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.

 

[Arkaign]

I'm more and more seeing this release akin to Zen1 or Nehalem. Extremely promising but some rough edges in a lot of areas.

Pros :

+Excellent single thread and multi thread capabilities
+Competitive CPU Pricing
+Introduction of new standards that will only become more influential and important such as PCIe5/DDR5/x86 Big.Little

Cons (not purely negative, but perhaps pragmatically looking at big picture and assessing the total state of things) :

-Win11 and this new more advanced core/thread allocation will take some time to get it's feet under itself. After 6-12 months I imagine it will mostly be settled, but it's not something I'd enjoy hassling with at present.
-Expensive Mobos and limited selection. B/H series won't be here for a while.
-12900K needs fairly extraordinarily cooling setup. Which is fine for diehard tuner enthusiasts, but I would not recommend this for most, get an i5/i7 and a Noctua and that's better, even if you're a gamer with a 3090.
-DDR5 is new, expensive, and will mature a LOT as time goes by. I imagine before long we'll see DDR5-6400, 7200, 8000, etc, and much improved latencies.
-PCIe 4/5 are still mostly useless for real world use outside of extremely rare use cases.
-Unluckily for Intel and basically all of us, mining craze and GPU scalping is still horrendous, somewhat deflating the launch. It's damned hard to recommend someone go build a new gaming rig when the cheapest decent GPU is overpriced by many hundreds of dollars.

I'm reading the tea leaves here a bit, and although I think I'll skip Alder, I like where it's going and think that by the time Alder refresh hits, THAT will be pretty awesome stuff with most of the negatives sorted out, cheaper/better mobos, better faster DDR5, etc etc. And at that time it's going to be a question of what Zen4 means. Zen3+ Cache actually sounds kind of stupid to me considering the yield and volume situation with TSMC. I doubt AMD wants to torpedo their margins to hell and back trying to match $200 12400, $300 12600K etc by making affordable alternatives that cost an absurd amount to produce in limited numbers (IIRC they're limited to 5900 class only anyway probably for that very reason). It smells like the Zen2+ XT series, which I looked and felt like an absolute tollbag for suggesting people wait and check out, but was just a craven excuse to jack up prices for 1% gains. Zen3 was the REAL deal, I think Zen3D will be mostly a joke and Zen4 probably amazing (and new socket, DDR5, etc, making it all the better to wait for Alder part 2 and Zen4 to make a decision if you can at all wait.

 

[Zebo]

This is Pentium 4 vs Athlon all over again. Yeah sure the highest end P4s were faster and you better have a fire extinguisher handy.

 

[clemsyn]

12600K using IGP, M.2 SSD and PCIe wifi

Average power draw from wall while doing Cinebench R23 is 175Watts (using a Bronze power supply)

R23 multiscore is 16213

 

[Bouowmx]

Since there are E cores, has anyone tried disabling Hyper-threading on P cores? To see if the change makes lightly threaded tasks faster on P cores.

 

[Zucker2k]

Igor's power consumption follow up..... Just amazing findings. Alder Lake-S is a beast in its own right!

Core i9-12900KF, Core i7-12700K and Core i5-12600 in a workstation test with amazing results and an old weakness | Part 2 | igor´sLAB

So today I’ll get serious and show you where Alder Lake S can really score aside from colorful gaming pixels. Gaming what? Completely overrated if you look at at least some of today’s results.

www.igorslab.de

 

[AtenRa]

Igors power consumption follow up.....

Core i9-12900KF, Core i7-12700K and Core i5-12600 in a workstation test with amazing results and an old weakness | Part 2 | igor´sLAB

So today I’ll get serious and show you where Alder Lake S can really score aside from colorful gaming pixels. Gaming what? Completely overrated if you look at at least some of today’s results.

www.igorslab.de

I really trying to understand how igor defines efficiency.

 

[Zucker2k]

I really trying to understand how igor defines efficiency.

I don't think you're questioning his competency, are you? Hehe

Meanwhile, the first undervolt results...

https://www.reddit.com/r/intel/comments/qnkgcb

 

[Zucker2k]

Gaming summary from 19 reviews...

