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.

 

[jpiniero]

Okay thanks. So 8+8+1 is for 8+8, 8+4, and 6+4. I understand that. It makes sense.

And 6+0 and 4+0 and 2+0... and there's also the chance they could do other combinations with Alder Lake-S BGA. Probably 8+0 for Xeon E too.

I guess another reason they chose 6+0 over 4+8 is that you can cut 8+8 dies that have both busted small core clusters. If they used 4+8 they would only be able to use dies where the small cores didn't get busted.

 

[Hulk]

So you've gotta love this pricing battle going. From my local Microcenter.
I grouped by approximate performance levels.
In my opinion 12600K at the low end. 12700K in the middle. And it's a toss up at the top depending on your workflow.

Does MC have price protection? My 12700K is $30 less than when I bought it a 10 days ago!

12600K - $270
5600X - $280
5800X - $300

12700K - $370
5900X - $470

12900K - $650
5950X - $700

 

[TESKATLIPOKA]

Hey guys, check out Perschistence's tweets here, here and here.
He posted graphs where he compared i9 12900K in Cyberpunk 2077 and GTA V at different PL settings.
He also did test It in a few applications like Cinebench R20.

 

[Hulk]

Hey guys, check out Perschistence's tweets here, here and here.
He posted graphs where he compared i9 12900K in Cyberpunk 2077 and GTA V at different PL settings.
He also did test It in a few applications like Cinebench R20.

Great info!
So it looks like 170W is about as high as you really need to push the 12900K. After that it's a lot of power for less than at most 7% increase in performance.
Also 65W is really enough in the games he tested!
Looks like CPU power draw in games is essentially a non-issue.

 

[blckgrffn]

So you've gotta love this pricing battle going. From my local Microcenter.
I grouped by approximate performance levels.
In my opinion 12600K at the low end. 12700K in the middle. And it's a toss up at the top depending on your workflow.

Does MC have price protection? My 12700K is $30 less than when I bought it a 10 days ago!

12600K - $270
5600X - $280
5800X - $300

12700K - $370
5900X - $470

12900K - $650
5950X - $700

They definitely do, I've had good luck with just using their chat support for low effort refunds.

 

[Hulk]

They definitely do, I've had good luck with just using their chat support for low effort refunds.

Thanks for the advice. I'm chatting with them now and they are telling me they can't process VISA refunds remotely and that I'll have to go do the store. Did you have this issue?

 

[DAPUNISHER]

I expect the prices available now will be held onto

For the 5600x, I disagree. We will know within days or weeks of when the 12400 launches. I have no problem admitting when I am wrong. I will accept my trout smacking with dignity. BTW, according to Newegg, 5600X is still the number 1 seller in CPUs. Amazon has it #4. All top 5 sellers are Ryzen on Amazon, top 4 on Newegg. Don't ask me why, I just meme here.

Newegg shoppers are choosing 12700K as top ADL. Amazon shoppers picked 12900K. I don't know if that is due to availability from each?

 

[blckgrffn]

Thanks for the advice. I'm chatting with them now and they are telling me they can't process VISA refunds remotely and that I'll have to go do the store. Did you have this issue?

Never. I've always used Visa and had zero issues.

Now, things might change... I think I last did this maybe a year ago?

 

[Hulk]

Never. I've always used Visa and had zero issues.

Now, things might change... I think I last did this maybe a year ago?

They are telling me it's a new policy but if the store manager agrees they'll send me a store credit in the mail.
Or I can just drive out there. It's always fun to look around. Anyone hear need me to pick up a CPU for them?!

 

[Hulk]

This is from the Anandtech CPU Buying Guide and the chart title is listed as "Gen-on-gen IPC."
What can account for the ever increasing Skylake core IPC? It's all DDR 4 and as far as I know there were no significant core difference that could account for a 24% IPC increase from Skylake to Comet Lake?
Perhaps this is "throughput" with frequency taken into account?

If you take the following cores and max ST frequency and then start the "scale" at 94% you get the following.

