News AMD Comes Roaring Back: Analyzing CPU Sales at Mindfactory Over the Last Five Years

[kinggremlin]

Well, Intel's lack of innovation is finally starting to catch up to em.

Indeed. Hopefully Jim Keller can deliver on his vision.

Intel's next architecture will be significantly bigger.

 

[jimmysmitty]

That's a definite no on the Intel side. Tick-Tock. As Intel defined it, Tock was the architecture change while tick was the die shrink.

Intel hasn't had an architectural change that dropped core speeds since they moved from Netburst to Core, 13 years ago. They've basically been tweaking core every generation since. There was no expected or announced drop in core speed when Intel announced Sunny Cove.

Intel still made tweaks even with die shrinks. Penryn (45nm) also included larger cache sizes, power tweaks and various other tweaks that Conroe did not have.

The only speed drop I could see is from the die shrinks themselves.

Indeed. Hopefully Jim Keller can deliver on his vision.

Intel's next architecture will be significantly bigger.

Considering the resources he will have compared to when he was at AMD I would hope so.

 

[TerryLaze]

Everyone here is talking about the process node transition from 14nm to 10/7nm.
Why are you talking about architecture?

I don't know what you are talking about but sunny cove has about 25% more IPC,from 4 to 5 and from 8 to 10 and it's slaughtering ryzen while using less power.
(The ryzen 7-3700U has a TDP of 35w)

https://www.tomshardware.com/reviews/intel-sunny-cove-gen11-xe-gpu-foveros,5932-4.html

https://hothardware.com/reviews/intel-10nm-ice-lake-performance-and-benchmarks?page=4

 

[joeblowsmynose]

I don't know what you are talking about but sunny cove has about 25% more IPC,from 4 to 5 and from 8 to 10 and it's slaughtering ryzen while using less power.
(The ryzen 7-3700U has a TDP of 35w)

https://www.tomshardware.com/reviews/intel-sunny-cove-gen11-xe-gpu-foveros,5932-4.html

https://hothardware.com/reviews/intel-10nm-ice-lake-performance-and-benchmarks?page=4

To correct you Re: the Hot Hardware "review" ...

Ryzen 3700U is a 12w-35w configurable TDP part (configurable by the vendor). There is no indication anywhere in the article that this one has been configured for 35w TDP, in fact, they didn't even name the laptop they used for the testing, which is a bit odd (what were they hiding it for?) ... at the end of the day - its possible that Ryzen CPU was running at 12w TDP, not 35w as you immediately assumed or where uninformed about, but we don't know, since the reviewer wouldn't let us know that for some reason.

(edit: The author commented in the comments section on that article that both the Whisky Lake part and the Ryzen part were technically 15w parts, so it would seem that "technically" the Ryzen and Whiskey lake part should be being compared to the Ice Lake part at 15w not 25w)

We also have to consider that AMD and Intel calculate TDP quite differently - with Intel never considering any boosting in their TDPs

They also noted that the 3700U system had extremely slow memory speeds ... "We should note that the memory in the Ryzen system was configured in dual-channel mode, but it ran at only DDR4-1866. "

Meanwhile the specs on the Ice Lake system in this test:

​

8GB LPDDR4X-3733 (dual channel)

That is a massive difference in ram speeds and Ryzen is very sensitive to ram speeds when it comes to performance (due to the infinity fabric speed running at half of RAM speed - whatever the ram speed is set to)

I'll note here that Ryzen 3700U is 12nm, not 7nm, and is not Zen2 architecture, but Zen+ architecture -- Zen2 APUs should be coming soon, this will be Ice lakes main competition, not this old architecture part.


So in summary,

1) (EDIT) The Ryzen part appears to be "technically" a 15w part according to the author.

2) The Ryzen part was running RAM at half the speed the Ice Lake part was; any enthusiast who owns a Ryzen processor know performance can go up significantly with simple RAM OCs and tweaks (Steve Burke from Gamer's Nexus, revealed that gaming FPS can increase up to an incredible 20% on Zen2 with just ram tweaks ).

3) This Ryzen APU is not Zen2, but Zen1 and is not a 7nm part, but old process and old architecture.



Add all that up and suddenly maybe Ice lake doesn't look that great. I'd like to see a real proper review / comparison -- this one at the article you linked is just plain poor, and appears to have purposefully hidden info regarding the Ryzen part.

And I'll also note that the old Ryzen APU still games better, despite Intel's advances in their iGPU.