Alder Lake Launchreviews: Die Spiele-Performance im Überblick | 3DCenter.org

Als kleinen Teaser zur kommenden Launch-Analyse zu Intels Alder Lake sollen hiermit bereits die Spiele-Benchmarks zur neuen Prozessoren-Generation ausgebreitet werden – und damit sicherlich jenes Benchmark-Feld, welches am meisten interessiert. Zum

www.3dcenter.org

 

[SKORPI0]

Priced a potential build. I have unused Thermaltake Core V71 Full-Tower Chassis , EVGA SuperNOVA 1000 G+ 80 Plus Gold 1000W, Corsiar H115i RGB Platinum 280mm (ordered the LGA 1700 standoffs) and a current using eVGA 1080Ti FTW3 ($200 at flea market- Aug 2019).

i9-12900K ----------------------- $649,99
Z690 Pro ------------------------- $329.99
64GB DDR5 4800---------------$547.99
-----------------------------------------------------------
Total ------------------------------ $1527.97 (shipping +taxes not included)

Of course a nice RTX 3080 at MSRP would be nice to get. I might end up waiting for the RTX 40 series (late 2022, early 2023).
Or wait until prices for the Alder-Lake goes or it's successor., 13th Gen. Raptor Lake

 

[AtenRa]

I don't think you're questioning his competency, are you? Hehe

Meanwhile, the first undervolt results...

https://www.reddit.com/r/intel/comments/qnkgcb

Well two things,

When measuring AlderLake 12900K at manual PL1 125W (base), then you should also measure Ryzen 5950X at 105W (Base) and then compare the efficiency.

Also to point out that efficiency should always be put in a context of either iso performance or iso power for CPUs at the same segment (12900K vs 5950X etc).

 

[epsilon84]

I don't think you're questioning his competency, are you? Hehe

Meanwhile, the first undervolt results...

https://www.reddit.com/r/intel/comments/qnkgcb

As I suspected. The 12900K is basically pre-overclocked out of the box to give Intel an overall 'win' against the 5950X (better single-thread, competitive multi-thread). It is pushed beyond its efficiency curve for the sake of winning benchmarks. Think of that what you will, personally I would undervolt a 12900K after seeing the insane temps on high end AIOs.

Of course, undervolting is nothing new. GPUs benefit greatly from this as well. I personally undervolt my 3070 Ti from 1.0V to 0.85V and power draw goes from 300W to 210W, for about a 2-3% loss in performance. It's funny how I've come full circle, I used to overclock every CPU/GPU that I got, nowadays it seems the first thing I'll do is undo the 'factory overclocks' instead and undervolt for better efficiency!

 

[mikk]

Interesting power consumption test from Raichu at the same clock speed and voltage:

https://twitter.com/x/status/1456939487003480074

1.5W more usage compared to Willow Cove. As we know Golden Cove performs about 20% better at the same clock. Intel didn't promise too much when they claimed 10-15% better performance per watt over 10SF.

 

[Carfax83]

As I suspected. The 12900K is basically pre-overclocked out of the box to give Intel an overall 'win' against the 5950X (better single-thread, competitive multi-thread). It is pushed beyond its efficiency curve for the sake of winning benchmarks.

Did the stock 5.2ghz single core and 5ghz all core "boost" give it away?

 

[TheELF]

As I suspected. The 12900K is basically pre-overclocked out of the box to give Intel an overall 'win' against the 5950X (better single-thread, competitive multi-thread). It is pushed beyond its efficiency curve for the sake of winning benchmarks. Think of that what you will, personally I would undervolt a 12900K after seeing the insane temps on high end AIOs.

It's not intel's fault, they make a CPU they tell people how to use it and everybody is ignoring them.
If intel would enforce reviewers to use certain settings there would be an even higher outrage.

The issue is with mobo makers pushing everything to the limit and with most reviewers being even more clueless than the average user.
Here is an example of everything that is wrong with reviews.
jayz runs the 12900k with MCE on at the beginning being completely honestly sure that overclocking is what intel considers PL lifted resulting in ~270W which is above PL2 because MCE applies an overclock on top of the lifted limit, then he goes into bios and disables overclocking/MCE honestly thinking that that is equal to using PBP/TDP because it says enforce all limits BUT ALL LIMITS ARE SET TO AUTO which means they are set to the highest possible or otherwise power lifted which is why he then gets a bit less power which should be the 241W that is max turbo power of constant PL2.

 

[JoeRambo]

Interesting power consumption test from Raichu at the same clock speed and voltage:

How the hell that 1.012V voltage point was even chosen? Is it driven by 14nm+++ that requires that voltage for 4.2Ghz ? Makes no sense to fix voltage as well, maybe ADL requires 0.95V and chews 60W instead ?

