Intel Core i9-13900K

[Mussels]

here is a review on this topic and E-core messes up in some games and helps in others.

Or TPU's review on just the E-cores, which shows them worse and slower than even ryzen 3000.
If the E-cores provide any benefit, it's because 6 P cores isnt enough - and having anything faster than them would provide even greater gains.


An 8P 0E intel CPU would be the gamers chip from intel, but they REALLY want to sell you E-cores for a higher price

 

[JustBenching]

Or TPU's review on just the E-cores, which shows them worse and slower than even ryzen 3000.
If the E-cores provide any benefit, it's because 6 P cores isnt enough - and having anything faster than them would provide even greater gains.


An 8P 0E intel CPU would be the gamers chip from intel, but they REALLY want to sell you E-cores for a higher price

My 13900k performs much better with ecores on as well, eg in cyberpunk

 

[Gica]

Or TPU's review on just the E-cores, which shows them worse and slower than even ryzen 3000.

Correct, but I'm starting again with comparisons that you don't like.
Archers (artillerists, now) are the most fragile in an army if they have no defense. They are even crushed by a gang of drunken soldiers led by a retarded sergeant. Protected, they can prove to be a lethal force, sometimes decisive.
This is about your E-core versus Ryzen 3600 comparison.
E-core is like CUDA at launch, with the problems related to the beginnings. Fortunately, they are small and do not get in the way in an annoying way. It was that beginning with x games that had problems with these E-cores, but ... problem solved. When, paradoxically, these small cores also help in single-threaded (in multi-threaded they help enormously) I don't see why they wouldn't help in gaming as well. In fact, it helps in some that the developers have done their homework on.

An 8P 0E intel CPU would be the gamers chip from intel, but they REALLY want to sell you E-cores for a higher price

Why? Thanks to these E-cores, i5 no longer has ryzen 5 as an opponent in multi-threaded, it fights with 7700/7700X, but costs less than ryzen 7.
13500 is placed between 7700 and 7700X in multithreading, thanks to this E-core. The cost is much lower, slightly above ryzen 5.
Conclusion: at least on some levels, these E-cores made the processor cheaper. Moreover, they also forced AMD to cut prices. What do you have with them?

 

[ThrashZone]

Or TPU's review on just the E-cores, which shows them worse and slower than even ryzen 3000.

Hi,
Of course e-voodoo threads are space heaters so getting performance is just in numbers not frequency performance.

 

[Gica]

They don't get hot for nothing. Those "E-Vodoo" bring a 1/3 increase in performance in multi-threading and even a marginal increase in single. Marginal, but it exists.
One more thing: when ryzen doesn't perform at Intel's level in gaming, AMD fans force the discussion to encodings, renderings... from these where ryzen does well. Now, when an i5 beats an r7 in many, no AMD fan renders or encodes anymore? Everyone wants gaming and only gaming? I'm asking for a friend.
I can redo the video with 12500, now with 13500. In idle, the same consumption, but in youtube (especially 4k@60FPS and 8K@60FPS) the consumption is lower. Not by much, but it is. Likewise in WoT, which surprised me because I was expecting somewhat higher consumption in these tasks with lower CPU usage due to the number of cores.
As the 13500 is a 12500 with a higher frequency and 8 extra E-cores, I attribute this lower consumption to these E-cores. This is their first role. The second is to bring a massive increase in performance in multi-threading.
Compare the two captures, maybe they will convince you.

 

[ThrashZone]

Hi,
Yes they do get hot that's why that usually max out at 4.3 turbo
What you're looking at is quantity of threads not quality of threads.

If they were all quality threads/ real cores you'd see better results.

 

[InVasMani]

I think the shared L2 cache on the E cores is something that gets overlooked a lot especially when the P core and E core comparisons at same frequency. When lightly loaded and with the right involved task a single E core could outperform a P core if the L2 cache difference is significant enough. A single E core accessing the full L2 cache has more than a individual P core. It's in a somewhat reminiscent situation to the CCX latency matter with AMD. The E cores do better when more lightly and moderately loaded and same could be said with CCX latency issue.

 

[Gica]

Hi,
Yes they do get hot that's why that usually max out at 4.3 turbo
What you're looking at is quantity of threads not quality of threads.

If they were all quality threads/ real cores you'd see better results.

Cinebench R23 is a tool to measure the performances in Cinema 4D, a rendering and modeling program.
The differences are visible:
12500 (does not have E-cores): 12000 pts
13500 P-cores: 14000 pts (runs at higher frequencies in multi-threading)
13500 P+E-cores: 21000 pts.
Notable difference, don't you think?
Thanks to these small cores, in multi-threading, core i5 rivals ryzen 7 7000, not ryzen 5 7000. It fights with ryzen 7 7000 in everything related to multi-threading, from renderings to encodings and what else you want.
P.S. Max. 3.5GHz for 13500.

 

[ThrashZone]

Hi,
20 threads against 12 threads and you wonder why the 20 thread chip does better, but then again almost double the threads and not double the performance, interesting

 

[JustBenching]

Hi,
20 threads against 12 threads and you wonder why the 20 thread chip does better, but then again almost double the threads and not double the performance, interesting

Why do the number of threads or cores matter? It's a useless metric. What matters is performance and price, that's it. Whether it needs 100000 cores or 1 core for the performance is completely irrelevant.

 

[ThrashZone]

Why do the number of threads or cores matter? It's a useless metric. What matters is performance and price, that's it. Whether it needs 100000 cores or 1 core for the performance is completely irrelevant.

Hi,
Yeah add frequency and you're done

Beside Gica brought up performance sorry you failed to follow the conversation.

