[Tweaktown] Jim Keller leaving Tesla to join Intel

[PeterScott]

2600X vs. 8700 at fixed 4Ghz

If people are interested in clock for clock IPC, this is the one.

It shows on a Variety of loads, that it is essentially +/-3% on them when running at the same clock speed.

Except for two areas where Intel pulls ahead. Gaming and Handbrake.

The only two loads I run where multi-core matters much are Gaming and Handbrake. So Intel is currently the better architecture for me.

Of course in actual products it gets more complicated, because Intel is running at much higher clock speeds and AMD is typically giving you more cores for the price.

But that still tends to put Intel ahead on Gaming, and about tied for Handbrake, despite the core deficit.

 

[tamz_msc]

If people are interested in clock for clock IPC, this is the one.

It shows on a Variety of loads, that it is essentially +/-3% on them when running at the same clock speed.

Except for two areas where Intel pulls ahead. Gaming and Handbrake.

The only two loads I run where multi-core matters much are Gaming and Handbrake. So Intel is currently the better architecture for me.

Of course in actual products it gets more complicated, because Intel is running at much higher clock speeds and AMD is typically giving you more cores for the price.

But that still tends to put Intel ahead on Gaming, and about tied for Handbrake, despite the core deficit.

Gaming 'IPC' is a misnomer because it introduces a variable which has nothing to do with the CPU.

 

[frozentundra123456]

Gaming 'IPC' is a misnomer because it introduces a variable which has nothing to do with the CPU.

Call it whatever you like, Intel is still faster at the same clockspeed and has about 20% greater clock speed headroom.

 

[TheELF]

If people are interested in clock for clock IPC, this is the one.

It shows on a Variety of loads, that it is essentially +/-3% on them when running at the same clock speed.

If you call 4 different 3d renderers, 2 video encoding and a synthetic excel workload variety...

 

[tamz_msc]

Call it whatever you like, Intel is still faster at the same clockspeed and has about 20% greater clock speed headroom.

So what? It's still a stupid idea to call it an IPC comparison if it's only about games.

 

[TheELF]

Gaming 'IPC' is a misnomer because it introduces a variable which has nothing to do with the CPU.

How do you mean?You don't run games on the CPU?

 

[tamz_msc]

If you call 4 different 3d renderers, 2 video encoding and a synthetic excel workload variety...

Monte-Carlo in Excel is 'synthetic'. Something new I learned today.

 

[tamz_msc]

How do you mean?You don't run games on the CPU?

When you're including games, it’s isn't just the CPU that is being tested.

 

[Gikaseixas]

Guys enough with the Intel vs AMD contest please

What we're witnessing here it's an invigorated Intel that wants to have capable graphics (Raja) and the best CPU in the business (Jim)

Nvidia and AMD better get ready because they woke up the beast (finally)

 

[Kenmitch]

Guys enough with the Intel vs AMD contest please

What we're witnessing here it's an invigorated Intel that wants to have capable graphics (Raja) and the best CPU in the business (Jim)

Nvidia and AMD better get ready because they woke up the beast (finally)

Will be interesting to see what comes of it.

 

[IllogicalGlory]

Gaming 'IPC' is a misnomer because it introduces a variable which has nothing to do with the CPU.

So then Skylake-X is also getting a pass for bad "gaming IPC" as well? It's in your sig.

 

[Zucker2k]

So what? It's still a stupid idea to call it an IPC comparison if it's only about games.

No, it's not only about games. If you need a reminder, the last time we had this argument this is how it went:

https://imgur.com/a/6tmEe
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

 

[IRobot23]

If people are interested in clock for clock IPC, this is the one.

It shows on a Variety of loads, that it is essentially +/-3% on them when running at the same clock speed.

Except for two areas where Intel pulls ahead. Gaming and Handbrake.

The only two loads I run where multi-core matters much are Gaming and Handbrake. So Intel is currently the better architecture for me.

Of course in actual products it gets more complicated, because Intel is running at much higher clock speeds and AMD is typically giving you more cores for the price.

But that still tends to put Intel ahead on Gaming, and about tied for Handbrake, despite the core deficit.