6700K	4.2	94%
7700K	4.5	101%
8700K	4.7	106%
9900K	5	113%
10900K	5.3	120%

 

[maddie]

This is from the Anandtech CPU Buying Guide and the chart title is listed as "Gen-on-gen IPC."
What can account for the ever increasing Skylake core IPC? It's all DDR 4 and as far as I know there were no significant core difference that could account for a 24% IPC increase from Skylake to Comet Lake?
Perhaps this is "throughput" with frequency taken into account?

If you take the following cores and max ST frequency and then start the "scale" at 94% you get the following.

6700K	4.2	94%
7700K	4.5	101%
8700K	4.7	106%
9900K	5	113%
10900K	5.3	120%

Where do you see gen-on-gen IPC?

 

[Hulk]

Where do you see gen-on-gen IPC?

That is the name of the file when you download it from the review.

 

[semiman]

This is from the Anandtech CPU Buying Guide and the chart title is listed as "Gen-on-gen IPC."
What can account for the ever increasing Skylake core IPC? It's all DDR 4 and as far as I know there were no significant core difference that could account for a 24% IPC increase from Skylake to Comet Lake?
Perhaps this is "throughput" with frequency taken into account?

If you take the following cores and max ST frequency and then start the "scale" at 94% you get the following.

6700K	4.2	94%
7700K	4.5	101%
8700K	4.7	106%
9900K	5	113%
10900K	5.3	120%

It's an ST performance chart. Coffee lake and Coffee lake refresh comparison makes no sense if it's an IPC chart. Their IPC cannot be different.

 

[ondma]

Yea, it it "core performance", the difference between Skylake and Coffee Lake is clockspeed.

 

[Hitman928]

It's an ST performance chart. Coffee lake and Coffee lake refresh comparison makes no sense if it's an IPC chart. Their IPC cannot be different.

It's not performance for AMD though, at least relative to each other, it is definitely IPC for AMD CPUs. I'm thinking they messed something up when compiling the chart.

 

[DrMrLordX]

ST performance may also be observed to increase on later-gen Skylake products due to the amount of L3 cache available.

 

[JoeRambo]

ST performance may also be observed to increase on later-gen Skylake products due to the amount of L3 cache available.

Yeah, and also memory speeds increased from 2133 to 2933.

There is also an elephant in the room in the form of hardware security support for Spectre and Meltdown and there is massive jump in Skylake performance @Coffee Lake Refresh aka 9900K gen.

The Intel 9th Gen Review: Core i9-9900K, Core i7-9700K and Core i5-9600K Tested

www.anandtech.com

No wonder, when it contains hw fixes, and previuos CPUs don't.

 

[coercitiv]

ST performance may also be observed to increase on later-gen Skylake products due to the amount of L3 cache available.

Still, arguing the chart is listed as "Gen-on-gen IPC" when the only indicator is the filename while neither the chart title nor any other wording in the article contain the word "IPC" is quite the strech.

 

[Zucker2k]

Still, arguing the chart is listed as "Gen-on-gen IPC" when the only indicator is the filename while neither the chart title nor any other wording in the article contain the word "IPC" is quite the strech.

Obviously, the only indicator is not the file name. What, in your opinion, accounts for the differences in Skylake generations, and how would you title the test? It seems the tester thinks "gen-on-gen ipc" is a fitting name? Also, a poster thinks the same characterization fits the AMD chart, why not Intel? Etc.

 

[coercitiv]

how would you title the test? It seems the tester thinks "gen-on-gen ipc" is a fitting name? Also, a poster thinks the same characterization fits the AMD chart, why not Intel? Etc.

The title of the test is "Core Performance (SPEC06/17 average)". It's right there in the chart, with big, bold letters. The context of the chart is also important, look at what the author writes in conjunction with the graph:

Most modern games can easily chew through four cores, and take advantage of six. When we’re getting up to that level, it also matters about single core performance too, and so trying to build in some headroom with what you can buy today obviously matters. But when buying, you also have to think about what’s coming up, and if you’re planning to upgrade or completely change systems. Something future proof has to work today, tomorrow, but also give options when tomorrow comes.

