Ryzen 4000 APU / Zen3 News

I have some questions for which I know we don't have answers - but I want your considered opinion:

1. If it is announced in Jan at CES2020 - do we expect immediate availability, or might it be closer to March/April before we see them in the shops?

2. Do we expect to see the launch of a mid-range motherboard chipset at the same time? Doesn't make sense to sell $150 APU's with $200 x570 motherboards...

3. Given i'm considering a 6-year cool-n-quiet htpc - is there any hope the video-decode and display-output portions will support AV1 decode and HDMI 2.1?

4. Do we expect the supported memory speed to increase beyond the DDR4 3200 used by Zen2/Ryzen3000 parts? We know that IF/mem-speed is fine one-to-one up to 3600 speeds, and APU parts do tend to arrive with faster memory speed support to feed the GPU portion. I ask because i'm not interested in overclocking - and 2x8GB DDR4 3200 is really cheap right now!

5. We 'know' the GPU is vega derived, but on 7nm. Do we have feel for the level of improvement in gaming performance offered over a 12nm Zen+ 3400G? For reference - my current haswell i5 mITX htpc has an AMD r7 270x in it (pitcairn).
 
Just wild guesses based on available info / past behavior

1) late Q1 is my guess
2) Lot of 300 series motherboards out there could probably handle these APUs just fine, and it's unlikely the target market would be willing to pay X570 prices for things like PCIe 4, etc, as you've said. My guess is - no new chipsets, just a "refresh" of 300/400 low end.
3) probably not, even if it's navi-based, since nothing announced so far supports HDMI 2.1.
4) maybe? Although that would mean an update to the i/o chiplet, and I'd think that's a refresh that would happen with "zen2+" or "zen3".
5) depends entirely on number of CPU cores, number of GPU compute units, speed of system memory. In other words: no idea. But probably safe to assume it will be faster.
 
5. We 'know' the GPU is vega derived, but on 7nm. Do we have feel for the level of improvement in gaming performance offered over a 12nm Zen+ 3400G? For reference - my current haswell i5 mITX htpc has an AMD r7 270x in it (pitcairn).

New Vegas are basically quite a few navi features added into them to keep them up to date. I wouldn't expect HDMI 2.1 but it would be nice...
 
You do know that Ice Lake architecture is immense better than skylake. It starts at about 20% better IPC but goes up with the number of cores. I bet the 38 core is better than the 48 cores but Intel wasn't probably sure the 10nm will be ready so they made piles of 48 cores they have to sell, maybe at a smaller price than the 38 cores 10nm.
In fact ICe lake architecture has much better IPC than Zen 2.

IPC is the Instructions Per Clock. It's the maximum instructions per clock cycle a chip can do. The IPC always remains the same. The number of cores means there are more cores operating at the same IPC. Adding more cores doesn't increase the IPC of each core. It never has, does not, and never will.
 
4. Do we expect the supported memory speed to increase beyond the DDR4 3200 used by Zen2/Ryzen3000 parts? We know that IF/mem-speed is fine one-to-one up to 3600 speeds, and APU parts do tend to arrive with faster memory speed support to feed the GPU portion. I ask because i'm not interested in overclocking - and 2x8GB DDR4 3200 is really cheap right now!

5. We 'know' the GPU is vega derived, but on 7nm. Do we have feel for the level of improvement in gaming performance offered over a 12nm Zen+ 3400G? For reference - my current haswell i5 mITX htpc has an AMD r7 270x in it (pitcairn).

Renoir will have more Vega units than 12nm APU. That's already counting on the DDR4 to be faster and on the 6 cores (so it seems) not to bottleneck the shared memory.
 
IPC is the Instructions Per Clock. It's the maximum instructions per clock cycle a chip can do. The IPC always remains the same. The number of cores means there are more cores operating at the same IPC. Adding more cores doesn't increase the IPC of each core. It never has, does not, and never will.
Okay. Don't be dumb !
The chips with multiple cores don't work as efficiently per core with the number of cores raising and busy and that depends on the architecture. Ice lake seems to be improved on that matter against Skylake and Ryzen 3000 series and Threadripper seem also to be better at this than Skylake. This is happening because of cache limitations or enhancements, bandwidth and plenty of other reasons, including OS support by the architecture, and also the use one program makes and which instructions used make a difference.
So the difference in IPC is something you find by making sophisticated tests, not something strait to test, precise, predictable from some specs and scalable. Some people working at AMD are saying, for instance, that Zen 3 has 10% improved IPC on integer instructions and up to 50% improvement on floating instructions. Will have to test to find the average result. Also the 8 core will be on a single CCX so probably IPC (it's per core) on multiple core use (tested while all cores used) will improve over Zen 2 even more in that case.
 
Renoir will have more Vega units than 12nm APU. That's already counting on the DDR4 to be faster and on the 6 cores (so it seems) not to bottleneck the shared memory.
Do we dare hope for DDR4 4000 support, or do you believe it is more likely to be a mild bump up to DDR4 3600?

Interesting that we're three weeks away from an announcement and we still know very little...
 
You already have that covered.
Wew lad, savage.
Memri - what is the cure for disorders.png
 
the 3400g came with DDR4 2933 support, and that only had the job of feeding 4C/8T and 704 shader units.

now we're looking at 8C/16T and and ~900 shaders at 33% plus clockspeed, doesn't feel like DDR4 3200 is a sufficient increase to feed this increased requirement...
 
the 3400g came with DDR4 2933 support, and that only had the job of feeding 4C/8T and 704 shader units.

now we're looking at 8C/16T and and ~900 shaders at 33% plus clockspeed, doesn't feel like DDR4 3200 is a sufficient increase to feed this increased requirement...
Maybe with more cache it's okay. Mind that the Zen APU have less cache per core on the CPU part than Zen CPU. Problem would be rather with the GPU but then again, there could be improvements based on what was done for Navi. However I'm betting on more IPC at lower clocks, higher core count and very low TDP for Renoir because this is what AMD needs to do to beat Intel on laptops and it seems AMD never missed its goal since Zen.
 
Back
Top