Intel to Launch 18-Core/36-Thread "Core i9-7980XE" Skylake-X CPU

[PCGuy_5960]

Just now, Valentyn said:

It does? Got a link.

Quote

Show me the money (Cache)

 

Intel have tweaked their cache hieracy again, this means that like Skylake-S (read i7-77xx) the L3 cache is reduced and L2 cache increased. The L3 now acts like a pseudo-victim cache with pre-fetching. This increased lower level cache means that the cache miss rate, which slow down operations are reduced and the cache hit rate is increased.

 

Tl;DR: IPC is increased over Skylake-S

 

Anandtech predicts around 11-18% IPC increase. (Isn't this the same or more than IPC gain we saw Skylake to Kabylake?)

 

[PCGuy_5960]

Just now, i_build_nanosuits said:

of course it doesn't...it's the same architecture, so 0% IPC improvement.

The same architecture with a 4x larger L2 cache

1 minute ago, i_build_nanosuits said:

4.5ghz on a 18 core chip?! ...good luck with that!

Sorry, confused it with the 12 core one.  But I do think that the 18 core should be able to clock to 3.5+GHz

[i_build_nanosuits]

Just now, PCGuy_5960 said:

The same architecture with a 4x larger L2 cache

..which impact IPC by...0%

[Valentyn]

Just now, PCGuy_5960 said:

 

Eh, no where in the Anandtech source is an increase in IPC mentioned.

[done12many2]

1 minute ago, i_build_nanosuits said:

4.5ghz on a 18 core chip?! ...good luck with that!

 

Yeah, that shit probably isn't happening, but Boost 3.0 will at least let you pick you two favorite cores (vice 1) and clock the shit out of them for daily use.  That combined with the driver that tells Windows to keep single-threaded apps on those cores as it did with Broadwell-E should suffice for a nice boost in single-threaded performance while still having an army of cores on standby to do some work!

[Guest Darth Revan]

I told you that Intel already had something better than what AMD had, and I told you they will release it and destroy AMD.

Intel has no reason to lower the price, no matter how much the consumers want that.

They know they have the best of the best. So they're going to price it as much as they want and people will buy it. Because it's the best.

[Valentyn]

Just now, done12many2 said:

 

Yeah, that shit probably isn't happening, but Boost 3.0 will at least let you pick you two favorite cores (vice 1) and clock the shit out of them for daily use.  That combined with the driver that tells Windows to keep single-threaded apps on those cores as it did with Broadwell-E should suffice for a nice boost in single-threaded performance while still having an army of cores on standby to do some work!

 

I was a bit skeptical of those Ghz as well, although the new Boost does look like it would be great.

Doing some AIDA 64 and Prime 95 tests will allow you to pick the two cores that produce the least heat as well, and see if they're also the most stable for that boost figure. 

Could be very good for people that need one hell of a workstation, but also want to use it for gaming, and other more single core workloads.

[leadeater]

Just now, PCGuy_5960 said:

It can clock up to 4.5GHz vs the 2.2GHz on the 24core one?

 

Just now, i_build_nanosuits said:

4.5ghz on a 18 core chip?! ...good luck with that!

There is no way that an 18 core chip will hit 4.5Ghz on all cores. There simply isn't the power envelope to do it. 

[i_build_nanosuits]

1 minute ago, Valentyn said:

Eh, no where in the Anandtech source is an increase in IPC mentioned.

chache increases proportinally with core count, and have no impact on what the CPU can process...doesn't increase IPC...has nothing to do with that.

Cache is a form of storage for the CPU, so it's like your system RAM...your system is not any faster because you have 64GB of RAM instead of 16...same with a CPU, even if it would have 1GB of L3 cache that wouldn't make it any faster.

[PCGuy_5960]

3 minutes ago, Valentyn said:

Eh, no where in the Anandtech source is an increase in IPC mentioned.

3 minutes ago, i_build_nanosuits said:

..which impact IPC by...0%

In this article, it is mentioned:

Quote

The good element of this design is that a larger L2 will increase the hit-rate and decrease the miss-rate. Depending on the level of associativity (which has not been disclosed yet, at least not in the basic slide decks), a general rule I have heard is that a double of cache size decreases the miss rate by the sqrt(2), and is liable for a 3-5% IPC uplift in a regular workflow. Thus here’s a conundrum for you: if the L2 has a factor 2 better hit rate, leading to an 8-13% IPC increase, it’s not the same performance as Skylake-S. It may be the same microarchitecture outside the caches, but we get a situation where performance will differ.

Fundamental Realisation: Skylake-S IPC and Skylake-X IPC will be different.

