• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Intel 12th Generation Alder Lake Platform Reportedly Brings 20% Single-Threaded Performance Uplift

Any news about how many Cores will the top dog have? Frequencies? Wattage? Something to forget this mizerabile fiasco called 11th generation :)
Top part will be 8+8 cores, no avx512 or just 8 cores with avx512 turned on (will need a restart to select the mode).
Performance is likely slightly better than current tiger lakes in single threaded loads and a bit less performant than 12 core amd chip in multi threaded loads. In the rare case where avx512 is of any actual use, it will be pretty ok.
 
Any news about how many Cores will the top dog have? Frequencies? Wattage? Something to forget this mizerabile fiasco called 11th generation :)
The top dog will be a 16C/24T thing, no idea on Freq/TDP yet as it relies on Intel fixing their FUBAR 10nm or pulling another Rocket Lake and backporting it to 14+++(+?)

It's basically going to be a "cove" 8C/16T big core with 8 puny Atom cores that handle low-priority, low power background stuff. The labelling convention that people are unofficially using (me included) is perhaps misleading though: 16C/24T implies all cores are equal which they obviously aren't. It also implies that instructions can be shifted around between big and little cores easily.

The reality is that the big cores will look very much like an 11th-gen i5/i7/19, but hopefully on 10nm if Intel can actually get it working properly. The little cores won't have the same instruction set. Sure, it's still basic x86-64 but "cove" and "mont" architectures are quite different internally - you can just move instructions from one core to another if they're not the same core and it won't even be a hardware issue. Remember how long it took Microsoft to get the scheduler right for Zen2? That was just two identical sets of four cores on indentical architecture. I suspect Intel's Big.Little architecture attempts might take Microsoft half a decade to really take advantage of properly. It would be nice if I was wrong - I genuinely WANT to be wrong but we're talking about a company that had 2+ years of prior notice to getsomething as simple as CCX scheduling worked out, and they still missed the launch by several months. This is the same company that, despite promising to do so a decade ago under Sinovsky still hasn't finished moving some of their underlying OS technology off Windows NT whilst they instead focus on new icon art and ways to sell and use your personal data for profit.

So yeah to get code to run seamlessly across two different core types, concurrently, with different cache, different instruction set support, at different clockspeeds, priorities, registers.... OMG. Google and Apple might be able to do it because their underlying OS isn't a dumpster fire that's been burning for two decades. Meanwhile, Microsoft - I, uh. I have no words. Here are some other peoples' words:

 
Last edited:
Is DDR5 confirmed for ALL Z960 chipset mobos, or just for the most expensive Z690 ones? And guessing it will come with 2 slots DDR 4 + 2 slots DDR 5 ?? Therefore limiting the total RAM it can be installed?
Also both the cipset and the CPU will support PCI-E 4.0 meaning at least double the PCI-E 4 lines compared to the Z590?
 
And guessing it will come with 2 slots DDR 4 + 2 slots DDR 5 ??
Maybe some silly mobo will do that, but most will definitely be either 4 slots of ddr5 or 4 slots of ddr4 depending on the model. There would be a bunch of unused components on board to support ddr4 making such multipurpose mobos more expensive than pure ddr5.
 
Maybe some silly mobo will do that, but most will definitely be either 4 slots of ddr5 or 4 slots of ddr4 depending on the model. There would be a bunch of unused components on board to support ddr4 making such multipurpose mobos more expensive than pure ddr5.
I hope you are right, but remember the DDR3->DDR4 transition. ;)
Btw, I'm still on DDR3. How interesting it will be to skip entirely DDR4 and go for DDR5 for my next upgrade build. But don't see anywhere those 8Ghz modules unfortunately....
 
I hope you are right, but remember the DDR3->DDR4 transition. ;)
Btw, I'm still on DDR3. How interesting it will be to skip entirely DDR4 and go for DDR5 for my next upgrade build. But don't see anywhere those 8Ghz modules unfortunately....
But who needs 8GHz, when you can get 512GB sticks at 7.2GHz ;)

All you need is a 10% overclock to get to the 8GHz you dream of.
 
But who needs 8GHz, when you can get 512GB sticks at 7.2GHz ;)

All you need is a 10% overclock to get to the 8GHz you dream of.
I only need 16GB, ok, 32 if they are cheap. And yeah. Those 7.2Ghz modules looks good.
Except for the timings.
 
I only need 16GB, ok, 32 if they are cheap. And yeah. Those 7.2Ghz modules looks good.
Except for the timings.
No-one has made chips available yet that would enable less than 32GB pairs of sticks to be made. So for 16 you’d be running single channel.

The specifications do permit 8GB sticks though, so maybe some chipmaker will eventually make the chips needed for those as well.
 
I’d settle for 20%, but 20% less heat, less cores, less Watts.

Just make a better CPU, not “better” marketing campaigns.

DDR5 and PCI 5 can wait. Put the Horse before the Cart this time.
 
Back
Top