News 52-core Xeon 8470 Demolishes Threadripper 3990X in New Benchmarks

Apr 29, 2022
1
2
10
So, the new 52 core chip is faster than the 3990x 64 core chip (a two year old CPU). Not that impressive! Or is it just me?
Absolutely pointless article and benchmark comparison. Why not compare the Xeon to a new Epyc chip?
Just makes Intel seem like a tryhard. But hey, at least they're now faster than a 2yo discontinued HEDT desktop cpu. Yay! Congrats Intel! 🥳
 

jacob249358

Commendable
Sep 8, 2021
636
215
1,290
So, the new 52 core chip is faster than the 3990x 64 core chip (a two year old CPU). Not that impressive! Or is it just me?
yeah especially because sapphire rapids is for servers not workstations like the thread rippers. And Im sure the 5990wx or whatever will beat it anyways. I'm hoping the 5000 thread rippers will have come decent single core so it can replace my gaming PC and be my workstation.
 
  • Like
Reactions: Roland Of Gilead

KyaraM

Admirable
If AMD was not around, we would see minimal improvements each year.
If AMD had not been stupid, we wouldn't even have seen that happen. Oooorrrr we would be in the same situation now, since improvement is limited. Take a guess.

Also, at least Intel did improve things, of only a little. When did Vermeer drop again? Oh, right. Almost 2 years ago. I wrirdly seem to remember that Intel did at least something in those two years. Funny.
 
  • Like
Reactions: jacob249358
D

Deleted member 14196

Guest
Absolutely pointless article and benchmark comparison. Why not compare the Xeon to a new Epyc chip?
Just makes Intel seem like a tryhard. But hey, at least they're now faster than a 2yo discontinued HEDT desktop cpu. Yay! Congrats Intel! 🥳
And the answer why not to compare those two is because the Intel would lose and loose badly
 
  • Like
Reactions: Roland Of Gilead

spongiemaster

Admirable
Dec 12, 2019
2,278
1,281
7,560
Absolutely pointless article and benchmark comparison. Why not compare the Xeon to a new Epyc chip?
Just makes Intel seem like a tryhard. But hey, at least they're now faster than a 2yo discontinued HEDT desktop cpu. Yay! Congrats Intel! 🥳
The leaker didn't post a comparison, just the results for the Intel system. The obvious answer to why Milan wasn't chosen for the comparison is because there are no posted results for the same tests. If the results are out there, why don't you go find them and post them here for the rest of us to see?
 
  • Like
Reactions: LolaGT

hosea_chan

Reputable
Jan 16, 2019
3
0
4,510
This 79k Cinebench R23 score is in a 2S configuration, which means a total of 104 cores are used. This is the only reasonable explanation for a 52-core part having 74x multi-core score over the single-core score. And for a fair comparison in the server sector, EPYC 7663 in 2S configuration has a Cinebench R23 score of 113k. (Source) Now 113k vs 79k is what we call demolishing. Hopefully Team Blue will come up with something more impressive when the actual product comes out.
 
If AMD had not been stupid, we wouldn't even have seen that happen. Oooorrrr we would be in the same situation now, since improvement is limited. Take a guess.

Also, at least Intel did improve things, of only a little. When did Vermeer drop again? Oh, right. Almost 2 years ago. I wrirdly seem to remember that Intel did at least something in those two years. Funny.
I know. AMD should've just kept on pushing clocks and power usage and not even bothered trying to design a better, faster and more efficient CPU.
 
This is the only reasonable explanation for a 52-core part having 74x multi-core score over the single-core score.
Or, you know, it's an es so it may have extremely limited single core boost, a 12900k at around 5Ghz has twice that score at around 2000 points.
https://www.tomshardware.com/review...-i5-12600k-review-retaking-the-gaming-crown/7

Or a single core just can't take full advantage of all the cache and ram of the system.
So there you go, at least two other reasonable explanations.
 
  • Like
Reactions: KyaraM

hosea_chan

Reputable
Jan 16, 2019
3
0
4,510
Or, you know, it's an es so it may have extremely limited single core boost, a 12900k at around 5Ghz has twice that score at around 2000 points.

I agree that it is an ES and I hope Intel would (and they should) come up with a better one in the end. It is still too early to judge the Intel product but the article title is way too misleading.

The 2000 single-core score at 2000 is way overboard for a server grade CPU because it would not practically be over clocked to 5GHz (cost-efrectiveness and of course, thermals). A 8P+8E with E-cores not overclocked and a full 52-core configuration are two different stories.

Or a single core just can't take full advantage of all the cache and ram of the system.

I seriously hope you are just trolling with this point. This means the Intel engineers have done a terrible job with their cores if what you said were true.
 
The 2000 single-core score at 2000 is way overboard for a server grade CPU because it would not practically be over clocked to 5GHz (cost-efrectiveness and of course, thermals). A 8P+8E with E-cores not overclocked and a full 52-core configuration are two different stories.
It's a single core that has to hit 5Ghz not all the cores, it doesn't need a lot of power or does it increase temps.
 
  • Like
Reactions: KyaraM