• 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.

***OFFICIAL*** Ryzen 5000 / Zen 3 Launch Thread REVIEWS BEGIN PAGE 39

Page 28 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Good god do we even know reg. Threadripper Zen 3 ? I feel for Intel now 😕

Zen3 Threadripper will be something like the first half of this gif:

hs1vo02qxwr51.gif
 
Man, if 5900x has 50% boost in AVX2 workloads I may actually get that instead of used 3950x. My heaviest use is BluRay to HVEC encoding which is highly AVX2 dependent. 3950x would provide 33% more cores compared to 3900X that I have, but 50% boost in AVX2 should make 5900x faster than 3950x in that particular area. Very excited to see actual benchmarks.
 
Man, if 5900x has 50% boost in AVX2 workloads I may actually get that instead of used 3950x. My heaviest use is BluRay to HVEC encoding which is highly AVX2 dependent. 3950x would provide 33% more cores compared to 3900X that I have, but 50% boost in AVX2 should make 5900x faster than 3950x in that particular area. Very excited to see actual benchmarks.

It makes sense. Unlike rendering encoding (at least H264 and H265) doesn't really benefit much from core-count past ~10-12 cores. At the very least it likes strong cores a lot more than more cores.

I mean just compare a 3990x to a 3950x The 64 core 3990x unsurprisingly absolutely smokes the 16 core 3950x in tasks like V-ray:
1603135668089.png

Yet in HEVC encoding, there barely is any difference:
1603135747512.png

Hell even the 10900K is really close (while losing significantly in rendering):

1603135476420.png
Base on these results I wouldn't be surprised at all if 5900X beats 3950X in this task.
 
AM4 has had a pretty great run.

From Bristol Ridge in 2016 through to Zen3 in 2020. Even within the Zen family, your looking at what? 50+% IPC improvement?
Meanwhile, what was the IPC gained by Intel on LGA 1151 (2015) -> LGA 1200 (2020)?
 
I assume since it is a new design that an improved AGESA in a few months from now may bring an extra uplift. (maybe for tests that didn't improve much over Zen2)
 
I assume since it is a new design that an improved AGESA in a few months from now may bring an extra uplift. (maybe for tests that didn't improve much over Zen2)
Yes. The new AGESA code is out in the wild in Beta BIOSes for a fair number X570 boards.
 
It makes sense. Unlike rendering encoding (at least H264 and H265) doesn't really benefit much from core-count past ~10-12 cores. At the very least it likes strong cores a lot more than more cores.

I mean just compare a 3990x to a 3950x The 64 core 3990x unsurprisingly absolutely smokes the 16 core 3950x in tasks like V-ray:
View attachment 32022

Yet in HEVC encoding, there barely is any difference:
View attachment 32023

Hell even the 10900K is really close (while losing significantly in rendering):

View attachment 32016
Base on these results I wouldn't be surprised at all if 5900X beats 3950X in this task.
I don't know how AT did HEVC testing, why would anyone upconvert 1080p30 to 4K60? That's not something anyone would ever do. But in my own testing ffmpeg HEVC scales linearly up to 8 physical cores. Which is why I run 2 instances of ffmpeg on my 3900x at the same time to maximize throughput. Either way 3950x or 5900x would be a big improvement, I guess I'll see which one would give me most HEVC performance for the buck, maybe rumored 12c24t 65W 5850x would be the best deal if it turns out to be real. Either way, exciting times, super cool to see actual meaningful CPU performance increases every 18 months.
 
What we call here "hype", there at Intel is "fear".
If the theory that AMD just tweaked the FPU instead of making it bigger is true, imagine how much can AMD still improve the Zen Family in the future generations? Zen 4 should be brutal and if Intel can't get things working again in time they're in serious trouble.
 
Last edited:
What we can here "hype", there at Intel is "fear".
If the theory that AMD just tweaked the FPU instead of making it bigger is true, imagine how much can AMD still improve the Zen Family in the future generations? Zen 4 should be brutal and if Intel can't get things working again in time they're in serious trouble.
I think Intel before 2017 was in "5% improvements per gen is enough" mode. Them Ryzen 1 struck and I believe only then They started revving their engines for more performance. But, as a new CPU generation takes five years from inception to final product, we might see Intel bearing fruit to that, hopefully, sometime in 2022... If all goes clockwork. But as the infighting goes on in there, maybe 2023? 🙂
 
I think Intel before 2017 was in "5% improvements per gen is enough" mode. Them Ryzen 1 struck and I believe only then They started revving their engines for more performance. But, as a new CPU generation takes five years from inception to final product, we might see Intel bearing fruit to that, hopefully, sometime in 2022... If all goes clockwork. But as the infighting goes on in there, maybe 2023? 🙂

This isn't really an Intel thread, but Intel's in more trouble than that. AMD has done pretty well up to this point, making it hard for anyone to catch up. And AMD has a future process roadmap, unlike Intel. With them you just don't know what process they'll be using or when anymore.
 
Geekbench 4 for the 5800X and 5900X:

View attachment 32134

Memory clock is 2133MHz (4266 Mhz actual) :O I wonder if they ran it in 1:1 FCLK mode, but i somewhat doubt it considering latency is almost exactly the same as my 3700X old run with same GB4 version @ 1733 Mhz:
5800X vs 3700X (Geekbench 4.3.4)

The 5800X vs my 3700X scored 27% better in ST test but "only" 18% better in MT test (probably some improvement there possible).

Still very strong results, especially when considering that this is an old version of GB (4.4.x improved scores all-around).
 
Wonder if it can do all core 5Ghz OC with normal cooling. Remember last year the 5Ghz hype train collapsed, I was also a victim *facepalm*

If the unrestricted per core overclock/voltage control hold true you might be able to get a couple of cores to 5GHz.

Guessing around 5.8-6GHz on LN2 for the crazies!
 
Back
Top