Question Alder Lake - Official Thread

[Hulk]

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.

 

[dullard]

So my take... Performance on sustained workloads, good, but a lot higher power use than 5950x. Gaming, they win by a small amount. For my use case (sustained 100% load, all cores) the power usage is a killer, so I won't be getting these. For a gamer ? its a winner. But the 12600k or 12700k....

Edit: I think the low power usage in games is due to the E cores being used primarily, as I doubt the CPU load is high, so the scheduler selects the E cores ????

What is the performance per watt comparison look like in your use if you limit Alder Lake power to say 125W?

 

[Markfw]

What is the performance per watt comparison look like in your use if you limit Alder Lake power to say 125W?

I did not see a benchmark out there on that. Did you see one ?

 

[blckgrffn]

So my take... Performance on sustained workloads, good, but a lot higher power use than 5950x. Gaming, they win by a small amount. For my use case (sustained 100% load, all cores) the power usage is a killer, so I won't be getting these. For a gamer ? its a winner. But the 12600k or 12700k....

Edit: I think the low power usage in games is due to the E cores being used primarily, as I doubt the CPU load is high, so the scheduler selects the E cores ????

If you are on Zen 3, it's probably a wait and see what comes next scenario.

But if you are on any Intel based system or earlier Zen platform and want to throw a bunch of money at it (probably maybe you got a big GPU already) then it is interesting.

Or for the HFR gamers types who care about 1080P/Low presets performance.

On the whole, it's good that it's competitive as that will continue to drive value, performance and forum arguments conversation topics for all of us

Intel releasing something solid is a really good thing.

 

[insertcarehere]

Gotta save the pennies for the GPUs, you know!

Speaking of GPUs, Alder Lake not needing any discrete GPUs in a PC build while the big boy Zen 3s do does seem particularly valuable during this time....

I did not see a benchmark out there on that. Did you see one ?

https://www.computerbase.de/2021-11/intel-core-i9-12900k-i7-12700k-i5-12600k-test/2/

According to this 12900k @125w is roughly equivalent to a 5900X in MT (at presumably similar power levels).

 

[dullard]

I did not see a benchmark out there on that. Did you see one ?

Oh, I thought that since you had reached a conclusion that you had seen a benchmark first. My mistake.

 

[igor_kavinski]

From the AT review:

What’s however a bit perplexing is that the core-to-core latencies between Gracemont cores is extremely slow, and that’s quite unintuitive as one would have expected coherency between them to be isolated purely on their local L2 cluster. Instead, what seems to be happening is that even between two cores in a cluster, requests have to travel out to the L3 ring, and come back to the very same pathway. That’s quite weird, and we don’t have a good explanation as to why Intel would do this.

Why is that weird? L3 cache lookup needs to happen during any request. Is Ian suggesting that Intel should prevent L3 lookups for all inter-core requests in the E-core cluster? That will lead to an even slower main memory access for most workloads.

 

[blckgrffn]

Speaking of GPUs, Alder Lake not needing any discrete GPUs in a PC build while the big boy Zen 3s do does seem particularly valuable during this time....


https://www.computerbase.de/2021-11/intel-core-i9-12900k-i7-12700k-i5-12600k-test/2/

According to this 12900k @125w is roughly equivalent to a 5900X in MT (at presumably similar power levels).

Meh. What does "particularly valuable" mean?

To me it's worth ~$15 to $20, max, even now you can get a used Quadro with DVI and display port for that price on eBay with its own memory to boot. I mean I won't build with any F processors because you basically never save more than but... the number/percentage of folks buying an i9 and using the integrated graphics is what?

To me on the high end Ryzens and even if the i9 didn't have an iGPU it woudn't change my mind about them at all.

All these comparisons to the 5600X and 5800X also seem to ignore that the 5600G and 5700G exist and are fine - most people would never know the difference CPU wise and you get a usable gaming GPU (by many standards) included instead of a integrated stand-in. When we are evaluating the relative value of the 12400 and the like I feel like that it will be more pertinent.

¯\_(ツ)_/¯

Maybe it's worth more in your estimation, and we can both have our own opinions

 

[Markfw]

Oh, I thought that since you had reached a conclusion that you had seen a benchmark first. My mistake.

This below quote from that link is what makes me leary of Alder Lake: (not to mention the temps)
"AMD Zen 3 is more efficient
At the upper end of the power, it becomes clear that a Ryzen 5 5950X with 16 "large" Zen 3 cores is faster with 142 watts (max. permanent power consumption ex works) than a Core i9-12900K with up to 241 watts, i.e. much more efficient. And even in the 65-watt region, things are getting tight for Alder Lake. Although Ryzen 5000 with Zen 3 cores consumes comparatively little energy even under full load, this architecture can also work much more efficiently if the clock is lowered.
In concrete terms, this means that a Ryzen 9 5950X in Eco mode with a maximum of 88 watts beats a Core i9-12900K with 88 watts by 8 percent in the editors' course, the 65-watt configuration highlighted by Intel even by 33 percent (with 35 percent higher consumption). A test of the Ryzen 9 5950X with 65 watts is still pending, but both platforms should not take much at this level – AMD's classic approach with a type core is intel's hybrid approach here at least equal and at the upper end of performance still clearly superior.
"

 

[nicalandia]

Yeah, in the Spec suite at least, the Golden Cove IPC over Zen 3 isn't much.