So "slaughtering", considering all that, is certainly not a word I'd use, until I see a proper review with a proper comparison system with all the information included. (in fact, with all that considered, many people could see the same review and see Ice lake (in this article anyway) as "meh" - one commentor on that article was wondering where Intel's claimed "18%" IPC uplift was ... (in reality, lower clocks eat into some that IPC uplift and the 18% was over Sandy Bridge - which explains it)

And where are you getting 25% IPC increase from ... and over what? Intel's marketing claimed only an average 18% increase over Sandy Bridge for Ice lake.

I don't know ... it seems that Intel's 10nm is potentially not much better than 14nm Whiskey lake and may actually be outperformed by Zen2 APUs when we get them. We'll have to wait for better quality info and comparison.

 

[joeblowsmynose]

Answer: Intel has massive amounts of money, in laymans terms or literal terms. They are heavily insulated and quite diverse. They DON'T put all their "Eggs" in one basket. Using this diversity, they can handle setbacks.
------
Question: If Intel is making crap processors, then why are they STILL selling more processors than even AMD?

You've answered your own question ... But you forgot about "marketing" and Intel "persuasion" -- each of which costs lots of money - but it's fine, Intel has lots to go around, because of their diversity, enough to "handle setbacks" like this.

 

[jimmysmitty]

To correct you Re: the Hot Hardware "review" ...

Ryzen 3700U is a 12w-35w configurable TDP part (configurable by the vendor). There is no indication anywhere in the article that this one has been configured for 35w TDP, in fact, they didn't even name the laptop they used for the testing, which is a bit odd (what were they hiding it for?) ... at the end of the day - its possible that Ryzen CPU was running at 12w TDP, not 35w as you immediately assumed or where uninformed about, but we don't know, since the reviewer wouldn't let us know that for some reason.

(edit: The author commented in the comments section on that article that both the Whisky Lake part and the Ryzen part were technically 15w parts, so it would seem that "technically" the Ryzen and Whiskey lake part should be being compared to the Ice Lake part at 15w not 25w)

We also have to consider that AMD and Intel calculate TDP quite differently - with Intel never considering any boosting in their TDPs

They also noted that the 3700U system had extremely slow memory speeds ... "We should note that the memory in the Ryzen system was configured in dual-channel mode, but it ran at only DDR4-1866. "

Meanwhile the specs on the Ice Lake system in this test:

​

8GB LPDDR4X-3733 (dual channel)

That is a massive difference in ram speeds and Ryzen is very sensitive to ram speeds when it comes to performance (due to the infinity fabric speed running at half of RAM speed - whatever the ram speed is set to)

I'll note here that Ryzen 3700U is 12nm+, not 7nm, and is not Zen2 architecture, but Zen+ architecture -- Zen2 APUs should be coming soon, this will be Ice lakes main competition, not this old architecture part.


So in summary,

1) (EDIT) The Ryzen part appears to be "technically" a 15w part according to the author.

2) The Ryzen part was running RAM at half the speed the Ice Lake part was; any enthusiast who owns a Ryzen processor know performance can go up significantly with simple RAM OCs and tweaks (Steve Burke from Gamer's Nexus, revealed that gaming FPS can increase up to an incredible 20% on Zen2 with just ram tweaks ).

3) This Ryzen APU is not Zen2, but Zen1 and is not a 7nm part, but old process and old architecture.



Add all that up and suddenly maybe Ice lake doesn't look that great. I'd like to see a real proper review / comparison -- this one at the article you linked is just plain poor, and appears to have purposefully hidden info regarding the Ryzen part.

And I'll also note that the old Ryzen APU still games better, despite Intel's advances in their iGPU.

So "slaughtering", considering all that, is certainly not a word I'd use, until I see a proper review with a proper comparison system with all the information included. (in fact, with all that considered, many people could see the same review and see Ice lake (in this article anyway) as "meh" - one commentor on that article was wondering where Intel's claimed "18%" IPC uplift was ... (in reality, lower clocks eat into some that IPC uplift and the 18% was over Sandy Bridge - which explains it)

And where are you getting 25% IPC increase from ... and over what? Intel's marketing claimed only an average 18% increase over Sandy Bridge for Ice lake.

I don't know ... it seems that Intel's 10nm is potentially not much better than 14nm Whiskey lake and may actually be outperformed by Zen2 APUs when we get them. We'll have to wait for better quality info and comparison.

