AMD Ryzen Threadripper 1950X Benchmarks Show It 45%+ Ahead Of The Skylake X Competition

[Mr_Troll]

Quote

AMD Ryzen Threadripper 1950X Benchmarks Show It 45%+ Ahead Of The Skylake X Competition

 

The benchamrks in question include four tests. Three of the tests were conducted in both multi-threaded and single-threaded mode whilst the fourth is multi-threaded only

 

 

No surprises here, the 16 core, 32 thread Ryzen Threadripper 1950X absolutely creams its similarly priced 10 core competition from Intel.  In fact, Threadripper manages to establish a monstrous lead, it came out ahead by no less than 42% in Cinebench R15 and no less than 47% in all three other tests.

Threadripper’s lead is so gigantic that if it weren’t for its price you would be hard pressed not to believe it was in an entirely different tier than the i9 7900X. At this juncture it wouldn’t surprise us at all if Intel ends up sharply cutting prices to its newly introduced 10 core in the coming weeks.

Moving on to the single threaded side of the tests.  Here, the Core i9 7900X manages to save some face thanks to its significantly higher single core Turbo clock speed — 4.5GHz — and ends up ahead by anywhere from 3-9% depending on the test. What’s truly astonishing however is that despite the clock speed advantage for the 7900X, Threadripper still manages to stay right on its tail.

This goes to show that there’s really not much in it between AMD’s Zen microarchitecture and Intel’s Skylake microarchitecture. Rather what’s giving Skylake X the edge in single threaded work seems to be primarily the result of higher clock speeds due to Intel’s mature 14nm process. It could be only a matter of time before AMD catches up to Intel’s clock speeds as Globalfoundries’ 14nm process matures. Although speaking of the here and now, Threadripper appears to already be a massive win for AMD.

 

 

Quote

 

TR has higher single core score than ryzen?.. hows that possible? 

 

Source:http://wccftech.com/amd-ryzen-threadripper-1950x-benchmarks-leaked-42-faster-competition/

 

[tom_w141]

3 minutes ago, Mr_Troll said:

TR has higher single core score than ryzen?.. hows that possible? 

Because Geekbench = Trash

 

Btw leaks i've seen put multi core between 3200-3300 with 3000+MHz RAM

 

EDIT: Also why is this all within a quote?

[Billy_Mays]

DIE SLUGS DIE

[Damascus]

2 minutes ago, Mr_Troll said:

TR has higher single core score than ryzen?.. hows that possible? 

 

1 minute ago, tom_w141 said:

Because Geekbench = Trash

I assume that TR is made of only the highest binned chips, combine that with better power delivery, cache and everything else and TR could easily push past a mainstream r7 processor.

[ApolloFury]

Does quad channel help in single-thread performance?

[Bananasplit_00]

Well isn't that R15 score a bit low? Maybe the ram was slower then Linus and AMD used? 

[TheCherryKing]

The Core i9-7980XE will outperform the Threadripper 1950X assuming there are not problems with thermal throttling(which there will probably be).

[ARikozuM]

I'm going to assume these benches use 3000+ MHz RAM and overclocks until proven otherwise. 

 

Edit: How is a 16-core CPU getting about 30% more performance over a 10-core? Shouldn't it be closer to 50%?

[PocketNerd]

16c/32t CPU vs 10c/20t CPU...

I WONDER WHICH WOULD WIN?!

More news @ 11

(That said, GO-GO ThreadRipper!)

[Billy_Mays]

Just now, PocketNerd said:

16c/32t CPU vs 10c/20t CPU...

I WONDER WHICH WOULD WIN?!

More news @ 11

Today?

[PocketNerd]

Just now, Billy_Mays said:

Today?

So long as you can follow-up with a new and exciting product that can be mine for 12 easy payments

[Billy_Mays]

Just now, PocketNerd said:

So long as you can follow-up with a new and exciting product that can be mine for 12 easy payments

Third time today

[Taf the Ghost]

6 minutes ago, ApolloFury said:

Does quad channel help in single-thread performance?

If the test can use the Quad-channel memory bandwidth, then it should.

3 minutes ago, Bananasplit_00 said:

Well isn't that R15 score a bit low? Maybe the ram was slower then Linus and AMD used? 

R15 Multi-core score is in line with expectations for a non-OC'd TR. Though maybe a little higher than others will get. 3000-3200 has been the expected range.

2 minutes ago, TheCherryKing said:

The Core i9-7980XE will outperform the Threadripper 1950X assuming there are not problems with thermal throttling(which there will probably be).

In theory it should (2 more cores), but that's going to depend on clocks. At stock, the 7980XE should max all-core Turbo to 3.4 Ghz, but we also don't know how well it's going to scale across the Mesh. (Bandwidth limitations?) We'll find out when it lands, but I would expect ~3500 to 3600 R15 scores for the 7980XE.

[ARikozuM]

Just now, PocketNerd said:

So long as you can follow-up with a new and exciting product that can be mine for 12 easy payments

[Princess Luna]

2 minutes ago, TheCherryKing said:

The Core i9-7980XE will outperform the Threadripper 1950X assuming there are not problems with thermal throttling(which there will probably be).

