Horus-Anhur:

Not as big as you think. Because people are counting all the caches. It's just a 1MB increase in L2 cache for the P-Cores and another 1MB for each 4 cluster of E-cores. But because these CPUs also have more E-cores, the overall cache number has to increase with it. The thing is, we could replace several of those E-cores with more cache, and have a greater IPC improvement. If Intel doesn't find a way to block it...

https://pyxis.nymag.com/v1/imgs/7a1/d19/dba7fbefd9f9c97c1df535b94a5d02eb6c-08-debbie-downer.2x.rsocial.w600.png

cucaulay malkin:

https://pyxis.nymag.com/v1/imgs/7a1/d19/dba7fbefd9f9c97c1df535b94a5d02eb6c-08-debbie-downer.2x.rsocial.w600.png

Have you changed your mind about e-cores? A few days ago you were trashing the idea of more e-cores.

Horus-Anhur:

A few days ago you were also against adding more e-cores, and preferred more P-cores.

on a +600 usd 13900k,your memory is very selective.adding e-cores to budget cpus is actually really nice,I said that a thousand times, again selective memory.

cucaulay malkin:

on a +600 usd 13900k,your memory is very selective.adding e-cores to budget cpus is actually really nice,I said that a thousand times, again selective memory.

But how many? Having one 4 e-core cluster is not a terrible idea. It helps with power usage and can improve select applications. My issue is with adding more and more e-cores. The 13900K is a prime example of this bad idea.

Horus-Anhur:

But how many? Having one 4 e-core cluster is not a terrible idea. It helps with power usage and can improve select applications. My issue is with adding more and more e-cores. The 13900K is a prime example of this bad idea.

yes,that is exactly what I said.

cucaulay malkin:

yes,that is exactly what I said.

Then we agree. Why are we arguing about...

cucaulay malkin:

yes,that is exactly what I said.

Horus-Anhur:

Then we agree. Why are we arguing about...

[youtube=t-4ADaq9og4]

Horus-Anhur:

Then we agree. Why are we arguing about...

were we arguing ? when ?

cucaulay malkin:

were we arguing ? when ?

Maybe a better term, would be "discussing with enthusiasm". 😀

When you compare CPU values (Intel vs AMD) you should take in consideration motherboard prices also... Do that and it is a whole new perspective.

wavetrex:

Some years ago when there were just 5% improvements gen to gen, the word around was that "all low hanging fruits have been picked", and there won't ever be large jumps in performance anymore. Yet, somehow, these days, both companies manage to jump 15-20% gen to gen in both single and multithreaded. Looks like there are plenty more fruits to be picked...

The quotes about low hanging fruits remain true because they are about the micro-architecture of CPUs, while in this case the performance is coming from higher clocks and more cache. In the end, there are no miracles or magic solutions, Intel and AMD are doing what they can to improve performance, but they are now using a more brute approach instead of a more elegant and sophisticated one, like improving their micro-architecture in such a way that they can extract more performance without having to increase clocks or caches, that have their own trade-offs.

Seems Cinebnch didn't get the memo that the CPU is called the 7700X, not the 7701X.

wavetrex:

Some years ago when there were just 5% improvements gen to gen, the word around was that "all low hanging fruits have been picked", and there won't ever be large jumps in performance anymore. Yet, somehow, these days, both companies manage to jump 15-20% gen to gen in both single and multithreaded. Looks like there are plenty more fruits to be picked...

Back then, that was legitimately true, and I'd say it still is to some degree. The only reason we've seen large jumps in performance is because devs are finally starting to optimize their code for more threads and more instructions. OSes are also getting a lot smarter with their schedulers. The bigger caches we see lately also make a bit of a difference. As @H83 said, a 15-20% jump is largely due to things like higher clock speeds. Where the IPC increase is higher, that's either because of newer instructions that devs or compilers have to use, or, it's because of other natural evolutions like faster memory and smaller transistors. There is still very little that can be done to improve the per-clock performance of single-threaded already-compiled applications.

Horus-Anhur:

Could be. It's curious how 5 years ago, it was AMD slapping more and more cores on their CPUs. But now it's Intel. Though it's those e-waste cores.

Calling the cores e-waste isn't fun, isn't either correct.