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/3

[TOMPPIX]

7 minutes ago, PCGuy_5960 said:

 

Quote

The good element of this design is that a larger L2 will increase the hit-rate and decrease the miss-rate. Depending on the level of associativity (which has not been disclosed yet, at least not in the basic slide decks), a general rule I have heard is that a double of cache size decreases the miss rate by the sqrt(2), and is liable for a 3-5% IPC uplift in a regular workflow. Thus here’s a conundrum for you: if the L2 has a factor 2 better hit rate, leading to an 8-13% IPC increase, it’s not the same performance as Skylake-S. It may be the same microarchitecture outside the caches, but we get a situation where performance will differ.

oops, you posted it already 

[i_build_nanosuits]

1 minute ago, PCGuy_5960 said:

In this article, it is mentioned:

http://www.anandtech.com/show/11464/intel-announces-skylakex-bringing-18core-hcc-silicon-to-consumers-for-1999/3

snip!

read my post just above yours.

[done12many2]

7 minutes ago, Valentyn said:

Eh, no where in the Anandtech source is an increase in IPC mentioned.

 

Quote

a general rule I have heard is that a double of cache size decreases the miss rate by the sqrt(2), and is liable for a 3-5% IPC uplift in a regular workflow. Thus here’s a conundrum for you: if the L2 has a factor 2 better hit rate, leading to an 8-13% IPC increase, it’s not the same performance as Skylake-S. It may be the same microarchitecture outside the caches, but we get a situation where performance will differ.

 

There's the quote from the article that @PCGuy_5960 was talking about.  

[i_build_nanosuits]

1 minute ago, done12many2 said:

 

 

There's the quote from the article that @PCGuy_5960 was talking about.  

...and i think thats uber bullshit...8-13% IPC increase...come on...intel would have done that years ago if that was the case...i think he's talking more along the lines of 0.08% to 0.13% ...

[Valentyn]

Just now, done12many2 said:

 

 

There's the quote from the article that @PCGuy_5960 was talking about.  

 

Thanks, ended up at a different anand article it seems.
Hmm, so just a prediction based on the new Cache.

That's a significant increase if true, I'll wait for reviews now; temps and IPC will be interesting to observe.

Even more so during multi-core workloads.

[PCGuy_5960]

1 minute ago, i_build_nanosuits said:

...and i think thats uber bullshit...8-13% IPC increase...come on...intel would have done that years ago if that was the case...i think he's talking more along the lines of 0.08% to 0.13% ...

No they wouldn't, they didn't need to  Now that AMD is competing again, they are forced to improve their CPUs as much as they can

[PCGuy_5960]

2 minutes ago, Valentyn said:

That's a significant increase if true, I'll wait for reviews now; temps and IPC will be interesting to observe.

Let's hope that they don't use too much adhesive this time around

[Valentyn]

Just now, PCGuy_5960 said:

No they wouldn't, they didn't need to  Now that AMD is competing again, they are forced to improve their CPUs as much as they can

 

I have no doubt that intel would have been content releasing a 12 core Skylake X as their next flagship if not for AMD and Ryzen/Threadripper appearing.

While the 8 core had its direct PCIe lanes reduced, it also got a big price reduction as a result as well. 

 

Zen + should hopefully push Intel to do better as well; if their easy gaines in IPC are as easy as AMD claims.

[Valentyn]

Just now, PCGuy_5960 said:

Let's hope that they don't use too much adhesive this time around

 

Have you seen the Der8bauer video? It didn't look too bad, he got it off rather easily for once.

 

 

[TOMPPIX]

3 minutes ago, PCGuy_5960 said:

No they wouldn't, they didn't need to  Now that AMD is competing again, they are forced to improve their CPUs as much as they can

and this is why i don't like intel

[done12many2]

Just now, Valentyn said:

 

Thanks, ended up at a different anand article it seems.
Hmm, so just a prediction based on the new Cache.

That's a significant increase if true, I'll wait for reviews now; temps and IPC will be interesting to observe.

Even more so during multi-core workloads.

 

Agreed.  As with anything, I think we'll observe an overall increase on the lower side of that estimate, but peaks in certain areas.

 

Either way, when you actually step back and look at what Intel is doing with the x299 release, they are giving us a hell of a lot more options, better pricing and better performance.

 

Even with the 8c/16t part.  Despite the reduction in CPU lanes when compared to the current x99 8c/16t parts, when you compare it to the AMD equivalent part, you end up with more CPU PCIe lanes, more chipset lanes, more IPC, more overclocking headroom and so on.  Grant it, it is more expensive, but I think that enthusiasts will be willing to pay the extra $100. 

[leadeater]

Well in a situation where the CPU does need more cache it will make a huge difference, if it doesn't need it it wont do anything. We saw this back in the origins of dual core when AMD had the superior architecture where Intel went in with a very large unified L2 cache and eventually ended up winning out due to having good across the board performance.

 

Sometimes more cache is the answer and sometimes it's not.

[done12many2]

1 minute ago, TOMPPIX said:

and this is why i don't like intel

 

It must pain you to turn your PC on everyday. 

[PCGuy_5960]

3 minutes ago, TOMPPIX said:

and this is why i don't like intel

AMD would have done the same thing, believe me

[TOMPPIX]

4 minutes ago, done12many2 said:

 

It must pain you to turn your PC on everyday. 

i never said i hated intel.