AMD's Future Developments: Ryzen 8000 and Navi 3.5
Click here to post a comment for AMD's Future Developments: Ryzen 8000 and Navi 3.5 on our message forum
Ryu5uzaku
AMD usually has had process node advantage or at least parity. After a good while they do not which is currently the 7900 series.
Most likely AMD also wasn't where they wanted to be versus their own cards either. But tbh if we remove the okish RT performance both 7900 are really good cards comparing to the new Nvidia also. Just that we expected more( at least I did except on RT side).
Only big jump from last gen is the ultra massive 4090.
Zen 5 has a good platform and it's a good release especially the big CPUs. Comparing to Intel. Also less power hungry.
In current climate one won't go wrong with building AMD rig only either. As long as they don't buy 7600 the monilithic pile of garbage.
Ryu5uzaku
H83
user1
given the untapped clock headroom on the 7900 series gpus, if they can get the power consumption under control, significant uplift is on the table without much work, so a respin makes alot of sense.
also,
chiplets are inferior to monolith in all domains except price, that is the point, it significantly reduces cost, that is why they do it. moving the memory controllers and the cache to a separate die is a clever move, as it allows them to share silicon allocation between many different products, and reallocate them as needed much more easily and much faster.
also leaves the door open to other kinds of products, for instance amd could build an HBM based MCD package, which could be " drop in" affair, without a creating a new die like with navi 12(hbm) vs navi 10(gddr6).
Denial
tunejunky
tunejunky
user1
https://occlub.ru/wp-content/uploads/2017/08/images_posts_news_2017_8_24_MCM-vs-monolithic-1.jpg
this is a slide based on zen 1, if you would to do a chip like genoa, it would be an even greater disparity, >1000 mm^2 monolithic chips are possible, but the cost increase is massive, definitely not worth single digit percentage increases in performance.
even for cpus monolith is better, but again you run into prohibitive cost, a monolithic epyc chip would use less power for instance, and have lower latency, but the cost is way higher, even on mature nodes.
cucaulay malkin
Com
it was the highest core count you could get. look where 7800x3d is on that application performance index compared to 7950x, for productivity it's like last gen 5900x asnd slower than new i5, that is also unlocked by the way.
Ryu5uzaku
https://tpucdn.com/review/amd-ryzen-7-7800x3d/images/efficiency-gaming.png
Where AMD currently does shine is efficiency tho. If one cares about it.
Tho gotta admit from Q6600 we have come a long way too. In terms of just inflation. Tho 7800x3D is funny since it's slower than 7700x in producitivity. I guess the 3D stack kind of sucks for clocks.
wavetrex
user1
moo100times
Crazy Joe
Horus-Anhur
tunejunky
user1
but this isn't true for all markets.
I don't really contest any of that, things like "wafer scale" have numerous problems, but it can be done as there are chips in serial production that use it, its just not cost effective unless you can charge big money, servers compete in the sub <10k per chip arena, and there, doing such a thing doesn't make sense, pharma
latest supercomputer listing as of May 23.
I think your listings might be incorrect. Here is the Ryu5uzaku
Crazy Joe
https://i0.wp.com/chipsandcheese.com/wp-content/uploads/2022/12/rdna3_cl_scalar_latency-2.png?ssl=1
As you can see, Ada Lovelace only loses the latency race between about 80 KB and 8000 KB, for the rest of the sizes it is either equally fast or faster.
Your graphs show Ampere vs RDNA3, not Ada Lovelace vs RDNA3.
This is the graph you should have used: