Have the differences in IPC and single core perf. been overstated in games, between Zen 3 and Golden Cove?

g67575 · 29 Jul 2022 at 01:03

Looking back at an old review on Techspot, the minimum framerates at 1080p appear quite similar between Golden Cove CPUs like the 12700K and Zen 3 CPUs like the 5800X:

From this 12600K review:

Intel Core i5-12600K Review: 5600X Defeated

The Core i5-12600K is Intel's latest mainstream CPU and a direct competitor to AMD's popular Ryzen 5 5600X. The Alder Lake chip packs 6 P-cores and 4...

www.techspot.com

My dad is curious about the likely differences in 'real world' performance between Zen 4 and Intel's 12th/13th generations, because he wants to build a new PC (upgrading from 1st gen i7!).

Looking at the above chart, my impression is that since Zen 4 has an 8% increase in IPC, it could end up about equal, or being slighty ahead of the Golden Cove architecture in games (min. framerate). If the IPC translates directly into increased min FPS, that would be a min. FPS of around 158 (10 game average, based on 5800X result).

Vish Petrol · 29 Jul 2022 at 01:32

What is your father going to use it for?

g67575 · 29 Jul 2022 at 01:38

Bit of everything really, I think he will be happy as long as the performance seems steady and consistent in games. He likes historical strategy games like some of the Total War series (he's not a fan of the fantasy games like I am, like WH3). I sold him a RTX 3070 GPU recently, so he should be all set, he's not bothered about a larger/higher resolution display though.

We've been noticing slowdowns when Windows 10 is booting and updating itself, so I think it will be easy to see an improvement. He transfers / moves around files quite a bit, so I think USB4 will help there, particularly with external backups.

I think he's been waiting for the clock speeds to hit 5ghz (all core) at the mainstream, it's taken a long time! I've told him if he goes with AM5, he will probably be able to upgrade the CPU for at least a few more years.

Emlyn_Dewar · 29 Jul 2022 at 05:32

Upgrading from first gen i7, is the prospect of numerous CPU upgrades a big deal? Got a lot of life out of that old one.
It’s always good to wait but even buying into any of the current highish end systems would have a lot to offer. 5800X3D does very well in Total War looking at reviews, but 12th gen seems pretty gold too.

The moving files part, I can’t imagine the USB3 on the current board is too great? Decent M2 drive, and a good motherboard will offer USB3.whateveritson…

g67575 · 29 Jul 2022 at 06:26

True, but there will be no new CPU releases for those platforms (after Intel's 13th gen). He wants DDR5 RAM too, partly because it's the latest thing. I don't think he minds waiting until the end of the year, if needed.

chroniclard · 29 Jul 2022 at 06:36

Will be literally sod all noticeable difference in the real world. All modern CPUs are great. Outside of benchmark arguing nobody cares.

Bencher · 29 Jul 2022 at 07:59

g67575 said:
Looking back at an old review on Techspot, the minimum framerates at 1080p appear quite similar between Golden Cove CPUs like the 12700K and Zen 3 CPUs like the 5800X:

From this 12600K review:

Intel Core i5-12600K Review: 5600X Defeated

The Core i5-12600K is Intel's latest mainstream CPU and a direct competitor to AMD's popular Ryzen 5 5600X. The Alder Lake chip packs 6 P-cores and 4...

www.techspot.com

My dad is curious about the likely differences in 'real world' performance between Zen 4 and Intel's 12th/13th generations, because he wants to build a new PC (upgrading from 1st gen i7!).

Looking at the above chart, my impression is that since Zen 4 has an 8% increase in IPC, it could end up about equal, or being slighty ahead of the Golden Cove architecture in games (min. framerate). If the IPC translates directly into increased min FPS, that would be a min. FPS of around 158 (10 game average, based on 5800X result).

The review hwunboxed runs is not supposed to show CPU differences. It's essentially a GPU review.

alec · 29 Jul 2022 at 08:34

Bencher said:
The review hwunboxed runs is not supposed to show CPU differences. It's essentially a GPU review.

have you looked at the chart? slow cpus on the bottom, fast cpus on the top?
what else do you want from cpu review?

g67575 said:
'real world' performance between Zen 4 and Intel's 12th/13th

Zen 3 and 12th gen are really close on IPC. Intel is currently on top thanks to higher boost clocks (with higher power use) and DDR5.
Next gen is hard to guess yet. We know AMD has caught up in CPU clocks. But Zen 4 could get a smaller IPC improvement than 13th gen.

Honestly, both next gen cpus will be a good choice

Bencher · 29 Jul 2022 at 08:44

alec said:
have you looked at the chart? slow cpus on the bottom, fast cpus on the top?
what else do you want from cpu review?

I want a cpu review to be fully cpu bound, else it's not a cpu review. If you check the individual games there are some that there is no difference between the top and the bottom. Even with the ones that there is a difference, its not as big as it should. Take cyberpunk for example. The difference between the 3700x and a 12700 is nowhere near close to 25%. When you activate RT, the cpu load gets huge and the 3700x struggles to maintain 60. Actually I've seen drops to the 40s in some areas. On the other hand my 12900k recorded a minimum of 115 at the heaviest spot of the game. Thats more like a 300% difference rather than a 30%. The 10600k and 11600k barely keep a minimum of 60 at those intensive areas as well.

Im talking about these cpus cause I actually have or had them.

uscool · 29 Jul 2022 at 10:14

AM5 you can be pretty confident will get 2+ CPU updates , 13th gen that will be it

Emlyn_Dewar · 29 Jul 2022 at 10:20

g67575 said:
True, but there will be no new CPU releases for those platforms (after Intel's 13th gen). He wants DDR5 RAM too, partly because it's the latest thing. I don't think he minds waiting until the end of the year, if needed.

My meaning was more a "why does this matter?" type one. Given that he’s lasted so long on a first gen i7, then he could probably drag a current gen i7 out for a similar amount of time, by then you’d likely want a platform upgrade again anyway.

uscool · 29 Jul 2022 at 10:25

both scheduled Fall 2022 which I dont think is that long

g67575 · 29 Jul 2022 at 10:42

alec said:
But Zen 4 could get a smaller IPC improvement than 13th gen.

13th gen is the same architecture (Golden Cove) as the 12th gen. It's an optimization of the 12th generation, with the same IPC, more cache and approx. 5-6% higher all core clocks for the 13900K compared to the 12900KS. Zen 4 on desktop has an 8% improvement in IPC according to AMD.

They can get 5.5ghz all core out of a recent engineering sample:

Early Core i9-13900K sample shows massive improvement in minimum framerates, boosts to 5.8 GHz

Hardware reviewer ExtremePlayer has posted some benchmarks of an Core i9-13900K qualification sample they managed to obtain. This CPU should offer near-identical performance to shipping products as...

www.techspot.com

It's interesting to see Intel trying to get more performance out of the E-cores, by clocking them upto 400mhz higher.

The new 700 series motherboards for the 13th gen, might ultimately be more interesting than the CPUs themselves.

arknor · 29 Jul 2022 at 15:40

alec said:
have you looked at the chart? slow cpus on the bottom, fast cpus on the top?
what else do you want from cpu review?

if you were seeing with cpu is fastest you would run the lowest res possible with all graphics options set to the lowest

Wolverine2349 · 29 Jul 2022 at 16:43

Golden Cove has 15-20% higher IPC at same clock speed compared to Zen 3. I tested with Cinebench and CPU-Z single core both at 4.7GHz with same 3600MHz CL14 RAM and 12700K at 4.7GHz had a 16-17% higher score than Ryzen 5900X.