If you look at the 1065 at 15W vs the 8565U it shows at least 15% in ST performance gains and he 8565U has a higher base and boost speed (1.8GHz/4.6GHz vs 1.3GHz/3.9GHz). The 8565U has a base TDP of 15W and can be configured for 25W or down to 10W. However from this article we can see its configured with its default 15W TDP:

HP Spectre x360 15 (15-df0000) review – the best 15-inch convertible for the money

Let’s imagine that HP’s wide range of convertible devices is a pyramid. At the bottom, you can find the Pavilions and ProBooks of x360 label. Next some EliteBooks and Envys. But on the …

laptopmedia.com

So it was matching TDP wise and the 10th gen did outperform it even with a clock speed deficiency which is impressive. How much of that is up to the RAM vs uArch changes is hard to say since the 8565U can't support that speed of RAM and I have no idea if the 1065 can go lower. However wouldn't you put your best available in both?

As for the AMD system it was a Lenovo T495. Per the specs it should be running at a 15W TDP. However this article mentions that the CPU will draw 25W for sustained workloads:

https://www.notebookcheck.net/Lenov...kPad-T490-in-initial-benchmarks.424019.0.html

Thats for a 3500U model. I would assume the 3700U version is the same laptop just a different chip so it might also be designed to be able to run at 25W for workloads when allowed.

BTW you are incorrect on the IPC uplift. Its not vs Sandy Bridge. Sandy Bridge was passed by Skylake by quite a bit. Its 18% over Skylake and in some cases can push to 40%:

https://www.techpowerup.com/256596/...-massive-40-uplift-over-skylake-18-on-average

Of course take that with a grain of salt, the rumors of a guy with a 6c/12t Ice Lake sample thats matching much faster CPUs at 1GHz+ advantage.

Intel's 10nm Ice Lake CPUs Achieves 18% IPC Improvement Over Skylake

Intel has dropped a ton of information on its upcoming 10nm Ice Lake processors and it is clear that this is going to be one of the biggest updates to its cadence so far - the first major change since 2015 by its own admission. Ice Lake, or more specifically Sunnycove, is going to be a major...

wccftech.com

 

[joeblowsmynose]

If you look at the 1065 at 15W vs the 8565U it shows at least 15% in ST performance gains and he 8565U has a higher base and boost speed (1.8GHz/4.6GHz vs 1.3GHz/3.9GHz). The 8565U has a base TDP of 15W and can be configured for 25W or down to 10W. However from this article we can see its configured with its default 15W TDP:

HP Spectre x360 15 (15-df0000) review – the best 15-inch convertible for the money

Let’s imagine that HP’s wide range of convertible devices is a pyramid. At the bottom, you can find the Pavilions and ProBooks of x360 label. Next some EliteBooks and Envys. But on the …

laptopmedia.com

So it was matching TDP wise and the 10th gen did outperform it even with a clock speed deficiency which is impressive. How much of that is up to the RAM vs uArch changes is hard to say since the 8565U can't support that speed of RAM and I have no idea if the 1065 can go lower. However wouldn't you put your best available in both?

As for the AMD system it was a Lenovo T495. Per the specs it should be running at a 15W TDP. However this article mentions that the CPU will draw 25W for sustained workloads:

https://www.notebookcheck.net/Lenov...kPad-T490-in-initial-benchmarks.424019.0.html

Thats for a 3500U model. I would assume the 3700U version is the same laptop just a different chip so it might also be designed to be able to run at 25W for workloads when allowed.

BTW you are incorrect on the IPC uplift. Its not vs Sandy Bridge. Sandy Bridge was passed by Skylake by quite a bit. Its 18% over Skylake and in some cases can push to 40%:

https://www.techpowerup.com/256596/...-massive-40-uplift-over-skylake-18-on-average

Of course take that with a grain of salt, the rumors of a guy with a 6c/12t Ice Lake sample thats matching much faster CPUs at 1GHz+ advantage.

Intel's 10nm Ice Lake CPUs Achieves 18% IPC Improvement Over Skylake

Intel has dropped a ton of information on its upcoming 10nm Ice Lake processors and it is clear that this is going to be one of the biggest updates to its cadence so far - the first major change since 2015 by its own admission. Ice Lake, or more specifically Sunnycove, is going to be a major...

wccftech.com

All fair ... and thanks for the correction on Sandy Bridge - I meant to write Sky Lake. The number is still 18% -- not 25%.

My only point wasn't that Ice lake was slower than Whiskey Lake or 12nm Ryzen APU, only that there wasn't any "slaughter" happening vs Whiskey Lake or Ryzen, with all things considered that the "review" showed, and didn't show.