 

[DrMrLordX]

It's not intel's fault, they make a CPU they tell people how to use it and everybody is ignoring them.

That's been an issue since at least Coffee Lake, and it keeps getting worse. If Intel would/could say "hey, our next-gen CPU will use less power but perform better" then the problem would be gone overnight. But from Coffee -> Comet -> Rocket -> Alder, power usage has steadily gotten worse, or at least hasn't gotten any better.

 

[igor_kavinski]

That's been an issue since at least Coffee Lake, and it keeps getting worse. If Intel would/could say "hey, our next-gen CPU will use less power but perform better" then the problem would be gone overnight. But from Coffee -> Comet -> Rocket -> Alder, power usage has steadily gotten worse, or at least hasn't gotten any better.

Unfortunately, it seems it will remain that way until RibbonFETs and PowerVia CPUs arrive in 2024, assuming Intel remains on schedule.

 

[DrMrLordX]

Unfortunately, it seems it will remain that way until RibbonFETs and PowerVia CPUs arrive in 2024, assuming Intel remains on schedule.

They apparently have some help from IBM this time around. Their 20a process might actually be good. That probably won't hit until 2025 though.

 

[majord]

How the hell that 1.012V voltage point was even chosen? Is it driven by 14nm+++ that requires that voltage for 4.2Ghz ? Makes no sense to fix voltage as well, maybe ADL requires 0.95V and chews 60W instead ?

or maybe it needs more.. you don't know (architectures are different) You just can't go fixing Vcore's and making such comparisons. The only valid way is using the official VID for that frequency on that part. If you want to represent a decent sample size and have $ to burn - do it on a number of CPU's.

 

[Abwx]

voltage is not shrinkable with a new process, that s inherent to laws of physics...

 

[Hulk]

In the beginning CPU's were all unlocked and we could adjust various parameters to squeeze out some extra MHz. Intel was not keen with us turning $200 parts into $500 ones and started locking the CPU's. They needed to protect the pricing structure of their stack. We all complained and it went something like "we bought the part, let is fool around with it."

Now Intel has a competitor and they are doing what was the hard overclocking work. That being figuring out what max frequency the CPU can safely operate (under warranty) and at what required voltage. In fact, you can even set power limits, which limit how far you want to push the part. Sounds great right? Intel had pre-tested their CPU's to within an inch of their lives and by simply setting unlimited power and providing adequate cooling you can eek every bit of performance out of your CPU. Or you can forego that last 10% or so of performance and cap power use at 125W, or 150W, or whatever.

But the complaints are still there. "Intel uses too much power." "Intel is killing the environment." "Intel should be fined for releasing power hog CPU's."

And that my friends is why we generally aren't allowed "full access" to so many devices in our daily lives. Not only to protect ourselves, but also the manufacturer from getting bad press because people operate the device in zones that don't always make the product look good.

With great power comes great responsibility!
(pun intended)

 

[eek2121]

Well two things,

When measuring AlderLake 12900K at manual PL1 125W (base), then you should also measure Ryzen 5950X at 105W (Base) and then compare the efficiency.

Also to point out that efficiency should always be put in a context of either iso performance or iso power for CPUs at the same segment (12900K vs 5950X etc).

That isn’t how you measure efficiency. CPU1 may use 2X more power than CPU2, but if CPU1 does the same work 3 times faster, CPU1 is more efficient than CPU2.

 

[Heartbreaker]

HWUB - 12600K Review.
My favorite of the bunch (because I'm poor). I'd wait for B660 MB, and use DDR4 to keep costs down:

Matches 5800X performance and power usage.

 

[AtenRa]

That isn’t how you measure efficiency. CPU1 may use 2X more power than CPU2, but if CPU1 does the same work 3 times faster, CPU1 is more efficient than CPU2.

Depends on the context, in general we use the perf/watt but if you want to see the efficiency at a specific performance or the performance you can get at iso power then just perf/watt desnt mean much.

In his case, he lowered the TDP for one CPU (PL1 125W) and left the other CPU at a higher TDP of 141W. Why not lower both CPU TDPs and measure the efficiency again ???

 

[gdansk]

HWUB - 12600K Review.
My favorite of the bunch (because I'm poor). I'd wait for B660 MB, and use DDR4 to keep costs down:

Matches 5800X performance and power usage.

Totally agree, it's the most compelling of the products presently available. Intel needs to make sure B660 are priced competitively to B550.