 

[Gica]

Hi,
20 threads against 12 threads and you wonder why the 20 thread chip does better, but then again almost double the threads and not double the performance, interesting

Core i5 13th (non K): price of ryzen 5 with performance of ryzen 7. This must be remembered. Maybe there are people who buy just to count the cores. I met a case that screamed loudly that 300 fps is better than 250 fps (on paper, it is, but only that) and he was using a monitor with 8 ms latency, from the time of Napoleon. He was getting gas from Putin to start it.

 

[ThrashZone]

Core i5 13th (non K): price of ryzen 5 with performance of ryzen 7. This must be remembered. Maybe there are people who buy just to count the cores. I met a case that screamed loudly that 300 fps is better than 250 fps (on paper, it is, but only that) and he was using a monitor with 8 ms latency, from the time of Napoleon. He was getting gas from Putin to start it.

Hi,
I'm not caring about amd-intel comparison performance wise

You threw out the 12500-13500 performance numbers and I countered the difference in threads as why 13500 was higher scoring
What I didn't add until later was thread frequency differences = this is an obvious factor so I did not state it to you.

If you want to split hairs add the memory frequency into the equation and while you're at it cache frequencies as well

 

[Gica]

I detect a logic fracture as big as the Pentagon.
Naturally, a number of E-cores cannot match the performance of the same number of P-cores, but they don't consume that much either. Strictly related to the 13500, with these E-cores it fights with the 7700X in multi-threading and, wow!!!, it's more efficient. The difference between them is small in terms of performance but colossal in terms of purchase price. For me, who does not dare to sell DDR4 memories, the price difference is astronomical between the two platforms.

 

[ValenOne]

If you are willing to spend ~2000$ on a video card, you will not look at the 200-250$ you pay extra for a processor. The most efficient are pentium, i3 or r3, not i9 or r9. The most balanced, for gaming, are i5 or r5.
If these E-cores raise the scores in the benchmark, they certainly also raise the performance in the applications. Even in some games.


I still don't understand why you bother to prove... what?
I said that those E-cores have their importance even in gaming and the reviews prove it. In applications that do not intensively use the processor, they also have their importance. As 1+1 = 2, when the 7950X consumes less than the 13900K in CPU killer applications, but consumes more in others, called "single-threaded", surely these E-cores have a role. This role translates into taking over P-core tasks and increasing energy efficiency. Even in applications, such as Premiere, with a processor that varies from 5-100% usage, the verdict of those who tested it was: same shit.
What is the connection between Cinebench and Cyberpunk?

PS: 6P + 8E = 20 hardware threads. Not 26.

Zen 4's tuned memory scaling. 3DV cache mitigates this issue.

 

[Gica]

...

Today I ran PCMark10 Extended and the comparison with 12500 offers something very interesting. 13500 achieves better results thanks to higher frequencies (a little from memory and cache), but the colossal jump, the undeniable impact of E-core, can be found in a synthetic test used for gaming performance (included in 3DMark). Interesting, right?

 

[ValenOne]

Today I ran PCMark10 Extended and the comparison with 12500 offers something very interesting. 13500 achieves better results thanks to higher frequencies (a little from memory and cache), but the colossal jump, the undeniable impact of E-core, can be found in a synthetic test used for gaming performance (included in 3DMark). Interesting, right?

View attachment 284749

From https://hothardware.com/reviews/amd-ryzen-5-7600x-and-ryzen-7-7700x-review

PC Mark 10 has a CPU scaling issue.

 

[Gica]

PC Mark 10 has a CPU scaling issue.

A large part of the results are influenced by the processing power of the video card, but Physics only uses the processor. That is the big difference between 13500 and 12500. This tool is integrated in 3DMark.

Comparison of 13500 with 7600X Xtreme overclocking in 3DMark (CPU Score and CPU Profile).

 

[Gica]

12th - AVX 512 enabled with BIOS version F2.
However, I prefer the latest version of BIOS, but it is good to know for situations in which I have to calculate the trajectory of dust particles when my wife uses the vacuum cleaner.
You can activate these functions only if the processor has embedded the old Intel logo.

 

[ValenOne]

A large part of the results are influenced by the processing power of the video card, but Physics only uses the processor. That is the big difference between 13500 and 12500. This tool is integrated in 3DMark.

Comparison of 13500 with 7600X Xtreme overclocking in 3DMark (CPU Score and CPU Profile).

7600X has 12 threads.
13500 has 12 threads from P-cores and 8 threads from E-cores.

LGA 1700 is still a dead platform with a missing discrete PCIe 5.0 4X NVMe.

 

[Gica]

7600X has 12 threads.
13500 has 12 threads from P-cores and 8 threads from E-cores.

LGA 1700 is still a dead platform with a missing discrete PCIe 5.0 4X NVMe.

7600X and 13500 they have the same price.
Total death. Without PCIe 5 you can't even start the computer.

 

[Mussels]

Total death. Without PCIe 5 you can't even start the computer.

What?

 

[Gica]

What?

It's just irony. PCIe 3.0, 4.0, 5.0 ... enormous differences in synthetic tests, negligible in practice. I don't think you'll notice that a PCIE 4.0 was mistakenly set to 3.0.
SATA classics still have life in them.

 

[mahanddeem]

It's just irony. PCIe 3.0, 4.0, 5.0 ... enormous differences in synthetic tests, negligible in practice. I don't think you'll notice that a PCIE 4.0 was mistakenly set to 3.0.
SATA classics still have life in them.

Obviously post you replied to was sarcastic

 

[ValenOne]

It's just irony. PCIe 3.0, 4.0, 5.0 ... enormous differences in synthetic tests, negligible in practice. I don't think you'll notice that a PCIE 4.0 was mistakenly set to 3.0.
SATA classics still have life in them.

NVMe is useful for video editing and database.