• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Raptor Lake Leaks + Intel 4 developments

What difference does it make if it's 4K or not? A CPU intensive game will still be CPU intensive regardless of the display resolution. It is true that limiting the framerate can help to prevent spikes in CPU utilization, which may cause drops in FPS.

It's easy to max out the cores on strategy games, just by forcing games to handle more units in a battle, or the size of the units (e.g. total war games).

I suppose the reason games don't generally consume as much power as they do in benchmarks, is mostly due to the CPU instructions used (not much AVX/AVX2), lower overall multithreaded utilization, and lower cache utilization in most cases.
 
Last edited:
it doesn't work like that, they don't just scale onto different cores infinitely.

some stuff can go on extra cores sure but they can't equally split everything, one part of the process will always limit the rests speed as well
splitting hairs now. It's true that single core performance counts for more in games.

The point I was making, is that it's easy to push CPUs to their limits in modern strategy games with lots of units (e.g Total War).
 
That's why pc games need direct storage, get rid of the cpu and ram
Watchdogs: Legion is another CPU intensive game that has high utilization across 8 threads /physical cores. Turning on hyperthreading definitely benefits this game.

I wonder if that's true for all games with high utilization on 8 CPU threads/physical cores?
 
but it's not because they won't use 20 cores or 50 cores
Yeah, tbf WH3 really doesn't benefit from hyperthreading at all (it can cause framerate dips). It doesn't seem to make a difference if more than 8 threads are set, or not. Also, they removed DX12 support due to stability problems, which doesn't help either.

Increasingly, I think L3 cache (and likely the cache ratio / clock rate) seems to be the thing that makes the most difference to framerates, particularly as IPC and clock rates have improved.
 
Last edited:
Adding more E-cores doesn't make CPUs more efficient. P-Cores running at their base frequency (same as Zen3/4 cores) will give the best performance per watt. E-cores are what Intel does when they want more multithreaded performance for laptops, that can't handle more P-cores due to cooling constraints. The sapphire rapids server CPUs will mostly be ran at, or near the base clocks.

Nuff said. So yeah, Intel needs a high end design without E-cores. Also, P cores on 10nm ESF lose their efficiency at very high clocks (probably around 5ghz), and start chewing up a ton of power.

Intel is going to need their 7nm EUV ('Intel 4') technology to improve the efficiency of future CPUs, so prioritizing the Meteor Lake release will be very important for their future progress.
 
It's not great, basically forces those buying high end h/w to use £100 or more water coolers and higher watt power supplies.

I'd say this is going to be less of a problem with Zen 4, it looks like the power consumption maxes out at 230w on the 12 and 16 core designs.

The 8 core 7700X should max out at <150w.

The reason for their apparent efficiency, is the much higher base clocks, compared to Zen 3. The 7700X has a base clock of 4.5ghz. The 5800X's was just 3.8ghz, at these frequencies both CPUs consume about 105w.
 
It's obvious that they do. The cooling systems on laptops can only handle upto a certain amount of power consumption (~150w?). E-cores weren't designed with desktop PCs (particularly gamers) in mind.

Adding more E-cores running at lower frequencies allows higher multithreaded perf, at a lower cost in power than P-Cores run at high frequencies.

There's still a strong argument to just use more P-cores at lower frequencies though, instead of using E-cores. EDIT - there shouldn't be a limit on the number of P-Cores in the design, as Sapphire Rapids will use a lot more than 8 P-Cores.
 
You do realize that your point is self refuting right? 8 p cores would outperform 8 e cores at the 150w you mentioned, so it's obvious what you are saying does not make any sense. E cores are more efficient only at very very low power (2-3w per core) and even then it's only for specific workloads. Again, the reason they use them is not power or thermals.
It makes sense if you actually think about it.

They can only have so many P-Cores, before hitting the cooling limit (~150w I'm guessing). If there's remaining cooling capacity, this can be used for E-cores.

Some times they use a lot of E-cores, just do reduce cost and cooling requirements on the lower end models.
 