Games not showing that difference probably has more to do with being GPU bottlenecked even at 1080P in some cases that the gains Golden Cove shows are capped because the GPU still cannot be fast enough to show more gains in 1% lows even at 1080P. And some of it depends on the game as well as some may be better optimized for certain architectures and such. When faster video cards come out and games get updated I think Golden Cove will age much better than Zen 3 as long as games do not start to meaningfully benefit from more than 8 good/strong cores. And no I do not count the e-waste cores. Intel 12th gen are super high powered 8 core 16 thread parts as far as games are concerned as e-waste cores should be off as they drag the ring clock down and cause more trouble than they are worth.

alec · 29 Jul 2022 at 16:48

Wolverine2349 said:
Cinebench and CPU-Z

cpuz lives in its own world, but cinebench?
what are you doing differently from https://www.guru3d.com/articles_pages/core_i9_12900k_review,7.html
Edit: ah, they are still using CB15

Wolverine2349 · 29 Jul 2022 at 17:20

alec said:
cpuz lives in its own world, but cinebench?
what are you doing differently from https://www.guru3d.com/articles_pages/core_i9_12900k_review,7.html
Edit: ah, they are still using CB15

I used Cinebench R23, not R20 and CPU-Z.

And both were set to 4.7GHz not 3.5GHz.

I sold off the 5900X and no longer have it to test, but I did record the results when I did have it.

Both used 3600MHZ DDR4 CL14 RAM, though I think RAM has little to no impact on single core score

Ryzen 5900X locked at 4.7GHz CCD1:

Cinebench R23 single core: 1548

CPU-Z Single core: 642

Intel Core i7 12700K with ring clock 4.7GHz and all 8 P CPU cores set fixed to 4.7GHz with e-waste cores disabled of course

Cinebench R23 single core: 1855

CPU-Z single core: 767

Intel Core i7 12700K with ring clock 4.8GHz and all 8 P CPU cores set fixed to 5GHz with e-waste cores disabled of course (This is my normal stable config I use all the time)

Cinebench single core: 1975

CPU-Z Single Core: 815

I had turned my 12700K down to 4.7GHz just to do an IPC test as I knew what the Ryzen 5900X was capable of at 4.7GHz and wanted to see IPC comparison at same clock speed. The 5900X I could not get any CCD rock stable above 4.7GHz, so had to stop there with that.

CPU-Z single thread benchmark results: https://valid.x86.fr/bench/1

Cinebench R23 CPU single core results: https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r23_single_core-15

The results are taken without doing a fixed clock frequency. The Ryzen 5900X probably boosts higher especially in the Cinebench test than mine at 4.7GHz static/fixed

The Ryzen 5900X probably boosted very close to same 4.7GHz speed on single core speed in CPU-Z tests as results were within margin of error mine fixed at 4.7GHz compared to the chart.

My 12700K 5GHz has almost as good of single core score in Cinebench R23 and CPU-Z despite 5MB less cache and only 5GHz. Not sure if the test they use the 12700K boosted to less than 5GHz single core?? The 12900K is supposed to be able to do 5.2GHz single core and yet it barely had a higher score with 5MB more L3 cache than my 12700K meaning maybe it boosted not any higher as it was tested in more thermally constrained than my 12700K fixed all core at 5GHz with ring at 4.8GHz.

Kirby Wurm · 29 Jul 2022 at 17:22

It's funny how fanboys (not aimed at anyone that's posted here) flip-flop on whether running at low settings/low resolution is a worthwhile test depending on how strongly their preferred manufacturer is doing. It happened when Zen3 was released, and then Alder Lake.

I think testing both at high quality 1080p and 1440p and then lower settings/lower res is the way to go. That way you get to see the current real world differences and then closer to true scaling when GPU limitations are reduced, important for those planning to keep their CPU for some time.

Wolverine2349 · 29 Jul 2022 at 17:35

Kirby Wurm said:
It's funny how fanboys (not aimed at anyone that's posted here) flip-flop on whether running at low settings/low resolution is a worthwhile test depending on how strongly their preferred manufacturer is doing. It happened when Zen3 was released, and then Alder Lake.