Every one knows this already, it is a 2000 dollars processor it HAS to beat it, not to mention 2c/4t extra etc... but this is not the point, who here really is soo blind seeking bleeding edge performance? you can actually build an entire 1950x system with motherboard, ram, psu at the cost of the 7980xe alone.

[Jito463]

Just now, Billy_Mays said:

Third time today

Maybe it's time to accept reality, and just ask the mods to change your name to Billy_M.  

[TheCherryKing]

Just now, Taf the Ghost said:

If the test can use the Quad-channel memory bandwidth, then it should.

R15 Multi-core score is in line with expectations for a non-OC'd TR. Though maybe a little higher than others will get. 3000-3200 has been the expected range.

In theory it should (2 more cores), but that's going to depend on clocks. At stock, the 7980XE should max all-core Turbo to 3.4 Ghz, but we also don't know how well it's going to scale across the Mesh. (Bandwidth limitations?) We'll find out when it lands, but I would expect ~3500 to 3600 R15 scores for the 7980XE.

I guess the 4.4 GHz max turbo is only for one core. The 7980XE should perform a little less than Dual Xeon E5-2695v3.

[Zyndo]

10 minutes ago, Mr_Troll said:

Snip

Would be good to see how these things perform in the real world, after appropriately high overclocks are accounted for (that is, what is reasonably attainable for most people, as well as consideration for a delid on the 7900x). stock operation is fine... Benchmarking programs are fine... but these things are not indicative of true performance levels, especially since some benchmark programs merely spit out a number based on an arbitrary formula which is not always on a linear scale.

 

But at the end of the day, any "benchmarks" we discuss before the actual launch cannot be trusted. Wait for the official launch and there will be plenty of benchmarks from dozens of users/reviwers out there to really get an idea of how good these CPU's are head to head in a variety of synthetic and real-world tests with varying RAM speeds, overclocks, and system configurations.

[Taf the Ghost]

4 minutes ago, ARikozuM said:

I'm going to assume these benches use 3000+ MHz RAM and overclocks until proven otherwise. 

 

Edit: How is a 16-core CPU getting about 30% more performance over a 10-core? Shouldn't it be closer to 50%?

Cinebench & RealBench are both around 45% for the 1950X over the 7900X. Given the core difference (16c vs 10c) and clock differences (7900X should be running higher clocks) & some IPC difference, 45% is about correct. Geekbench is trash and I still don't know why anyone uses it.

[Billy_Mays]

Just now, Jito463 said:

Maybe it's time to accept reality, and just ask the mods to change your name to Billy_M.  

I'm fine with it but atleast I know people know Im here 

[Taf the Ghost]

2 minutes ago, TheCherryKing said:

I guess the 4.4 GHz max turbo is only for one core. The 7980XE should perform a little less than Dual Xeon E5-2695v3.

The charts cropped up in the other thread about the SKL-X release for the full turbo speed.

[Taf the Ghost]

3 minutes ago, TheCherryKing said:

I guess the 4.4 GHz max turbo is only for one core. The 7980XE should perform a little less than Dual Xeon E5-2695v3.

 

[Bananasplit_00]

4 minutes ago, Taf the Ghost said:

If the test can use the Quad-channel memory bandwidth, then it should.

R15 Multi-core score is in line with expectations for a non-OC'd TR. Though maybe a little higher than others will get. 3000-3200 has been the expected range.

In theory it should (2 more cores), but that's going to depend on clocks. At stock, the 7980XE should max all-core Turbo to 3.4 Ghz, but we also don't know how well it's going to scale across the Mesh. (Bandwidth limitations?) We'll find out when it lands, but I would expect ~3500 to 3600 R15 scores for the 7980XE.

Linus showed it putting out over 3200 at LTX I'm pretty sure, and so has AMD in an official video a while back. 

[Curufinwe_wins]

6 minutes ago, Taf the Ghost said:

Cinebench & RealBench are both around 45% for the 1950X over the 7900X. Given the core difference (16c vs 10c) and clock differences (7900X should be running higher clocks) & some IPC difference, 45% is about correct. Geekbench is trash and I still don't know why anyone uses it.

 

12 minutes ago, ARikozuM said:

I'm going to assume these benches use 3000+ MHz RAM and overclocks until proven otherwise. 

 

Edit: How is a 16-core CPU getting about 30% more performance over a 10-core? Shouldn't it be closer to 50%?

 

 

Also Amadahl's Law (and for that matter "Braess's paradox").... Speedup scaling always decreases as core-count decreases due to increased cpu cycles merely swapping data around cores.

[dtaflorida]

6 minutes ago, PocketNerd said:

16c/32t CPU vs 10c/20t CPU...

I WONDER WHICH WOULD WIN?!

They probably are in the same price bracket actually. So as far as I'm concerned it is a valid apples to apples comparison. $1k cpu vs  $1k cpu.