Lower base clocks for the entire 13th generation, at the same TDPs (vs 12th gen parts):

These aren't going to be more power efficient parts than Alder Lake.
 
AMD is significantly increasing the base clocks with Zen 4 (at the same power usage as Zen 3), not decreasing them. So, it matters.

EDIT - In fairness though, the 12/16 core CPUs will undoubtedly draw more power than Zen 3 CPUs, but will be coolable with many air coolers.
 
The new Kingston 6400 CL32 kits look very decent, and while still kind of expensive are not silly money like they have been for high speed kits, starting to be a good time to buy into DDR5 :)
I would go with the 6400 MT/s kits if on Intel's 12/13th gen also. Not sure yet if this will be optimal on AM5 systems though.

They haven't produced 6400 MT/s 8GB modules yet though, maybe soon?
 
Almost switched to Alder Lake this year from Skylake but decided to wait for better value and to get some more E cores. Now realising that Intel is putting up processor prices, that 6## series motherboard prices went up by about £30 a couple of months ago (and raptor lake mobo prices will be even higher). Normally in this game you'd just buy the lower end of the next generation for more performance and lower cost, could alder lake to raptor lake be the first time that's not the case?
I think you'll get more multithreaded perf. from the extra E-cores at the lower end, but about the same single core perf. There could be a slight bump up in cost. On the other hand, Id expect the 13700K to be a lot cheaper than a 12900K (and KS) is now (very similar spec).
 
No mention at Hotchips of the 13th gen. Arrow Lake and Lunar Lake apparently confirmed. Link:

Intel is thinking well ahead now. Foveros packaging confirmed for the 14th, 15th and 16th generations.
 
Anyone with a 12th gen Intel CPU gonna skip the 13th gen, and buy a new platform + Meteor Lake CPU (which will probably release on desktops) next year? If you have DDR5, you could carry that over too.
 
People will buy whatever CPU is cheapest and gives the best single core performance (unless they already have a LGA1700 board).

Multicore performance is what Intel focuses when they are struggling to improve IPC (which would improve overall ST or MT performance anyway).

They are doing this more with the 13th gen, as it looks like models without E-cores will be only at the very low end.

Next time around (with Meteor Lake), Intel will actually be able to improve IPC and reduce power consumption (multithreaded gains from E-cores will simply be a bonus to performance).
 
Intels single core performance on ADL was already good and RPL will increase this, zen 4 is only bridging this gap. Also it's the platform costs that need to be taken into account and this time Intel will have the advantage despite the increased CPU costs.
Not sure. It does look like the Z790 boards are limited to ~6800 MT/s with a single module, 6400 MT/s with 2 modules. At least according to this rumour:

This certainly isn't a bad spec, should be plenty for most. But, it looks like it will be possible to use higher frequency modules (>6000 MT/s) on AM5 boards releasing this year, in the future.
 
I don't think the "supported" spec matters for either brand, people will push past it anyway, people have already gotten ddr5 to 8000mt/s on alder lake
Don't you have to drop down the imc to 'gear 4' on the 12th gen though, at some point?

RAM speed support matters, because you want it to be stable and perform well.

It doesn't seem a stretch to think that 7000 MT/s modules will be available to buy within the next 3-6 months, considering that 6600 MT/s modules run at 1.4v.
 
I've been looking at the top spec modules to date on DDR5 just now. My impression is that Samsung is probably the furthest ahead with DDR5, with 7200 MT/s 512GB modules possible (for data centres etc), and this was back in August 2021. The most impressive thing is, this could be achieved with just 1.1v!

Spec here:

eUMDaTfecBDb9J4bQiHii3-970-80.jpg.webp


Article here:

Shows that there's still plenty of room for DDR5 to develop, but it makes me wonder why we haven't seen voltages come down on consumer DDR5 modules? With lower voltages possible, there should be no need to resort to costly cooling solutions for RAM modules...
 
Back
Top Bottom