• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Speculation: Ryzen 4000 series/Zen 3

Page 186 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
jamescox said:
Well, it only has a couple red LED lights (no RGB), which seem to be only on and off, so it is kind of display limited. If they could get my dishes cleaned better and more quietly, then I guess put a 5 nm chip in there.
Click to expand...

You should have purchased a samsung. At least it would play a song for you when it finishes it's wash cycle. Overclocking it could have made it polyphonic!
 
LightningZ71 said:
GloFo, by choosing to cease development work on their leading edge node processes, has essentially written their epitaph, just years ahead of when their demise finally happens.
Click to expand...
SOI wafers are largely dependent on SOI providers. So, GloFo's node introduction on FDX is largely built on SOITEC providing.

GlobalFoundries' FinFETs are like UMC's FinFETs. So, them getting money from leading edge FinFETs is pretty much like UMC getting money from leading edge FinFETs.
UMCFF.png

Just because it exists doesn't mean there is demand for it. GloFo can just force AMD to use it and still not make a profit as they did with 14LPP/12LP. Fab8 basically has run at a loss from 2014 onward.

Whereas 22FDX has been a profit.

22FDX+ apparently has removed the sandbagging that in turn reintroduces high performance to the platform. Before, 12FDX would have popped up which was suppose to be the high performance driver.

I see AMD abandoning GlobalFoundries for most if not all modern products. Following the N-1 rule, instantly AMD does 5nm it means all HVM APUs will be on 7nm or smaller.

The only products AMD has with GlobalFoundries that are available for new product replacements are for Stoney (Entry-level Chromebooks) and Polaris 24 (Iceland/Topaz). Both of which launched 2019 and are still 28nm. We have yet to see small CU/WGP RDNAx and I'm withholding statements on the other till Zen3. Specifically, if the rumor of the FPU co-processor model finally being shuttered. General Purpose gets to use all Map/NSQ(Unified NSQ Map) or Computational gets to use all those resources. No more 50/50 Int/Fp to get all resources used.
 
Last edited:
So I have been busy. I visited briefly to see if anything interesting popped up, yet the first thing I see is nonsense about Intel 10nm and AMD 5nm not happening until 2022.

You guys need to bury the hatchet.

Alder Lake and Rocket Lake will be neck-in next performance wise regardless of the process. While I would love to see AMD beat Intel in ST performance by 20%, that isn’t going to happen. They will be a single digit percentage between each other and I suspect Intel will somehow take the lead in multiple specific workloads just like with Comet Lake.
 
Didn't see this posted, but this is a possible Zen 3 5800x benchmark and it appears to be faster than a stock 10900K in the Ashes of the Singularity CPU benchmark:

EjDiQzaVkAU9-Qy


APISAK on Twitter: "R7 5800X VS i9-10900K https://t.co/eXLc9WmMz8" / Twitter
 
I'm not familiar with Ashes benches, but this looks Crazy good.

The important metric is CPU framerate. Here is the 5800x with RTX 2080 with Crazy_1080p preset (btw the benchmark itself is still up):
AMD-Ryzen-7-5800X-Ashes-of-the-Singularity-Single-Run-1200x571.png


Here is the best 10900K result i could find with the same GPU (again, look at the green graphs, the GPU might be a higher clocked model, etc):

OWDJxOb.png


Here is the best 3900X with the same GPU:
JCtBboD.png

And here is the best 3900X with a RTX 3090, to prove that the while GPU score increases a lot (75 -> 121 FPS) the CPU score barely moves the needle (116 -> 121) :
1yaPNaf.png

TL;DR:

Crazy_1080p CPU Framerates (all batches) with RTX 2080:

CPUGPUAverage CPU Framerate (All Batches)% of 10900K performance
Ryzen 3900XRTX 2080116.4 FPS92.6%
Ryzen 3900XRTX 3090121.5 FPS96.6%
Intel 10900KRTX 2080125.7 FPS100%
Ryzen 5800XRTX 2080142.4 FPS113%

Bottom line:

While it's only one benchmark (that's usually even a bit AMD favored) that is still an extremely strong result.
 
Last edited:
Gideon said:
I'm not familiar with Ashes benches, but this looks Crazy good.

Crazy_1080p CPU Framerates (all batches) with RTX 2080:

CPUGPUAverage CPU Framerate (All Batches)% of 10900K performance
Ryzen 3900XRTX 2080116.4 FPS92.6%
Ryzen 3900XRTX 3090121.5 FPS96.6%
Intel 10900KRTX 2080125.7 FPS100%
Ryzen 5800XRTX 2080142.4 FPS113%
Click to expand...

