD get more wafer capacity. Right now they are trying to get into new segments with the sub 10 Watt chips
Click to expand...

Probably cost as well. Van Gogh appears to be very close to the console SOCs they have already designed so scaling that design down is probably a lot cheaper than designing a 4c + RDNA 2 zen 3 based SOC.
 
TSMC 4nm Mass Production Preponed to Q4 2021; AMD 5nm Chips to Start Production in H2 2021

As for the 5nm node, most of TSMC’s primary clients are going to make the transition starting from the second half of the year. The biggest names include AMD, Qualcomm, and MediaTek. The 7nm shares of the same chipmakers are also expected to grow significantly in the same time period, with AMD already confirmed to have its 7nm capacity nearly doubled by Q4.
Click to expand...

https://www.hardwaretimes.com/tsmc-...amd-5nm-chips-to-start-production-in-h2-2021/
 
Those Van Gogh rumours only make sense if their agreement with Sony or Microsoft would allow them to sell salvaged parts.
Taping out a and a designing a 200mm² or even 300mm² 7nm with die with Zen2 just doesn't make sense.
I know that they are trying to play segmentation games with the Zen2 Luciennes but those use the same 7nm wafers as the Zen3 parts.
 
Last edited:
AMD Ryzen 7000 Series ‘Raphael’ CPUs Leaked In New Roadmap – 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022
AMD Ryzen 7000 Series 'Raphael' CPUs Leaked In New Roadmap - 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022 (wccftech.com)



Grim5 said:
Highlights

* Zen3 will be followed by: Zen3+, TSMC 6nm on AM4 with ddr4 and pcie4

* Followed by: Zen 4, TSMC 5nm on AM5 with DDR5 and pcie5. Launch quarter is Q3 2022. Also Zen 4 includes an RDNA2 based iGPU, all Zen 4 CPU include an iGPU chiplet.

RDNA3: TSMC 5nm, launch quarter is Q3 2022, target performance is 50% increase per watt

Zen 3 Threadripper target launch month: August 2021

https://www.chiphell.com/thread-2314832-1-1.html

https://www.reddit.com/r/Amd/comments/mjst0p/latest_amd_cpu_roadmap/

https://mobile.twitter.com/Olrak29_/status/1378488719787786240
Click to expand...

Including an iGPU that no one needs?

AMD will be in a trouble. A serious one.
 
4K8KW10 said:
AMD Ryzen 7000 Series ‘Raphael’ CPUs Leaked In New Roadmap – 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022
AMD Ryzen 7000 Series 'Raphael' CPUs Leaked In New Roadmap - 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022 (wccftech.com)

Click to expand...

Certainly hope Raphael is as good as everyone believes though, it lines up neatly with my plans for a major upgrade. Not that anyone can say what will be on the cards in 2 years time.

4K8KW10 said:
Including an iGPU that no one needs?

AMD will be in a trouble. A serious one.
Click to expand...

Maybe they know something about AM5 that they're keeping tight lipped.
 
4K8KW10 said:
Including an iGPU that no one needs?

AMD will be in a trouble. A serious one.
Click to expand...

eh? naw. iGPU's do enough for a large number of people, certainly not no one, and probably account for a majority of gpu ability in main desktop sales.

I wonder whether future CPU's will have a threadripper style chiplet layout with 4 chiplets around a central IO, and that the iGPU chiplet could be swapped into the place of a CPU chiplet. That way, you could have a standard layout across the range, with the GPU in one corner, then a combination of cpu chiplets added as needed .... then for the top line chip, the gpu gets dumped in place of another set of cores.
 
What's the point of two identical APUs - Raphael and Phoenix? :confused:

If Phoenix is enough with its iGPU for the intended market tiers, why would AMD want to lose its performance advantage against the Intel counterparts by wasting the precious die area for a weak GPU that no high-end or enthusiast buyer would want, instead of cramming the much needed more cores?
 
I think the slide has been put together from the previous slide, and it'll turn out to be partially gubbins.
why have a navi igup, when you can use those navi cores in 500 gfx cards, makes no sense
 
4K8KW10 said:
What's the point of two identical APUs - Raphael and Phoenix? :confused:

If Phoenix is enough with its iGPU for the intended market tiers, why would AMD want to lose its performance advantage against the Intel counterparts by wasting the precious die area for a weak GPU that no high-end or enthusiast buyer would want, instead of cramming the much needed more cores?
Click to expand...
It's possible the difference between "... Yes, it'll drive a desktop ok..." to "it's ok for light/medium gaming at 1080p".

Work machines Vs moderate home use, is my thinking.
 
q974739 said:
It's possible the difference between "... Yes, it'll drive a desktop ok..." to "it's ok for light/medium gaming at 1080p".

Work machines Vs moderate home use, is my thinking.
Click to expand...

I don't see how this is correct.

The serious work machines require serious discrete graphics cards, while most offices use no more than something like the Ryzen 5 2500U.
Moderate home use is also served by 7-watt, max 15-watt configurations.
 
4K8KW10 said:
I don't see how this is correct.

The serious work machines require serious discrete graphics cards, while most offices use no more than something like the Ryzen 5 2500U.
Moderate home use is also served by 7-watt, max 15-watt configurations.
Click to expand...
And the cost saving of 3000-odd cheaper CPUs, good for word processing only? Yep, I see a use.

My corporate overlords would love them for the rank and file... As they use their MS surfaces.
 
4K8KW10 said:
The serious work machines require serious discrete graphics cards, while most offices use no more than something like the Ryzen 5 2500U.
Click to expand...
Sounds like a serious niche...
None of my work machines have ever required a graphic card, although I did baulk at being expected to do dev work on a 8GB HDD (after demonstrating that power on until spinning up the dev environments took 20+minutes of thumb fiddling, they let me upgrade to 16GB and an SSD - all the others wanted that too then...). Zero need for a dGPU.
Been WfH the last year from my machine, again zero use for a dGPU.
Lots of memory, reasonable CPU grunt, fast SSD or NVMe... Yes to all of those.
GPU? Why? Onboard can drive 3 monitors what more could I need?
 
4K8KW10 said:
AMD Ryzen 7000 Series ‘Raphael’ CPUs Leaked In New Roadmap – 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022
AMD Ryzen 7000 Series 'Raphael' CPUs Leaked In New Roadmap - 5nm, Zen4, PCIe 5.0, DDR5 And Navi 2 Graphics Coming In 2022 (wccftech.com)





Including an iGPU that no one needs?

AMD will be in a trouble. A serious one.
Click to expand...

Tbh I like the iGPU because it helps with debug and problem solving. For instance my coilwhine is awful but I can't send my 6900xt back to get replaced as it my work PC and without an iGPU I have no GPU to run the system even to connect to VPN so am now stuck with a very coilwhine heavy card. If I could have had the same CPU with an iGPU I would have got that instead.
 
I find that slide intresting, AMD has never offered an iGPU on their enthusiast grade CPU's, they take up valuable die space on a Chip they know will be used with a discrete GPU.

Don't get me wrong i think its great if they have found some space in the IO die for it and it doesn't take anything away from the CPU, that IO die may change from 14nm GloFo to 7nm TSMC with Raphael so i guess its possible.

On APU's, the reason AMD have stuck with Vega for this long is System memory bandwidth, there isn't enough in DDR4 to justify the move to a much faster iGPU, it would bottleneck, however DDR5 has 50% more bandwidth at the same clock speed.
 
AMD's original plan with Fusion was to use the graphics shaders instead of the classic floating point units in order to accelerate the apps and make it a true heterogenous computing system.
Today, I have no idea why they will waste their so important performance advantage.
 
You must log in or register to reply here.
Back
Top Bottom