They are even on IPC

 

[Racan]

Yeah, in the Spec suite at least, the Golden Cove IPC over Zen 3 isn't much. GC should be boosting to 5.2 GHz vs. 5.05 GHz for the 5950x. When both are on the same DD4, it shows GC with ~1.7% and ~9% IPC lead in integer and fp respectively over Zen 3. That's a little weaker than I expected. Did anyone else try to isolate IPC or ST performance?

Here's Computerbase:
Intel Core i9-12900K, i7-12700K & i5-12600K reviewed - ComputerBase

 

[eek2121]

They are even on IPC

View attachment 52350

Not really:



AnandTech deliverin’ the goods! 😍

The E-Cores perform surprisingly well. I wonder if they can be overclocked? Seems like Intel needs to retire *cove.

 

[Abwx]

They are even on IPC

View attachment 52350

You can see the pattern from CB R15 to R23 here :

Intel Core i9-12900K, i7-12700K & i5-12600K im Test: Benchmarks in Anwendungen

Intel Alder Lake im Test: Benchmarks in Anwendungen / Ein rekordverdächtiger Testumfang / Windows 10 vs. Windows 11

www.computerbase.de

Computerbase give a small advantage for ADL in R15 perf wise wich is close to the frequency difference since their 5950X hit 5GHz occasionaly.

 

[igor_kavinski]

ADL needs expensive DDR5 to match 5950X in this workload.

 

[Zucker2k]

Puget Systems Review

 

[Ajay]

So my take... Performance on sustained workloads, good, but a lot higher power use than 5950x. Gaming, they win by a small amount. For my use case (sustained 100% load, all cores) the power usage is a killer, so I won't be getting these. For a gamer ? its a winner. But the 12600k or 12700k....

Same here. If I was buying now with gaming as my main activity - I’d buy a 12600K and be done with it. Saving dollars towards an expensive GPU would be the main driver.

That’s only one class of application I use on my system; another class is virtual machines, and core count wins. I’m not fussy on power as I'm not running any DC projects on my system anymore, so long as I don’t have to shell out for an expensive 1.2KW PSU, I’m fine 🙂.

 

[Asterox]

Blah, but why use DDR4 3200mhz CL22 for test vs DDR5 4800mhz CL40? This is completely unnecessary, 3200mhz CL16 DDR4 memory is very cheep.

There is no need to help brutally overpriced DDR5 memory.

The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity

www.anandtech.com

https://www.amazon.com/Patriot-PC4-25600-288-Pin-Desktop-PVS416G320C6K/dp/B07KXMWHBY

 

[Dayman1225]

High res pic of the 8+8 Alderlake die if anyone is interested

https://twitter.com/x/status/1456307727488065561

 

[nicalandia]

Not really:

Where on that page does it shows that they are at same clock?

 

[Hitman928]

I think due to JEDEC timings and speeds Anandtech is showing the baseline that is pretty much the best case for Zen3 vs ADL comparison. ZEN is able to buffer horrible memory latency with huge caches and somewhat faster memory controller.
While people value Anandtech results as baseline, they are only relevant for OEM systems and frankly should not even be discussed on enthusiast forums, nor those IPC values hold much relevance. Chips with 1.7% SPEC IPC advantage don't go and beat ZEN3 by tens of % all around the web.

I fully expect Z3D to have better IPC in Anandtech testing, while at same time getting soundly beaten by those monster DDR5 6000 machines. The gap once memory subsystem is unchained is just huge and shown everywhere else except than on Anandtech.

ADL has over double the L2 cache per core (0.5MB vs 1.25MB) than Zen 3 and 30 MB vs 32 MB of LLC. I don't think cache capacity is really an advantage for Zen3. Yes, DDR5 will help ADL, but that will be mostly for heavy MT scenarios or a couple of benchmarks that are more memory bound than anything, that doesn't really tell us much about the IPC or architecture performance though. When you try to get an even comparison, ADL on average is faster per clock, but just not as much as I had expected, especially given the core power use.

 

[Aikouka]

Check out the TPU link.

Thanks for that; it was pretty helpful. Amusingly enough, the one game that I was likely going to start up again soon -- Borderlands 3 -- is the one game they tested that would give me an appreciable difference. Outside of that, there just isn't much of a reason. I know a lot of folks in this thread are comparing it to AMD, but even with AMD, I'd rather wait for Zen 4 given it will introduce a new socket. I know I'm pretty used to one/two-generation motherboards with Intel, but I don't want that with AMD.

