I9 7960X benchmarks released on LN2

[Lays]

http://hwbot.org/hardware/processor/core_i9_7960x/

 

The embargo on manufacturers benchmarks lifted this morning, we should see more coming soon. It beats the current fastest cinebench r15 Threadripper Ln2 score by roughly 400 points. Which looks to be about 9%.

 

http://hwbot.org/submission/3622583_sofos1990_cinebench___r15_core_i9_7960x_4931_cb/

 

 

This is a good sign for everyone that said it would be slower than TR, and an even better sign for how insane the 18 core will be.  Obviously the 16 is more expensive, but you get what ya pay for. The fact that it does 5.6 on Ln2 is a good sign as well, if it behaves like the 10 core does we should see similar clock speeds on watercooling since the 16c most likely has stronger binning behind it. 

[tom_w141]

Quote

I7 7960X benchmarks released...

 *clicks*

 

Quote

...on LN2

 *closes*

[Lays]

1 minute ago, tom_w141 said:

 *clicks*

 

 *closes*

Read the rest of the post. 

[The Sloth]

4 minutes ago, Lays said:

you get what ya pay for.

yea, lets pay magnitude more more for a small performance gain. 

[tom_w141]

Just now, Lays said:

Read the rest of the post. 

TLDR rest of post says:

 

i9 (stupid cooling method used by no one day to day) beats TR (stupid cooling method used by no one day to day). LN2 threads are totally irrelevant and show nothing meaningful.

[PCGuy_5960]

Impressive stuff  It will be interesting to see how it OCs on water.

[The Benjamins]

Just now, Lays said:

Read the rest of the post. 

ya that's neat and all but it really doesn't matter.

 

the 7960x still cost $700 more, still does not have top tier motherboards out.

 

I want to know what it can score at stock and custom loop OC.

[leadeater]

I think one of the Threadripper reviews mentioned that Intel released cinebench scores for their 16 core Skylake-X CPU, supposed to be about 3200 points so higher than TR. The 18 core should be 3500 ish.

 

https://www.reddit.com/r/intel/comments/6s7g8z/sneak_peek_at_intel_core_i97960x_performance/

[tom_w141]

Just now, PCGuy_5960 said:

It will be interesting to see how it OCs on water.

My thoughts exactly

[PCGuy_5960]

1 minute ago, leadeater said:

I think one of the Threadripper reviews mentioned that Intel released cinebench scores for their 16 core Skylake-X CPU, supposed to be about 3200 points so higher than TR. The 18 core should be 3500 ish.

 

https://www.reddit.com/r/intel/comments/6s7g8z/sneak_peek_at_intel_core_i97960x_performance/

Clock speeds? If it gets 3200 at stock, I'm impressed

[leadeater]

2 minutes ago, PCGuy_5960 said:

Clock speeds? If it gets 3200 at stock, I'm impressed

Yea stock, I don't think Intel would be giving out anything other than stock figures.

[The Benjamins]

4 minutes ago, PCGuy_5960 said:

Clock speeds? If it gets 3200 at stock, I'm impressed

it sounds like to me that would be the 3.6 all core boost. that would make it 6.6% faster then 1950x at its 3.7 all core boost which makes sense to me. that's like 7.5% faster IPC. (didn't math, I lazy)

[PCGuy_5960]

1 minute ago, leadeater said:

Yea stock, I don't think Intel would be giving out anything other than stock figures.

Quite impressive. Let's hope that they are soldered.

1 minute ago, The Benjamins said:

it sounds like to me that would be the 3.6 all core boost. that would make it 6.6% faster then 1950x at its 3.7 all core boost which makes sense to me. that's like 7.5% faster IPC. (didn't math, I lazy)

Yeah, if it overclocks like the 7900X, it should be quite a bit faster than TR.

[Lays]

5 minutes ago, tom_w141 said:

TLDR rest of post says:

 

i9 (stupid cooling method used by no one day to day) beats TR (stupid cooling method used by no one day to day). LN2 threads are totally irrelevant and show nothing meaningful.

It still tells us it should clock well based on how it does on Ln2.  It is a few hundred mhz behind the best 7900x's on Ln2 so if expect similar on water. Amd actually has a huge advantage on LN2 because it can run at - 196c, and x299 cold bugs at - 100ish. 

Just now, The Benjamins said:

it sounds like to me that would be the 3.6 all core boost. that would make it 6.6% faster then 1950x at its 3.7 all core boost which makes sense to me. that's like 7.5% faster IPC.

Based on what we see In the 10 it will oc well with proper cooling I reckon. 

[The Benjamins]

Just now, PCGuy_5960 said:

Quite impressive. Let's hope that they are soldered.

Yeah, if it overclocks like the 7900X, it should be quite a bit faster than TR.

