Broadwell-E 6950x benchmarks leaked

[NumLock21]

OC cpu vendor silicon lottery who sells oc chips, got a 6950x es, and ran it against the current extreme edition 5960x. The 6950x is the first deca-core for the desktop with 20mb of L3 cache. The same amont on the 5960x. Both have a stock clock of 3ghz. They're both oc to 4ghz during the bench. And the 6950x was further oc to 4.5ghz.

 

Cb15

5960x @4ghz 1592

6950x @4ghz 1904

6950x @4.5ghz 2327

 

 

http://www.overclock.net/t/1599263/6950x-benchmarked

http://wccftech.com/intel-core-i7-6950x-vs-core-i7-5960x-benchmark-results/

[Senzelian]

It has cores.
It's fast.

WHO WOULD HAVE THOUGHT?! O_O

[Sakkura]

6950X running same clocks as 5960X but getting less single-core performance. Presumably that means performance will improve another 10-15% by launch, with release-quality microcode. So the multithreaded performance will be maybe 30-35% higher than the 5960X.

[NumLock21]

We also would like you to know that motherboard makers are already prepping up new X99 products for the Broadwell-E launch. We have seen

three brand new motherboards from Gigabyte, EVGA has also teased their X99 FTW K motherboard but now we know that ASUS and MSI are also in the bandwagon. MSI is ready with a all new X99 Gaming Pro motherboard that comes with a revised design while ASUS will be deploying the latest X99 A II, X99 Deluxe II and X99 STRIX Gaming series.

 

Hmm... Strix motherboards from Asus. I might have a guess how the X99 A and Deluxe version 2 might look like. It'll look like their Z170s and have USB type C along with TB3 and onboard U.2

[PCgamer324]

hah

 

 

 

still can't beat my dual X5675 OC setup from 2010

[Doobeedoo]

Odd single-core performance being lower on same clocks though. Probably will differ on launch.
Also shame always one gen behind with extreme editions, not to mention they come after =/

[Jaggsta]

Guess release is next month review chips sent to people have been sold on ebay already.

 

• Intel Core i7-6950X – $1950.00 US
• Intel Core i7-6900K – $875.00 US
• Intel Core i7-6850K – $575.00 US

[ClownFace1511]

I'm happy but unhappy with my 4.7 5960x, not gonna upgrade to the new one but sad that I don't have the best one

[Energycore]

The funny thing is, given the same clockspeed, it has 19% better performance while having 25% more cores (10 rather than 8).

Does that mean worse IPC? I'd like to see single threaded cinebench from the 6950X (With that name, you might think it's an old Radeon :P)

[patrickjp93]

17 minutes ago, Energycore said:

The funny thing is, given the same clockspeed, it has 19% better performance while having 25% more cores (10 rather than 8).

Does that mean worse IPC? I'd like to see single threaded cinebench from the 6950X (With that name, you might think it's an old Radeon :P)

Ever heard of Amdahl's Law? Programs which are not 100% parallelizable have diminishing returns when adding more cores. This is why improving single-core performance regardless of core count is paramount from now into the future.

[Energycore]

1 minute ago, patrickjp93 said:

Ever heard of Amdahl's Law? Programs which are not 100% parallelizable have diminishing returns when adding more cores. This is why improving single-core performance regardless of core count is paramount from now into the future.

I haven't, but it sounds like I'll read up on that. So is the GPU workload closer to 100% parallelizable compared to say, physics and AI calculations in games?

[patrickjp93]

2 minutes ago, Energycore said:

I haven't, but it sounds like I'll read up on that. So is the GPU workload closer to 100% parallelizable compared to say, physics and AI calculations in games?

Yes, most of graphics rendering (not lighting) is an embarrassingly parallel problem (100% parallel). Certain implementations of neural network AIs are embarrassingly parallel, but they're not particularly useful. Physics is embarrassingly parallel outside of interactions (collisions, multi-body gravitation, reflections, etc..)

[FPSwithaWacomTablet]

21 hours ago, patrickjp93 said:

Yes, most of graphics rendering (not lighting) is an embarrassingly parallel problem (100% parallel). Certain implementations of neural network AIs are embarrassingly parallel, but they're not particularly useful. Physics is embarrassingly parallel outside of interactions (collisions, multi-body gravitation, reflections, etc..)

Would keyboards count as parallel as well?

If so, why hasn't gaming keyboards been using GPU's for processing keyboard inputs?

 

/s

 

As a side project in the future however, I'd like to use an FPGA to emulate a seperate microcontroller for 100+ keys and a serial connection to a PC (not USB, but RS232). I wonder how parallel that can get...

 

Not that I know what I'm getting myself into though. I only know C after finishing a course on one (also my first programming language)

[Master Disaster]

On 12/05/2016 at 1:19 AM, patrickjp93 said:

Ever heard of Amdahl's Law? Programs which are not 100% parallelizable have diminishing returns when adding more cores. This is why improving single-core performance regardless of core count is paramount from now into the future.

Not to mention its a pre release CPU running on pre release microcode. I'm sure Intel will have some improvements when they drop the finalised microcode closer to release.

[Gdourado]

Any news on release date? 

[Sakkura]

4 hours ago, Gdourado said:

Any news on release date? 

 

Officially launching in "however long, I actually have no idea."