a look at (ry)Zen's IPC - from Excavator to KabyLake

[Imglidinhere]

Just now, zMeul said:

really?! I wonder what you're reading

 

explain to me how it beats Haswell in just about every test when the absolute IPC compared to Excavator sums up to:

Explain to me what ABSOLUTE IPC means? Because obviously you don't get it either. Kaby Lake does not have a clock for clock advantage over Haswell of THIRTY PERCENT. Those are advanced instruction sets optimizing the architecture, NOT IPC IMPROVEMENTS.

[zMeul]

4 minutes ago, Imglidinhere said:

Explain to me what ABSOLUTE IPC means? Because obviously you don't get it either. Kaby Lake does not have a clock for clock advantage over Haswell of THIRTY PERCENT. Those are advanced instruction sets optimizing the architecture, NOT IPC IMPROVEMENTS.

because you are reading it wrong, that's not what the data shows

@Glenwing explained it perfectly here:

the comparison basis is Excavator IPC, not Zen, not Haswell, not KabyLake

even the chart's sub-title sais Relative to Excavator

[79wjd]

4 minutes ago, zMeul said:

because you are reading it wrong, that's not what the data shows

@Glenwing explained it perfectly here:

the comparison basis is Excavator IPC, not Zen, not Haswell, not KabyLake

even the chart's sub-title sais Relative to Excavator

Kaby Lake doesn't have a 17% IPC advantage over Haswell either. 

[zMeul]

Just now, djdwosk97 said:

Kaby Lake doesn't have a 17% advantage over Haswell either. 

that's not what I said, I never said that

[Imglidinhere]

1 minute ago, zMeul said:

because you are reading it wrong, that's not what the data shows

@Glenwing explained it perfectly here:

the comparison basis is Excavator IPC, not Zen, not Haswell, not KabyLake

even the chart's title sais Relative to Excavator

So it's a ~15% improvement from Haswell to Kabylake? Even that's a bit much.

 

 

Because this shows a different story.

 

I'll admit I misread it, apologies for the explosive reaction, but it was not clarified and you were openly calling me stupid for not getting it and still not explaining it, adding fuel to the fire.

[Imglidinhere]

2 minutes ago, zMeul said:

that's not what I said, I never said that

Actually... you posted something of someone else stating that KL has a 17.3% lead over Haswell in these graphs... and even THAT is wrong. Again... it's the instruction sets and optimizations. Plus there's advanced encoding that happens between the CPU and the iGPU to enhance performance in certain applications. That's the BIG THING with KL.

[zMeul]

1 minute ago, Imglidinhere said:

So it's a ~15% improvement from Haswell to Kabylake? Even that's a bit much.

 

I'll admit I misread it, apologies for the explosive reaction, but it was not clarified and you were openly calling me stupid for not getting it and still not explaining it, adding fuel to the fire.

the IPC should be calculated over a very wide sets of benches that attack the CPU in all areas, not just in one test (CineBench)

 

also, I didn't called you stupid at all

when you don't understand something, ask - because I can't telepathically understand what are you asking

[zMeul]

1 minute ago, Imglidinhere said:

Actually... you posted something of someone else stating that KL has a 17.3% lead over Haswell in these graphs

I sincerely don't remember that

[Imglidinhere]

14 minutes ago, zMeul said:

because you are reading it wrong, that's not what the data shows

@Glenwing explained it perfectly here:

the comparison basis is Excavator IPC, not Zen, not Haswell, not KabyLake

even the chart's sub-title sais Relative to Excavator

Remember this post from like... +/-14 minutes ago? It even says right there, scores 17% higher than Haswell, which, on average is being calculated as an ABSOLUTE IPC improvement and it's got literally nothing to do with IPC at all. It's showing the efficiency of the architecture in those applications. The only way to judge IPC is with something that's going to draw out the raw performance of the processor, not run some number crunching algorithm.  

[79wjd]

2 minutes ago, zMeul said:

I sincerely don't remember that

 

16 hours ago, zMeul said:

source: https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/

 

"The Slit", established member over Anand's forums has done some testing that feature more technical benchmarks that you don't usually see

go check the source for all the benches and results plus in depth technical details on the CPU and chipset (Zeppelin)

 

the tests were conducted at 3.5Ghz

mobo used ASUS CROSSHAIR VI HERO

results with ER mean extremes removed (highest and lowest)

 

let's talk IPC and how it compares to CPUs ranging from Excavator era to Intel's latest KabyLake:

[Bouzoo]

6 hours ago, Glenwing said:

No, the scores here show Kaby Lake to be about 17.3% higher than Haswell. It's just because they're percentages, not absolute numbers, and percentages depend on reference point.