I would hope so, If you haven't seen the 1950x gets 3350-3450 when OCed to 4Ghz

[Princess Luna]

Impressive? no not at all, when your product costs significantly more it has to be superior one way or another, Intel still has too bad value for money to actually be any attractive even if it is 7~8% faster

[PCGuy_5960]

Just now, The Benjamins said:

I would hope so, If you haven't seen the 1950x gets 3350-3450 when OCed to 4Ghz

I know, if the 7960X gets 3200cb at 3.6GHz, at 4.5GHz it should be able to hit 4000cb. 

 

Although, if it's not soldered I doubt that it will be able to hit 4.5GHz without a delid.

[tom_w141]

4 minutes ago, PCGuy_5960 said:

Quite impressive. Let's hope that they are soldered.

I admire your hope  

[DXMember]

HOLY SHIT....

I thought you were dead...

 

On the other hand, to spur up some discussion

http://hwbot.org/submission/3622692_theoverclocker_cinebench___r15_ryzen_threadripper_1950x_4514_cb/

 

if you adjust for the clock speed, it comes out to less than 5% in favor of Intel

[Lays]

Just now, Princess Cadence said:

Impressive? no not at all, when your product costs significantly more it has to be superior one way or another, Intel still has too bad value for money to actually be any attractive even if it is 7~9% faster

There's more to that price tag than just performance, but that's your opinion.  Some people want the best and don't care how much more it is.  Also remember that since TR gets a bigger boost from Ln2 than intel does due to cold bug, you'll probably see an even bigger performance gap on watercooling.

[Princess Luna]

1 minute ago, Lays said:

.

I also want the best....... best value, which more people should start caring for too, would turn the market better for sure.... when you're dealing with 100% price increase like the 1950x to 7980xe for 15~20% increase in performance I really find ludicrous there's people so stubborn to go like that... "BEST OR NUTHING" usually people like that are the ones that don't even use the PC to its full potential also

 

 

[tom_w141]

5 minutes ago, Lays said:

you'll probably see an even bigger performance gap on watercooling.

I'd agree if the chip was soldered. Doesn't matter how sexy your loop is when diminishing returns kick in real fast because the thermal bottleneck is between the die and IHS

[MageTank]

11 minutes ago, Lays said:

http://hwbot.org/hardware/processor/core_i9_7960x/

 

The embargo on manufacturers benchmarks lifted this morning, we should see more coming soon. It beats the current fastest cinebench r15 Threadripper Ln2 score by roughly 400 points. Which looks to be about 9%.

 

http://hwbot.org/submission/3622583_sofos1990_cinebench___r15_core_i9_7960x_4931_cb/

 

 

This is a good sign for everyone that said it would be slower than TR, and an even better sign for how insane the 18 core will be.  Obviously the 16 is more expensive, but you get what ya pay for. The fact that it does 5.6 on Ln2 is a good sign as well, if it behaves like the 10 core does we should see similar clock speeds on watercooling since the 16c most likely has stronger binning behind it. 

Where is the one person that said this? Also, did anyone even believe them? Why on earth would Intel's superior IPC and clock speeds, be slower than a CPU whose MCM design is based on two slower dies? Come on @Lays,  you are making this part up, lol.

 

Also, TR was clocked at 5370mhz when it scored 4514. This 7900X is clocked at 5600mhz scoring 4931. That is a 4.3% difference in clock speeds, but a 9.3% difference in scores. If we ignore memory impact on this benchmark, that's still a 5% difference clock for clock (again, this is ignoring ram, which is likely also having an impact here). 5% is about where we all previously mathed out the IPC difference between Skylake and Ryzen, so it adds up in my books.

 

The REAL difference between real world TR, and real world X299, is cooling requirements. I bet you any amount of money, that Intel result was delidded. You know it, I know it, let's get that out of the way. The average enthusiast will still be too afraid to delid their X299 chips. Ryzen consumers won't need to delid at all. Thermals will certainly be in Ryzen's favor (except for an apparent issue with socket thermals, i'll study that in more detail when I get the free time), but it's going to be difficult to hit 4.5ghz on X299 on these 16 core chips without a custom loop. AIO's will certainly be pushing it, unless they go for an AVX offset and have decent ambients. 

 

TL:DR? Nobody thought TR was going to be faster core for core. If they did, they must have clearly ignored Ryzen entirely, as an MCM design made it extremely easy to extrapolate TR's performance. Aside from the extremely shoddy memory implementation, we all knew roughly how strong the multi-core performance was going to be. None of us thought for a second that TR was going to beat the 7960X or 7980XE, but being priced to compete against the 7900X, it can certainly deliver under very niche conditions. 

[PCGuy_5960]

2 minutes ago, Princess Cadence said:

best value, which more people should start caring for too, would turn the market better for sure....

Yeah, but some people prefer maximum performance. Why did you choose Titan X's instead of 980 Tis, you paid more for a 5% max performance increase?

[dexT]

When are these coming out?

 

AMD has the most powerful CPU on the market.. whoa.. it feels weird saying that.