Looks like a huge improvement in CPU framerates (Green bars). Comparing 5800X vs best 3800X:

AS_5800x_vs_3800x.jpg

GPU Benchmark - Ashes of the Singularity

Video Card GPU Benchmarks, High quality testing of your video card and gaming system.
www.ashesofthesingularity.com www.ashesofthesingularity.com

GPU Benchmark - Ashes of the Singularity

Video Card GPU Benchmarks, High quality testing of your video card and gaming system.
www.ashesofthesingularity.com www.ashesofthesingularity.com
 
Last edited:
Gideon said:
While it's only one benchmark (that's usually even a bit AMD favored) that is still an extremely strong result.
Click to expand...

+1. And this is an 8 core CPU vs a 10 core CPU from Intel.

If AMD nails gaming performance this time, Intel's back will be up against the wall for sure......at least until Rocket Lake and Alder Lake come out.

Just a few more days until the big reveal! I was honestly shocked to find that benchmark. It really satisfied the intense itch I've been having to see some kind of Zen 3 benchmark! 😀

AMD did a great job this cycle in controlling leaks, but I guess now the cat is out of the bag and hopefully there will be a huge torrent of benchmarks before the official reveal.
 
What a beating! Awesome!

Even more so considering those leaks are usually done on crappy unoptimized settings (low fclk/mclk, subpar memory timings, etc) so there's probably even more headroom in that 5800x to keep scaling, at least in games.

I guess there should be more leaks, about a week left for Oct 8. Launch day reviews will be quite a fun read.
 
uzzi38 said:
For comparison I offer pretty much close to the best 8 core scores you'll see on Intel's side right now.

This is a 10700K with a 5GHz all core OC using DDR4-4000cl14... and it's ~7% better
Click to expand...

Where did you get the info about overclock and DDR? As the Ryzen benchmark is still up maybe some info can be dug out about that as well?
 
Anyway, if this bench is even remotely representative of overall gaming perfomance, Intel will lose a ton of mindshare and AMD should gain some even in GPUs.

Gaming perfomance has been their saving grace for 3 years.
 
From that VTune profile we know that AotS loves tight memory subsystem, so great to see AMD finally has one. I find it hard to explain the results with just unified L3 cache, maybe something more is going on, like straight up better latency of memory.
 
JoeRambo said:
From that VTune profile we know that AotS loves tight memory subsystem, so great to see AMD finally has one. I find it hard to explain the results with just unified L3 cache, maybe something more is going on, like straight up better latency of memory.
Click to expand...
Zen3 is a new uarchitecture, it's not just the unified L3 cache. The new CCX design surely helps lower the memory latency but there are some significant changes that will show up more in some specific tests. Then there is better power/clock management that will keep the clocks higher for longer periods of time, all of this adds up. I'm afraid AMD will jack up the prices of initial 5000 series SKUs, for cheaper(lower end) parts we'll have to wait a few months.
 
inf64 said:
Zen3 is a new uarchitecture, it's not just the unified L3 cache. The new CCX design surely helps lower the memory latency but there are some significant changes that will show up more in some specific tests. Then there is better power/clock management that will keep the clocks higher for longer periods of time, all of this adds up. I'm afraid AMD will jack up the prices of initial 5000 series SKUs, for cheaper(lower end) parts we'll have to wait a few months.
Click to expand...

When workload is 30% memory latency bound ( that is mem op missed L1, L2, L3 and is now pending from memory), the best way to improve it, is to tighten memory latency. uArch can help here with more outstanding memory operations, speculation magic, but I think it is safe to say that ZEN3 definitely has better memory latency than previous AMD designs.

AotS workload has characteristics similar to stuff like Linpack ( except just 33% of FP utilization ). Easy to scale Linpack flops with memory BW/latency. Intel Skylake can go +25% GFlops just from tightening memory.
 
JoeRambo said:
From that VTune profile we know that AotS loves tight memory subsystem, so great to see AMD finally has one. I find it hard to explain the results with just unified L3 cache, maybe something more is going on, like straight up better latency of memory.
Click to expand...

Well, that is what you get from 15% higher IPC.It is expected, if CPU core has several additional improvements.
 
Asterox said:
Well, that is what you get from 15% higher IPC.It is expected, if CPU core has several additional improvements.
Click to expand...

Isn't IPC tangential to memory latency/BW ? ZEN2 already has IPC advantage over Skylake cores and very sizable deficit in games and AotS benchmark.
My point is, to get results these good, AMD must have done something to memory subsystem, like cutting quite some latency, core/L3 redesign comes as part of it and on top of that.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top