With the release of Alder Lake less than a week away and the "Lakes" thread having turned into a nightmare to navigate I thought it might be a good time to start a discussion thread solely for Alder Lake.
Since ADL is thought, in some quarters, to be a mobile first design adapted to desktop, that'll be interesting if Intel could manipulate Intel 7 at the silicon level like that since that'll probably be a historical first, and a departure from good old traditional binning as a differentiator between Desktop and Mobile.Just for having a physically smaller die, I would expect that the uncore (excluding the large iGPU and supporting circuits) draw of the 2+8 die to be lower simply due to the shorter distances that the electrical currents will have to traverse in most cases as every unit of length has a level of resistance associated with it, no matter how high a quality the silicon and metals are, and shorter distances will experience lower resistive losses. This is compared to the 8+8, 6+8 and 6+0 die. Yes, I also expect binning as to what product and market the individual packages will go into, but, I am questioning if Intel does anything at all at the process level for the lines that manufacture the 2+8 die that's different to the other Alder Lake die to increase efficiency.
| # P Cores | P Core Frequency | # E Cores | E Core Frequency | CB R23 |
| 2 | 3 | 8 | 3 | 9452 |
| 2 | 2.5 | 8 | 3 | 8994 |
| 8 | 2 | 4468 | ||
| 8 | 4.8 | 16 | 3.8 | 34582 |
| 6 | 4 | 8 | 3.4 | 18598 |
On voltage and power, are we even 100% sure that the 2+8 Alder Lake parts are being manufactured on the EXACT same process mix as the 8+8, 6+8 and 6+0 parts are?
[@Hulk NEVER assume linear.
That's quite similar to my calculations based on CB R23 score.@TESKATLIPOKA I stick with the expectations that 15W 2+8 will beat 28W Tigerlake by 30-50%.
I think part of the reason is the limited video memory available to Radeon iGPUs. I don't know about 5700G but Ryzen 3450U had only 1GB available to it on a 8GB laptop while i3-1115G4 had 4GB video memory it could use out of total 8GB. When a game can't even fit all its textures in there, it will lead to tons of swapping with the iGPU mostly just waiting.Even the 5700G is mostly agonizing and I'd rather be on a 1050ti or even ancient 770/960 level GPU.
I think part of the reason is the limited video memory available to Radeon iGPUs. I don't know about 5700G but Ryzen 3450U had only 1GB available to it on a 8GB laptop while i3-1115G4 had 4GB video memory it could use out of total 8GB.
Unigine Heaven can be forced to run despite insufficient video memory but on Ryzen 3450U, the medium benchmark spent more than 5 minutes on the loading screen at which point I had to kill it.Supposedly, some software isn't coded flexibly enough and bails on a run because the gpu reports what is CURRENTLY allocated istead of what can be moved as needed.
Alder Lake-N patches are in Linux:
[Intel-gfx] [PATCH] drm/i915/adl-n: Enable ADL-N platform
lists.freedesktop.org
"ADL-N from i915 point of view is subplatform of ADL-P."
I think part of the reason is the limited video memory available to Radeon iGPUs. I don't know about 5700G but Ryzen 3450U had only 1GB available to it on a 8GB laptop while i3-1115G4 had 4GB video memory it could use out of total 8GB. When a game can't even fit all its textures in there, it will lead to tons of swapping with the iGPU mostly just waiting.
I think the hypothetical 2P/2E part would have done pretty well as a 1080P gaming CPU, judging from these benchmarks and would have saved a lot of power too.No Surprise. E-Cores suck for gaming. I previously speculated this is why Non-K i5 have 6+0, instead of 4+8:
I've used only one Radeon iGPU so far, so it's quite possible that the Gateway laptop I used had a really bad implementation of it, BIOS-wise. I couldn't find any BIOS option to increase the iGPU's usable RAM.Actually, thats a fairly complicated subject...
Radeon iGPUs should have the memory that is set as fixed in bios + 2GB dynamic VRAM.
But you have some games that:
-Only use the fixed memory (Jedi Knight The Fallen Order)
-Only use the dynamic memory
-Use both.
But im not sure if this is a problem with the games, the DX version used, the drivers or wharever. In the case of AMD, once "4G decoding" is enabled, it allows to set up to 50% of the installed ram as fixed memory. So it is possible to have 4, 6, 8GB VRAM. The problem is that you cant really disable dynamic vram.
On Intel im not sure if the behaviour is the same per se, but i know the dynamic vram is larger than 2GB, i think it can go up to 4 or 8GB. Not sure if "4G decoding" is needed for that.
I mean this testing is irrelevant though, because they're never used in isolation. It's a nice thought exercise with no real world impact.No Surprise. E-Cores suck for gaming. I previously speculated this is why Non-K i5 have 6+0, instead of 4+8:
I mean this testing is irrelevant though, because they're never used in isolation. It's a nice thought exercise with no real world impact.
The same review is also on Techspot. Are they collaborating or something?No Surprise. E-Cores suck for gaming. I previously speculated this is why Non-K i5 have 6+0, instead of 4+8:
On Intel im not sure if the behaviour is the same per se, but i know the dynamic vram is larger than 2GB, i think it can go up to 4 or 8GB. Not sure if "4G decoding" is needed for that.