You are correct, the e-cores are pretty much the speed of Skylake without HT, so 4 of those E-cores are roughly the strength of a 6600K, while consuming a lot less energy. It's not bad at all... ... the problem is the P-cores are gas guzzling SUVs, and a lot less efficient than AMD's Zen3 or 4 cores. No wonder they can't put more than 8 of them in a cpu

tty8k:

Since multithread is what matters in 3D render, this 7700x gets literally destroyed by its counterpart the 13700k as well as 12700k. But hey it finally beats i5 12600k!

Lol.

Wccftech? 771 cb20 4.5GHz,5.3GHz...?,the IPC increase is above 8٪.

what all this e-core bs talkin, ofcoruse it will add up to multi score, its more threads afterall.. Real picture would be only those P-cores, my Pcore RL gets ~6350 at 5Ghz. So this 7700x is quite strong at 4.5ghz imo

I do not care for moar fps: Will it increase Lua single-threaded interpreter performance? Will it improve image (like JPEG) single-threaded decoding times and by proxy PDF decoding? Will it increase JavaScript single-threaded interpreter/JIT performance? Will it increase Java/C#/whatever run-time machine code compilation performance? Will it improve turn-times of strategy games like Total War and Civilization?

schmidtbag:

Seems Cinebnch didn't get the memo that the CPU is called the 7700X, not the 7701X. Back then, that was legitimately true, and I'd say it still is to some degree. The only reason we've seen large jumps in performance is because devs are finally starting to optimize their code for more threads and more instructions. OSes are also getting a lot smarter with their schedulers. The bigger caches we see lately also make a bit of a difference. As @H83 said, a 15-20% jump is largely due to things like higher clock speeds. Where the IPC increase is higher, that's either because of newer instructions that devs or compilers have to use, or, it's because of other natural evolutions like faster memory and smaller transistors. There is still very little that can be done to improve the per-clock performance of single-threaded already-compiled applications.

Really? Prove it. You just said the performance increases are almost solely due to higher clock speeds. Then explain why the jump from Zen+ to Zen 2 was so dramatic? The jump from the 2600 to the 3600 was a near 40% bump in multithreaded performance across the board at near the same operating frequency. The 3600 clocked at 4.1GHz and the 2600 clocked at 4GHz for me, so I'm genuinely curious as to how you acquire a sub 10% clockspeed increase but somehow gain nearly 40% performance? And yeah, Zen 4 is largely a clock boost over Zen 3. It's reported roughly to be a 20-25% increase. That's nothing to scoff at, especially when the maintained operating frequency in programs is nearing 1GHz higher than the previous generation. Also I really don't understand how anyone can possibly make the claim of "There's not much performance left on the table" when we've already seen AMD completely obliterate Intel at their own game. They aren't even remotely concerned about Intel anymore. They're overtaking them in the server market share, they're likely to overtake them in the consumer market within the next few years at the VERY least, and they've set their sights on Apple as a true competitor for the future. Not to sound like a fanboy of AMD or anything, but unless Intel can do anything with this hybrid x86 arch they've got, I'm afraid the writing is on the wall. Zen is more efficient, which matters faaar more than raw performance. Also worth noting that almost zero programs are actually truly multithreaded in practice. It's actually quite hard to code something to utilize several cores simultaneously and properly at that.

Not sure why people suddenly have faith in Microsoft Windows schedular , last I looked the E-Core just sit idle in windows 11. So like people are going to buy 12/13 gen CPUs just to get half outta it. lol

~AngusHades~:

Not sure why people suddenly have faith in Microsoft Windows schedular , last I looked the E-Core just sit idle in windows 11. So like people are going to buy 12/13 gen CPUs just to get half outta it. lol

Are you sure that is the Scheduler fault? Is there enough load on the CPU to use the E-cores instead of sending all work to the faster p-cores? Loading up all cores on a 12900k or 13900k needs quite a bit of work and good coding that allows a lot of parallel instructions. For gaming a 12900k is mostly overkill and games are not able to use all cores, but if background work is done on the pc at the same time, the work should find its way to the e-cores. A 5950 or a threadripper has the exact same problem, if the workload is not there, the excess CPU cores will just sleep waiting for work. Rendering already works fine with e-cores.