I think testing both at high quality 1080p and 1440p and then lower settings/lower res is the way to go. That way you get to see the current real world differences and then closer to true scaling when GPU limitations are reduced, important for those planning to keep their CPU for some time.

Yeah I hear you. I am not a fanboy of either. I just like what is better. I did an IPC test to really see how much better at same clock speeds Golden Cove was compared to Zen 3. Many say that well Golden Cove only pulls so much ahead of Zen 3 because of higher clock speeds when that is only partially true. It also has modestly to much better IPC. Not Conroe destroying K8 IPC better (that was closer to 30%) but 15-20% per my tests. Specifically 16-17% in both Cinebench R23 and CPU-Z which means flat out Alder Lake P cores are clearly superior to Zen 3. Though I do not like the e-cores at all and I wish Intel did have an option for more than 8 P cores, though fortunately now like very few games have any benefit from more than 6 cores let alone 8 unless you are doing heavy streaming and such in background which I do not and it is so hard for developers in something like games to just scale to more and more threads and parallelization for something like games.

Though I applaud AMD for having more than 8 good cores, they still only have 8 on a single CCD/ring which means big latency penalty for threads that have to communicate across CCD/rings. Though Intel has less latency going to e-core, but the e-cores are crap so hurts in a different way for game thread getting stuck on e-core. That need fast consistent communication with each other beyond 6 or 8 cores which have to cross CCX.

That's why IMHO Intel is clearly better choice for 8 cores or less. For more than 8 cores for pure parallelized workloads especially where the threads do not need fast communication with each other, AMD is better option as they scale beyond more than 8 good cores.

Bencher · 29 Jul 2022 at 18:21

Wolverine2349 said:
I used Cinebench R23, not R20 and CPU-Z.

And both were set to 4.7GHz not 3.5GHz.

I sold off the 5900X and no longer have it to test, but I did record the results when I did have it.

Both used 3600MHZ DDR4 CL14 RAM, though I think RAM has little to no impact on single core score

Ryzen 5900X locked at 4.7GHz CCD1:

Cinebench R23 single core: 1548

CPU-Z Single core: 642

Intel Core i7 12700K with ring clock 4.7GHz and all 8 P CPU cores set fixed to 4.7GHz with e-waste cores disabled of course

Cinebench R23 single core: 1855

CPU-Z single core: 767

Intel Core i7 12700K with ring clock 4.8GHz and all 8 P CPU cores set fixed to 5GHz with e-waste cores disabled of course (This is my normal stable config I use all the time)

Cinebench single core: 1975

CPU-Z Single Core: 815

I had turned my 12700K down to 4.7GHz just to do an IPC test as I knew what the Ryzen 5900X was capable of at 4.7GHz and wanted to see IPC comparison at same clock speed. The 5900X I could not get any CCD rock stable above 4.7GHz, so had to stop there with that.

CPU-Z single thread benchmark results: https://valid.x86.fr/bench/1

Cinebench R23 CPU single core results: https://www.cpu-monkey.com/en/cpu_benchmark-cinebench_r23_single_core-15

The results are taken without doing a fixed clock frequency. The Ryzen 5900X probably boosts higher especially in the Cinebench test than mine at 4.7GHz static/fixed

The Ryzen 5900X probably boosted very close to same 4.7GHz speed on single core speed in CPU-Z tests as results were within margin of error mine fixed at 4.7GHz compared to the chart.

My 12700K 5GHz has almost as good of single core score in Cinebench R23 and CPU-Z despite 5MB less cache and only 5GHz. Not sure if the test they use the 12700K boosted to less than 5GHz single core?? The 12900K is supposed to be able to do 5.2GHz single core and yet it barely had a higher score with 5MB more L3 cache than my 12700K meaning maybe it boosted not any higher as it was tested in more thermally constrained than my 12700K fixed all core at 5GHz with ring at 4.8GHz.

My 12900k can get 927 cpu z ST with e cores on btw.