Zen2 APUs probably won't be out till early 2020 so we'll have to wait a bit to compare Intel's 10nm directly with Zen2 7nm -- since it doesn't look like we get 10nm desktop anytime soon for that comparison ...

 

[jimmysmitty]

All fair ... and thanks for the correction on Sandy Bridge - I meant to write Sky Lake. The number is still 18% -- not 25%.

My only point wasn't that Ice lake was slower than Whiskey Lake or 12nm+ Ryzen APU, only that there wasn't any "slaughter" happening vs Whiskey Lake or Ryzen, with all things considered that the "review" showed, and didn't show.

Zen2 APUs probably won't be out till early 2020 so we'll have to wait a bit to compare Intel's 10nm directly with Zen2 7nm -- since it doesn't look like we get 10nm desktop anytime soon for that comparison ...

I am still of the belief that 10nm wont ever make it to desktop. I think they will skip it for 7nm. And they could still give us Sunny Cove on desktop using 14nm but I doubt they will. My guess is Intel will push 7nm with Willow or Golden Cove to the desktop. My guess is the latter since Willow Cove will be just an enhancement while Golden Cove is supposed to be all performance gains.

The biggest issue with any performance gains is real world application. Some may see it, others may not.

I do wonder about that 6c/12t Sunny Cove CPU. It would be rather insane if we could see upwards of 40% in some scenarios but I highly doubt it. Although when I look at what Intel has done to the uArch it is very much capable of quite a bit of performance gains. Time shall tell.

 

[Olle P]

... Then you have ... support etc. Its a lot of work. AMD dropped out of the server market and it will take time and hard work to win back some companies trust. ...

... which is why AMD now began by contracting some really major users. That way their small service organization (compared to what Intel has) will be able to provide a good service while expanding.
AMD is expected to ten-fold their market share (in the server market) in about a year from now. Expect that share to continue growing unless Intel come up with some magic soon.

 

[joeblowsmynose]

I am still of the belief that 10nm wont ever make it to desktop. I think they will skip it for 7nm. And they could still give us Sunny Cove on desktop using 14nm but I doubt they will. My guess is Intel will push 7nm with Willow or Golden Cove to the desktop. My guess is the latter since Willow Cove will be just an enhancement while Golden Cove is supposed to be all performance gains.

The biggest issue with any performance gains is real world application. Some may see it, others may not.

I do wonder about that 6c/12t Sunny Cove CPU. It would be rather insane if we could see upwards of 40% in some scenarios but I highly doubt it. Although when I look at what Intel has done to the uArch it is very much capable of quite a bit of performance gains. Time shall tell.

Fully agreed (I think wev'e discussed this possible 10nm bypass for desktop previously). And I have the same curiosities as you.

 

[Loadedaxe]

They dont get a medal, but they do get intel crushing sales.

And intel didnt improve a lot over 9 years, so they are now seeing the effects of that.

You guys are acting like I am hating on AMD. I am not. I own Ryzen.


But the sky is not falling at Intel.....

 

[TerryLaze]

And where are you getting 25% IPC increase from ... and over what? Intel's marketing claimed only an average 18% increase over Sandy Bridge for Ice lake.

They go from being able to run 4 instructions to 5 instructions,that's 25% ,that's what IPC is instructions per clock.they go from 4 to 5.
Same for execution ports, they go from 8 to 10 which is also 25%.
They are able to process 25% more instructions choosing from a 25% bigger pool.

I am still of the belief that 10nm wont ever make it to desktop. I think they will skip it for 7nm.

Why would they give up so much opportunity for milking?
It's not like the first gen of 7nm is somehow magically going to achieve high clocks which is the only problem right now.

... which is why AMD now began by contracting some really major users. That way their small service organization (compared to what Intel has) will be able to provide a good service while expanding.
AMD is expected to ten-fold their market share (in the server market) in about a year from now. Expect that share to continue growing unless Intel come up with some magic soon.

Intel is selling optane dimms for $8000 a pop and laptop CPUs on nvme cards so that clients can buy dozens of them from intel for the same one system they are selling them to the server market at ridiculous prices,I doubt very much that they are all that concerned about selling low margin CPU cores.
The only server clients buying ryzen are the ones still stuck in the stone ages that only need to do heavy computation.
https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html
https://www.extremetech.com/computing/296990-intel-nervana-nnp-i-nnp-t-a-training-inference

 

[jimmysmitty]

They go from being able to run 4 instructions to 5 instructions,that's 25% ,that's what IPC is instructions per clock.they go from 4 to 5.
Same for execution ports, they go from 8 to 10 which is also 25%.
They are able to process 25% more instructions choosing from a 25% bigger pool.