The point I'm trying to make is that in such a case one doesn't simply assume the filename is the only relevant piece of information, while the chart title and the writing about "single core performance" in the context of gaming is not. I care less what some other poster thinks of the chart relevance to AMD IPC, I care more about what the author of the chart measured. Why initiate a sterile debate when we don't even have a proper clue about what and how is being measured?

Even IPC can be measured with different methodologies, and IIRC Anandtech prefers running the chips at stock and derive "IPC" through normalization. Other reviewers prefer testing directly at ISO frequency. Both methodologies have their shortcomings, but we want to at least know which one the reviewer is using in the first place in order to give meaning to the results.

What, in your opinion, accounts for the differences in Skylake generations

Considering we're talking Anandtech review methodology, there's 3 variables between Skylake generations:

• L3 cache size increase in size from 8M to 20M
• uncore speed increase
• DDR4 support increase from 2133MT/s to 2933MT/s

Newer Skylake cores are fed with data faster, both in terms of bandwidth and latency, from every memory tier.

 

[coercitiv]

Took a look at Anandtech's benchmark charts which contain SPEC2006 1T and SPEC2007 1T measurements:

• The difference in performance from 6700K to 7700K is ~9% in SPEC2006 and ~9% in SPEC 2017. If we take the 94% rating of Skylake and multiply that by x1.09, we obtain ~102.46% which coincides with the 102% rating Kaby Lake received in the debated chart.
• The difference in performance from 6700K to 10900K is ~27.7% in SPEC2006 and ~24.2% in SPEC2017, with an average of ~26%. If we take the 94% rating of Skylake and multiply that by x1.26, we obtain ~118.44% which coincides with the 118% rating Comet Lake received in the debated chart.

Conclusion? It's a performance chart. If there's anything wrong with the AMD data we should ask the author.

 

[Zucker2k]

Took a look at Anandtech's benchmark charts which contain SPEC2006 1T and SPEC2007 1T measurements:

• The difference in performance from 6700K to 7700K is ~9% in SPEC2006 and ~9% in SPEC 2017. If we take the 94% rating of Skylake and multiply that by x1.09, we obtain ~102.46% which coincides with the 102% rating Kaby Lake received in the debated chart.
• The difference in performance from 6700K to 10900K is ~27.7% in SPEC2006 and ~24.2% in SPEC2017, with an average of ~26%. If we take the 94% rating of Skylake and multiply that by x1.26, we obtain ~118.44% which coincides with the 118% rating Comet Lake received in the debated chart.

Conclusion? It's a performance chart. If there's anything wrong with the AMD data we should ask the author.

Just wow! Thanks for going through the trouble.

 

[JoeRambo]

Some interesting results in that bench. For example
(9-1d) SPEC2006 1T - 429.mcf

Intel Core i9-12900K + Win10 + DDR5 : 70
Intel Core i9-10900K (10C/20T, 125W, $488) : 69.7
Intel Core i9-11900K (8C/16T, 125W, $539) : 73.1
and
Intel Core i9-12900K + Win10 + DDR4 : 53.6

so either a typo, or test ran on E cores. In most other SPEC17 1T results DDR4 12900K is beating the hell out of others and some tests like https://www.anandtech.com/bench/CPU-2020/2798 seem to be utterly broken.

 

[igor_kavinski]

Some interesting results in that bench. For example
(9-1d) SPEC2006 1T - 429.mcf

Intel Core i9-12900K + Win10 + DDR5 : 70
Intel Core i9-12900K + Win10 + DDR4 : 53.6

429.mcf: SPEC CPU2006 Benchmark Description

429.mcf requires about 860 and 1700 megabyte for a 32 and a 64 bit data model, respectively.

A test using that much memory would benefit from DDR5's increased bandwidth, no?

 

[JoeRambo]

A test using that much memory would benefit from DDR5's increased bandwidth, no?

If that was the case, 10th and 11th gen couples with DDR4 and 33%/50% less L3 cache would be left in dust as well. More like a case of results that make no sense.