 

[Hitman928]

Same here. If I was buying now with gaming as my main activity - I’d buy a 12600K and be done with it. Saving dollars towards an expensive GPU would be the main driver.

That’s only one class of application I use on my system; another class is virtual machines, and core count wins. I’m not fussy on power as I'm not running any DC projects on my system anymore, so long as I don’t have to shell out for an expensive 1.2KW PSU, I’m fine 🙂.

Only concern for gaming is if you still play older games, particularly if they have DRM built in. It seems some games will not work right on ADL without getting patched and getting patches for older games is less and less likely the older the game is. Newer games should be just fine now that ADL is released. If the previous doesn't apply to you or Intel works out a global solution, then the 12600k looks like a fantastic choice for a gaming CPU, although it might be worth it to wait for cheaper boards to appear.

Edit: Or I guess you could just disable the E cores but then you're spending extra money just to disable a significant portion of the CPU.

 

[TheELF]

This below quote from that link is what makes me leary of Alder Lake: (not to mention the temps)
"AMD Zen 3 is more efficient
At the upper end of the power, it becomes clear that a Ryzen 5 5950X with 16 "large" Zen 3 cores is faster with 142 watts (max. permanent power consumption ex works) than a Core i9-12900K with up to 241 watts, i.e. much more efficient. And even in the 65-watt region, things are getting tight for Alder Lake. Although Ryzen 5000 with Zen 3 cores consumes comparatively little energy even under full load, this architecture can also work much more efficiently if the clock is lowered.
In concrete terms, this means that a Ryzen 9 5950X in Eco mode with a maximum of 88 watts beats a Core i9-12900K with 88 watts by 8 percent in the editors' course, the 65-watt configuration highlighted by Intel even by 33 percent (with 35 percent higher consumption). A test of the Ryzen 9 5950X with 65 watts is still pending, but both platforms should not take much at this level – AMD's classic approach with a type core is intel's hybrid approach here at least equal and at the upper end of performance still clearly superior.
"

"In concrete terms, this means that a Ryzen 9 5950X in Eco mode with a maximum of 88 watts beats a Core i9-12900K with 88 watts by 8 percent in the editors' course"
So 8% less efficient (if that's how you calculate that) for $200 less, I guess in some countries electricity is expensive enough for this to make your money back pretty fast but otherwise...

 

[Arkaign]

So my take... Performance on sustained workloads, good, but a lot higher power use than 5950x. Gaming, they win by a small amount. For my use case (sustained 100% load, all cores) the power usage is a killer, so I won't be getting these. For a gamer ? its a winner. But the 12600k or 12700k....

Edit: I think the low power usage in games is due to the E cores being used primarily, as I doubt the CPU load is high, so the scheduler selects the E cores ????

Edit2: and the temps ??? I did not see what cooler, but a lot higher than Ryzen.

I'm very much with you.

12900K is hot and super niche.

12600K and 12700K would make nice partners with high end GPUs for gaming.

Platform being so new means expensive mobos, immature OS support, and the DDR5 first wave is probably worth skipping.

I could foresee 12400F being a budget monster in 3-6 months, by that time hopefully $100ish reasonable mobos swing by.

I wonder if this is a Lynnwood moment. It would be so Intel for the next revision to be far better and live up to the potential. Eg; fall 2022, matured process tech, much faster DDR5, improved IMC, better Mobo support, matured OS/threading optimization, etc.

This is purely early adopter toy stuff IMHO for the time being.

I hope cheaper mobos come quickly, we need a price war for the $150-$250 CPU range.

And finally, I could see an Alder Lake range Core i3 being just excellent for making good gaming rigs, hopefully with some affordable GPUs as Ethereum goes PoS.

 

[Arkaign]

"In concrete terms, this means that a Ryzen 9 5950X in Eco mode with a maximum of 88 watts beats a Core i9-12900K with 88 watts by 8 percent in the editors' course"
So 8% less efficient (if that's how you calculate that) for $200 less, I guess in some countries electricity is expensive enough for this to make your money back pretty fast but otherwise...

I think the Venn diagram of people buying flagship CPUs but fussing over pennies in electric costs are basically non-existent.

Now pure heat output as a concern for someone literally running full load 24/7, that is actually a type of person that exists.

 

[JoeRambo]

Why is that weird? L3 cache lookup needs to happen during any request. Is Ian suggesting that Intel should prevent L3 lookups for all inter-core requests in the E-core cluster? That will lead to an even slower main memory access for most workloads.

That is not how things work, if that was the case, HT threads that share the core and L1/L2 would also operate at L3 latency, yet they are optimized and don't take as long.
The problem with E core cluster, is that they share L2, but probably there is some "optimization" that partitions the L2 for small cores, that not only does not allow quick cache line transfer between cores, but imparts extra 10ns latency on tranfers between same cluster cores.