Overclock.net banner

[SiSoftware.net] Skylake-X i7-7900X Performance Leaked: 55% faster than i7-6950X @ 4.5GHz

17K views 109 replies 60 participants last post by  guttheslayer 
#1 ·
Quote:
i7-7900X (4.0GHz base 4.5GHz boost - running @ 4.0GHz on 9 cores and 4.5GHz on 1 core) - Score: 1386.94Mpix/s



i7-6950X @ 4.50 GHz on all 10 cores - Score: 897.28Mpix/s



i7-6950X (3.5GHz base 4.0GHz boost - running @ 3.5GHz on 9 cores and 4.0GHz on 1 core) - Score: 746.64Mpix/s

Source

Source 2

Source 3

The i7-7900X @ stock is 55% faster than i7-6950X @ 4.5GHz and 86% faster than i7-6950X @ stock; looking good Intel, some serious IPC gains here.
thumb.gif


The only other parts that come close to this performance is:
Quote:


]Intel(R) Xeon(R) Platinum 8176 CPU @ 2.10GHz (28C 56T 2.8GHz/3.8GHz, 2.4GHz IMC, 28x 1MB L2, 38.5MB L3) - Score: 2429.90Mpix/s



and

Intel(R) Xeon(R) Gold 6130 CPU @ 2.10GHz (16C 32T 3.4GHz/3.7GHz, 2.4GHz IMC, 16x 1MB L2, 22MB L3) - Score: 1356.70Mpix/s



These are the Skylake Xeons with new branding:

http://ranker.sisoftware.net/top_run.php?q=c2ffcee885e8d5e5c3b18cbc9af3ceffd9b18cbc9ae2dfeec8adc8f5c5e390ad9d&l=en

http://ranker.sisoftware.net/show_run.php?q=c2ffcee889e8d5e3d4e2d5e4d4f280bd8dabceab96a680f3cefe&l=en

http://ranker.sisoftware.net/show_run.php?q=c2ffcee889e8d5e3d4e2d5e4d4f280bd8dabceab96a680f3cefe&l=en

Now while these have way more L3, Cores and Threads, it seems that the Xeon Gold 6130 is scoring similarly to this i7-7900X. Other than that the only CPUs ranking close to it, below third place, are previous Gen Xeons with 22C/44T.
mad.gif
 
See less See more
2
#2 ·
The only way for Skylake-X to have serious IPC gains over broadwell is if it was an entirely different core than thee normal Skylake, which is highly doubtful as Intel has never done that to date.
This is explained more by a bad reading of the benchmark, not detecting cores and such correctly
 
#3 ·
Wow, suspiciously awesome. I wonder why it's reporting 175W when I thought these were supposed to be 160W. Is Intel gonna finally stop doling out 5-10% IPC increases?
 
#4 ·
Quote:
Originally Posted by EniGma1987 View Post

The only way for Skylake-X to have serious IPC gains over broadwell is if it was an entirely different core than thee normal Skylake, which is highly doubtful as Intel has never done that to date.
This is explained more by a bad reading of the benchmark, not detecting cores and such correctly
I'm wondering if this just isnt the result of having massively more L2 cache. Sorta like the CPU-Z benchmark when Ryzen was released.
 
#5 ·
There's no way the i9 7900X is 55% faster than 6950X at 4.5Ghz. Not all 6950X chips can even hit 4.5Ghz. D:
 
#7 ·
Ohhh the title is a little misleading. If you look at the tables and the clock speeds you'll see the actual comparison.

The 7900X @ 4.5Ghz is 55% faster than a 6950X @ 3.0Ghz.

That makes more sense.

EDIT: Nope I guess both are clocked at 4.5 Ghz. It's crazy if true.
 
#8 ·
Quote:
Originally Posted by Syan48306 View Post

Ohhh the title is a little misleading. If you look at the tables and the clock speeds you'll see the actual comparison.

The 7900X @ 4.5Ghz is 55% faster than a 6950X @ 3.0Ghz.

That makes more sense.
You are looking at the factory clock speeds listed in the top section. Neither of the 6050x tests were run at 3.0ghz.

Not sure that I have any faith in this.
 
#9 ·
Quote:
Originally Posted by Syan48306 View Post

Ohhh the title is a little misleading. If you look at the tables and the clock speeds you'll see the actual comparison.

The 7900X @ 4.5Ghz is 55% faster than a 6950X @ 3.0Ghz.

That makes more sense.
No, its presumably at stock with Turbo Boost 2.0 and 3.0 working as intended. 7900X @ 4.5GHz would imply an all core OC.

Edit: @bigjdubb, the second 6950x result is also presumably stock with boost doing its thing again.
 
#10 ·
Quote:
Originally Posted by bigjdubb View Post

Not sure that I have any faith in this.
Seconded
 
#14 ·
Quote:
Originally Posted by BinaryDemon View Post

Quote:
Originally Posted by EniGma1987 View Post

The only way for Skylake-X to have serious IPC gains over broadwell is if it was an entirely different core than thee normal Skylake, which is highly doubtful as Intel has never done that to date.
This is explained more by a bad reading of the benchmark, not detecting cores and such correctly
I'm wondering if this just isnt the result of having massively more L2 cache. Sorta like the CPU-Z benchmark when Ryzen was released.
Exactly what I was thinking. These CPUs will supposedly have 4x more L2 cache than before, at 1 MB per core vs 256 KB; if a certain benchmark benefits immensely from fitting inside the L2 cache, then it can skew the results - as far as real world general performance is concerned that is.

I'm really curious to see if Intel was leaving any general IPC on the table with mainstream Skylake though, or if this will only increase very specific workloads' performance.
 
#19 ·
55% faster

...in this one benchmark that probably likes l2 cache
 
#20 ·


If we ignore the aggregate results, and look at individual test parts....it starts to look like a cherry picked test. I am just going to poke a guess that the cores themselves have seen typical IPC like the 6700k-7700k, and we are really just seeing a cache test..

I am not complaining! I have a feeling that there are going to be some games that see some pretty good gains, or at least will have a bottleneck removed. Specifically, looking at Bethesda's Gamebryo engine and the way it handles shadows, but surely others as well.
 
#21 ·
Quote:
Originally Posted by Jim86 View Post

R.I.P. AMD
smile.gif
Yea, I'm sure people looking to buy AMD's $300-$500 line is really upset about this $1500 chip.
 
#22 ·
10MB L2 definitely did the trick, even broadwell-C's 128MB eDRAM cache did some magical things on certain benchmarks.

Quote:
Originally Posted by tpi2007 View Post

I'm really curious to see if Intel was leaving any general IPC on the table with mainstream Skylake though, or if this will only increase very specific workloads' performance.
its possible but unlikely, the things that could've been different are the instruction sets and certain xeon-only pipeline tweaks.
 
#24 ·
I'm not buying this for one second, though if true it would be a massive feat by Intel. Anyway, I'll believe it when I see it.
 
#25 ·
Less massive feat, more deliberate sandbagging if true.
wink.gif
 
  • Rep+
Reactions: Majin SSJ Eric
#26 ·
I'd be happy to finally see a proper performance increase from Intel for a change but I'm taking this with a mountain of salt and being realistic.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top