I think you would be surprised to see ZEN+ core gaming "IPC". Most of efficiency on i7 8700K/Intel comes from lower memory latency. If we would want to isolate all other factors CORE should run at 1.6GHz with LLC at 1.6GHz and MEMORY at 3200MHz and then do gaming comparison. I think ZEN+ has great IPC, Ryzen arch is just not best way to efficiently use it.
If we could run Ryzen 2700X at 5GHz with DDR4 at 4500MHz... I think it would be really hard for I7 8700K to beat it.

 

[William Gaatjes]

No, it's not only about games. If you need a reminder, the last time we had this argument this is how it went:

https://imgur.com/a/6tmEe
https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

Those are strange numbers.

Full zen design explanation and ipc numbers:
https://arstechnica.com/gadgets/2017/03/amds-moment-of-zen-finally-an-architecture-that-can-compete/

 

[cytg111]

Presumably he's going to announce today.

The enthusiast in me cannot wait to see what comes out of this, the stoic in me is prepared for the (painful) hype train to come.

Intel formally announces the move.

Haha ... Nice! Lets see what Keller can dream up with Intel IP portfolio and fabbing capabilities... I already dream of faster chips.

 

[Zucker2k]

Those are strange numbers.

Full zen design explanation and ipc numbers:
https://arstechnica.com/gadgets/2017/03/amds-moment-of-zen-finally-an-architecture-that-can-compete/

Sub-heading: Intel's architecture is still better, but AMD has significantly narrowed the gap.

 

[Markfw]

Sub-heading: Intel's architecture is still better, but AMD has significantly narrowed the gap.

Thats a year old, and all of this is OFF TOPIC. Make your own thread to discuss this matter !

 

[esquared]

Next one that continues this AMD vs Intel talk, in this thread, gets a vacation.


esquared
Anandtech Forum Director

 

[frozentundra123456]

I think you would be surprised to see ZEN+ core gaming "IPC". Most of efficiency on i7 8700K/Intel comes from lower memory latency. If we would want to isolate all other factors CORE should run at 1.6GHz with LLC at 1.6GHz and MEMORY at 3200MHz and then do gaming comparison. I think ZEN+ has great IPC, Ryzen arch is just not best way to efficiently use it.
If we could run Ryzen 2700X at 5GHz with DDR4 at 4500MHz... I think it would be really hard for I7 8700K to beat it.

nvm



Couldn't see my mod warning post above?


esquared
Ananantech Forum Director

 

[tamz_msc]

So then Skylake-X is also getting a pass for bad "gaming IPC" as well? It's in your sig.

It's in quotes for a reason.



Another who didn't heed my mod warning.


esquared
Anandtech Forum Director

 

[tamz_msc]

Interesting tweet from Ashraf:

So, Jim Keller at Intel isn't there to work on cores, but on SoC integration. Intel has been trying to develop a good SoC methodology for years but still hasn't quite cracked it. Perhaps Keller will help them get there.

Ties into what BK said during the earnings call:

....we are using some very unique packaging technologies and such that allow us. At 7 nanometers and beyond, we're really moving to a world where you're not going to look at any piece of silicon as being a single node. You're going to use what we're going to call heterogeneous techniques that allow us to use silicon for multiple nodes. So you may use cores from 7 nanometers and IP from 14 nanometers and even as far back as 22 nanometers for the parts that don't need the high performance. And we're able to put those together and make them perform and behave like a single piece of silicon in the package.

So Keller might be working on a new SoC paradigm for integrating a wider range of IP on EMIB, beginning with Intel's 7nm.

 

[Nothingness]

So you changed your mind finally?

 

[moinmoin]

So Keller might be working on a new SoC paradigm for integrating a wider range of IP on EMIB, beginning with Intel's 7nm.

Hm, maybe somebody can refresh my memory, I recall reading that Keller's input with Zen wasn't so much the cores themselves but the on chip validation, i.e. the control fabric and the comparably huge amount of sensors that come along with it to have it all work smoothly without the complexity increasing development time. Having that in mind Intel seeking his expertise for a heterogeneous application of EMIB makes perfect sense.