Why would they give up so much opportunity for milking?
It's not like the first gen of 7nm is somehow magically going to achieve high clocks which is the only problem right now.

Intel is selling optane dimms for $8000 a pop and laptop CPUs on nvme cards so that clients can buy dozens of them from intel for the same one system they are selling them to the server market at ridiculous prices,I doubt very much that they are all that concerned about selling low margin CPU cores.
The only server clients buying ryzen are the ones still stuck in the stone ages that only need to do heavy computation.
https://www.tomshardware.com/news/intel-optane-dimm-pricing-performance,39007.html
https://www.extremetech.com/computing/296990-intel-nervana-nnp-i-nnp-t-a-training-inference

My only reasoning is the fast rollout to 7nm after 10nm will barely be getting established. From a marketing perspective it would make sense. And we have no information on Intels 7nm. I am not assuming higher clocks but there is the possibility.

And I agree. Intel has been focusing on high margin markets which is why I find it funny that people think Intel is going to suffer with lower desktop sales. The same Xeon CPU as a desktop CPU sells for vastly higher margins. Plus Intel has a ton of irons in the fire right now while AMD has CPUs and GPUs for the most part.

Intel will be fine.

 

[NightHawkRMX]

Epyc seems pretty impressive tho

 

[TJ Hooker]

They go from being able to run 4 instructions to 5 instructions,that's 25% ,that's what IPC is instructions per clock.they go from 4 to 5.
[...]
They are able to process 25% more instructions [...]

(Real) IPC is not a nice, fixed number like 4 or 5. It depends on a variety of factors, e.g. what type of instruction being executed. Where are you getting that from?

 

[NightHawkRMX]

(Real) IPC is not a nice, fixed number like 4 or 5. It depends on a variety of factors, e.g. what type of instruction being executed. Where are you getting that from?

I agree that IPC varies highly depending on workload and instructions utilized. Some people have said Zen2 has higher IPC compared to Intel Coffee Lake Refreh, however thats not the full story.

I have seen a test where AMDs IPC in workloads like maxon software is superior to intel. However, intels IPC is above amd for gaming.

The test compared a Ryzen 9 3700x @ 4.0ghz and i9 9900k @ 4ghz. The 9900k had superior gaming performance, however the 3700x performed better in creative workloads ike Cinebench.

I cent find these tests to link.

 

[jimmysmitty]

I agree that IPC varies highly depending on workload and instructions utilized. Some people have said Zen2 has higher IPC compared to Intel Coffee Lake Refreh, however thats not the full story.

I have seen a test where AMDs IPC in workloads like maxon software is superior to intel. However, intels IPC is above amd for gaming.

The test compared a Ryzen 9 3700x @ 4.0ghz and i9 9900k @ 4ghz. The 9900k had superior gaming performance, however the 3700x performed better in creative workloads ike Cinebench.

I cent find these tests to link.

4GHz CPU Battle: Ryzen 3900X vs. 3700X vs. Core i9-9900K

Expanding upon all the testing we performed in our day-one 3rd-gen Ryzen coverage, today we'll be running a clock-for-clock comparison benchmark. IPC can be a good indicator...

www.techspot.com

Something like that? Biggest issue for Ryzen is memory latency and that might be due to memory being on a separate I/O die vs the IMC the 9900K has. Its the biggest reason the 9900K is better in games. It might also be that Intels CPUs performance is not as affected by memory speeds.

Also strangely the 3700x has a bit over half the memory write speed of the 3900x set to 4 cores which is very odd. I would imagine it would be the same no matter the amount of CCXs or cores.

To be fair though Intels 10th gen laptop CPUs also show a good performance vs the 9th gen at lower clock speeds without the extra cache. But they made a lot of changed to the uOP and prefetch which we wont see on desktop for a while.

 

[TJ Hooker]

Also strangely the 3700x has a bit over half the memory write speed of the 3900x set to 4 cores which is very odd. I would imagine it would be the same no matter the amount of CCXs or cores.