That is the proper math yes, but it seems far to huge from everything we knew until now. And correct me if I'm wrong, but even the The Stilt removed absolute IPC graph from Anandtech? I can't seem to find it. And it's against the tests from Anandtech themselves.

[zMeul]

Just now, Imglidinhere said:

The only way to judge IPC is with something that's going to draw out the raw performance of the processor, not run some number crunching algorithm.  

that's exactly what all those individual benches do, express raw performance from each section of the CPU 

[79wjd]

1 minute ago, Bouzoo said:

That is the proper math yes, but it seems far to huge from everything we knew until now. And correct me if I'm wrong, but even the The Stilt removed absolute IPC graph from Anandtech? I can't seem to find it. 

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/#post-38770111

[zMeul]

1 minute ago, djdwosk97 said:

 

mate?! what are you quoting my OP for?

where do I say what?

[Bouzoo]

Just now, djdwosk97 said:

https://forums.anandtech.com/threads/ryzen-strictly-technical.2500572/#post-38770111

Yes, but where is the first graph from zMeuls post?

[79wjd]

1 minute ago, zMeul said:

mate?! what are you quoting my OP for?

where do I say what?

Your main point was the Absolute IPC graph, which shows Kaby Lake IPC being 17% above Haswell. Therefore, by posting it (and not disagreeing with it in your post), you're clearly agreeing with it. 

1 minute ago, Bouzoo said:

Yes, but where is the first graph from zMeuls post?

Last graph in the first spoiler (from post #3).

[zMeul]

2 minutes ago, Bouzoo said:

That is the proper math yes, but it seems far to huge from everything we knew until now. And correct me if I'm wrong, but even the The Stilt removed absolute IPC graph from Anandtech? I can't seem to find it. 

yes, from the absolute IPC graph he removed extremes, I even said it in the OP

[Bouzoo]

Just now, zMeul said:

yes, from the absolute IPC graph he removed extremes, I even said it in the OP

I haven't read your OP again, that's why.  It seems to be wrong (kind of).

[zMeul]

Just now, djdwosk97 said:

You main point was the Absolute IPC graph, which shows Kaby Lake IPC being 17% above Haswell. Therefore, by posting it (and not disagreeing with it in your post), you're clearly agreeing with it. 

that's not what that graph shows .. wtf mate!

even @Imglidinhere understood in the end

[Imglidinhere]

1 minute ago, zMeul said:

that's exactly what all those individual benches do, express raw performance from each section of the CPU 

So it's as unnecessary of a graph as stating that the 1800X will be a better buy than the 6900K. Gooootcha.

 

I wanted to see how well the CPU fared in an apples to apples comparison with games, not these dumbass number crunching programs no one uses.  It's literally just showcasing the architectural improvements Intel has actually made, which is nice to see that they've made SOME improvements worth noting (not discounting them here btw), but what does gaming look like at the same clock speeds?

[Bouzoo]

2 minutes ago, djdwosk97 said:

Last graph in the first spoiler (from post #3).

Oh I missed so many graphs. Tnx!

[zMeul]

4 minutes ago, Imglidinhere said:

I wanted to see how well the CPU fared in an apples to apples comparison with games, not these dumbass number crunching programs no one uses.  It's literally just showcasing the architectural improvements Intel has actually made, which is nice to see that they've made SOME improvements worth noting (not discounting them here btw), but what does gaming look like at the same clock speeds?

mate, the title is IPC and you expect gaming benchmarks?! 

 

Quote

So it's as unnecessary of a graph

no it's not!

when you compare CPUs you don't benchmark games 

[Bouzoo]

12 minutes ago, zMeul said:

the IPC should be calculated over a very wide sets of benches that attack the CPU in all areas, not just in one test (CineBench)

Even according to Anandtech it is still way off:

 

Quote

Sandy Bridge to Ivy Bridge: Average ~5.8% Up
Ivy Bridge to Haswell: Average ~11.2% Up
Haswell to Broadwell: Average ~3.3% Up
Broadwell to Skylake (DDR3): Average ~2.4% Up
Broadwell to Skylake (DDR4): Average ~2.7% Up

Haswell to Skylake (DDR3): Average ~5.7% Up.

 

[79wjd]

5 minutes ago, zMeul said:

that's not what that graph shows .. wtf mate!

even @Imglidinhere understood in the end

That's exactly what that graph implies and, I'm sure why you chose to use it -- because it exacerbates the lead of Haswell/Kaby Lake over Zen. 

 

That graph implies that Kaby Lake's IPC is 17% faster than Haswell. That's not exactly the case. In terms of raw IPC it's nowhere near that much of an improvement. 

[zMeul]

1 minute ago, Bouzoo said:

Even according to Anandtech it is still way off:

mate, what are we discussing?! what is way off?!