This is a 'known' but somewhat poorly reported quirk of Zen 2's design, relating to the links between the compute and I/O die. See the "Chips, chiplets, PCIe 4" section here for more detail: https://hexus.net/tech/reviews/cpu/132374-amd-ryzen-9-3900x-ryzen-7-3700x/

 

[jimmysmitty]

This is a 'known' but somewhat poorly reported quirk of Zen 2's design, relating to the links between the compute and I/O die. See the "Chips, chiplets, PCIe 4" section here for more detail: https://hexus.net/tech/reviews/cpu/132374-amd-ryzen-9-3900x-ryzen-7-3700x/

Interesting read. So basically its by design. And that hurts them in some ways.

The latency is also probably known and by design.

I wonder if Intels EMIB will present similar "quirks" or due to its design if it will be better....

 

[TJ Hooker]

Interesting read. So basically its by design. And that hurts them in some ways.

Doesn't look like it makes more than a couple % difference from what I've seen, so most people would probably never notice the difference.

 

[TerryLaze]

(Real) IPC is not a nice, fixed number like 4 or 5. It depends on a variety of factors, e.g. what type of instruction being executed. Where are you getting that from?

Yes it is a fixed number.
The thing that depends on a variety of factors, e.g. what type of instruction being executed is the performance(per core) .
It's like saying horsepower is the same as mph but ipc/horsepower always stays the same while performance/mph depends on what you do.

4GHz CPU Battle: Ryzen 3900X vs. 3700X vs. Core i9-9900K

Expanding upon all the testing we performed in our day-one 3rd-gen Ryzen coverage, today we'll be running a clock-for-clock comparison benchmark. IPC can be a good indicator...

www.techspot.com

Something like that? Biggest issue for Ryzen is memory latency and that might be due to memory being on a separate I/O die vs the IMC the 9900K has.

Biggest issue for ryzen is that the 9900k has another 25% in clock overhead while the ryzen maybe has 10% at best.
*all core

 

[TJ Hooker]

Yes it is a fixed number.
The thing that depends on a variety of factors, e.g. what type of instruction being executed is the performance(per core) .
It's like saying horsepower is the same as mph but ipc/horsepower always stays the same while performance/mph depends on what you do.

No, it's not. Performance (instructions/operations per second) = IPC x clock speed. Given that performance obviously varies from one load to the next, even with constant frequency, it's obvious that effective IPC must vary as well.

For very old CPUs at least there was a fixed number of cycles required to perform some operations, but it still varied by instruction. https://en.wikipedia.org/wiki/Intel_8086#Performance
For modern CPUs with out of order execution, pipelining, various caches, etc., executing a thread with a variety of instructions, it is much more complicated to come up with a specific number for effective IPC.

And even if it was a fixed number (it's not), I'll again ask for a source for your claim that Intel's IPC went from 4 to 5.

Also, your analogy is a poor one, as the power output (hp) of a motor is not constant (varies with RPM, air-fuel ratio, etc).

 

[NightHawkRMX]

I agree that IPC varies highly depending on workload and instructions utilized. Some people have said Zen2 has higher IPC compared to Intel Coffee Lake Refreh, however thats not the full story.

I have seen a test where AMDs IPC in workloads like maxon software is superior to intel. However, intels IPC is above amd for gaming.

The test compared a Ryzen 9 3700x @ 4.0ghz and i9 9900k @ 4ghz. The 9900k had superior gaming performance, however the 3700x performed better in creative workloads ike Cinebench.

I cent find these tests to link.

I will also add to my previous post, some of the performance differences between the 2 CPUs could have to do with latency, cache size, or other variables excluding IPC.

I would imagine the reason above and how varied IPC is from workload to workload is why IPC is so hard to pinpoint.

 

[jimmysmitty]

I will also add to my previous post, some of the performance differences between the 2 CPUs could have to do with latency, cache size, or other variables excluding IPC.

I would imagine the reason above and how varied IPC is from workload to workload is why IPC is so hard to pinpoint.

Correct. Some applications respond differently to cahce sizes, cache/memory latency or even just clock speed. Its why at the same clock AMD performs better on some while Intel performs better on others. This even happened during the Athlon 64 days. AMD had a better overall CPU but the Pentium 4 did outperform it in certain tasks.

Intels biggest advantage is clock speed. They can push an 8 core to 5GHz on all cores. AMD cannot even if they upped the power limit for the CPU. Its either a limitation of the process, 7nm, or of the uArch design.

One other is going to be extensions supported by each chip.

 

[TJ Hooker]

I will also add to my previous post, some of the performance differences between the 2 CPUs could have to do with latency, cache size, or other variables excluding IPC.

Factors like cache are part and parcel of IPC. Even if you could come up with some theoretical value for IPC for the CPU core in isolation that excludes data access time, I don't see the point.