The Intel 12th Gen Core i9-12900K Review: Hybrid Performance Brings Hybrid Complexity
by Dr. Ian Cutress & Andrei Frumusanu on November 4, 2021 9:00 AM ESTCPU Tests: SPEC MT Performance - P and E-Core Scaling
Update Nov 6th:
We’ve finished our MT breakdown for the platform, investigating the various combination of cores and memory configurations for Alder Lake and the i9-12900K. We're posting the detailed scores for the DDR5 results, following up the aggregate results for DDR4 as well.
The results here solely cover the i9-12900K and various combinations of MT performance, such as 8 E-cores, 8 P-cores with 1T as well as 2T, and the full 24T 8P2T+8E scenario. The results here were done on Linux due to easier way to set affinities to the various cores, and they’re not completely comparable to the WSL results on the previous page, however should be within small margins of error for most tests.
In the integer suite, the E-cores are quite powerful, reaching scores of around 50% of the 8P2T results, or more.
Many of the more core-bound workloads appear to very much enjoy just having more cores added to the suite, and these are also the workloads that have the largest gains in terms of gaining performance when we add 8 E-cores on top of the 8P2T results.
Workloads that are more cache-heavy, or rely on memory bandwidth, both shared resources on the chip, don’t scale too well at the top-end of things when adding the 8 E-cores. Most surprising to me was the 502.gcc_r result which barely saw any improvement with the added 8 E-cores.
More memory-bound workloads such as 520.omnetpp or 505.mcf are not surprising to see them not scale with the added E-cores – mcf even seeing a performance regression as the added cores mean more memory contention on the L3 and memory controllers.
In the FP suite, the E-cores more clearly showcase a lower % of performance relative to the P-cores, and this makes sense given their design. Only few more compute-bound tests, such as 508.namd, 511.povray, or 538.imagick see larger contributions of the E-cores when they’re added in on top of the P-cores.
The FP suite also has a lot more memory-hungry workload. When it comes to DRAM bandwidth, having either E-cores or P-cores doesn’t matter much for the workload, as it’s the memory which is bottlenecked. Here, the E-cores are able to achieve extremely large performance figures compared to the P-cores. 503.bwaves and 519.lbm for example are pure DRAM bandwidth limited, and using the E-cores in MT scenarios allows for similar performance to the P-cores, however at only 35-40W package power, versus 110-125W for the P-cores result set.
Some of these workloads also see regressions in performance when adding in more cores or threads, as it just means more memory traffic contention on the chip, such as seen in the 8P2T+8E, 8P2T regressions over the 8P1T results.
What’s most interesting here is the scaling of performance and the attribution between the P-cores and the E-cores. Focusing on the DDR5 set, the 8 E-cores are able to provide around 52-55% of the performance of 8 P-cores without SMT, and 47-51% of the P-cores with SMT. At first glance this could be argued that the 8P+8E setup can be somewhat similar to a 12P setup in MT performance, however the combined performance of both clusters only raises the MT scores by respectively 25% in the integer suite, and 5% in the FP suite, as we are hitting near package power limits with just 8P2T, and there’s diminishing returns on performance given the shared L3. What the E-cores do seem to allow the system is to allows to reduce every-day average power usage and increase the efficiency of the socket, as less P-cores need to be active at any one time.
Facebook
Twitter
RSS
474 Comments
View All Comments
mode_13h - Saturday, November 6, 2021 - link
On what basis do you reach that verdict? Based on your posts, I wouldn't trust you to run a corner shop, much less one of the biggest and most advanced tech & manufacturing companies on the planet.And where did they said it's "fused off"? AFAIK, all they said is that it's not available and this will not change. And we've seen no evidence of that being untrue.
Also, I think you're getting a bit too worked up over the messaging. In the grand scheme, that's not the decision that really matters.
SystemsBuilder - Saturday, November 6, 2021 - link
no I did no miss that. I'm just happy that ASUS found a way to enable it.Intel screwed up of course - battel between different departments and managers, marketing etc I'm sure - that's a given and did not think it was necessary to repeat that. And yes it is absurd - even incompetent.
Still I'm happy ASUS found it and exposed it, because, as I said they they actually seam to have gotten AVX-512 right in Golden cove.
Intel should of course work with Microsoft to get the scheduler to work with any E/P mix, make the support official, enable it in the base BIOS, have base BIOS sent over to all OEMs and lastly fire/reassign the idiot that took AVX-512 off the POR for Alder lake.
In any case it give me something to look forward to with Sapphire rapids which should come with more Golden cove P cores.
I only by ASUS boards so
SystemsBuilder - Saturday, November 6, 2021 - link
can't edit post so continuing:I only buy and use ASUS boards so for me it's fine but I sucks for others.
Also doubt that Pat was involved. Decisions were likely made before his arrival. I'm thinking about the Microsoft dependency. They would have needed to lock the POR towards Microsoft a while back to give MS enough time to get the scheduler and other stuff right...
Oxford Guy - Saturday, November 6, 2021 - link
The product was released on his watch, on these incompetent terms. Gelsinger is absolutely responsible. He now has a serious black mark on his leadership card. A competent CEO wouldn’t have allowed this situation to occur.This is an outstanding example of how the claim that only engineers make good CEOs for tech companies is suspect.
‘I only by ASUS boards so’
Lies of omission are lies.
SystemsBuilder - Saturday, November 6, 2021 - link
I totally agree that it goes against putting engineers in charge.for me the whole AVX-512 POR decision and "AVX-512 is fused off" message is coming out of a incompetent marketing department when they were still in charge.
Oxford Guy - Saturday, November 6, 2021 - link
‘when they were still in charge.’Gelsinger isn’t the CEO? Gelsinger wasn’t the CEO when Alder Lake was released? Marketeers outrank the CEO?
The buck stops with him.
The implication that having an engineer run a business makes said engineer a skilled businessperson is safely dead.
The success of Steve Jobs also problematized that claim well before this episode, as he was not an engineer.
SystemsBuilder - Saturday, November 6, 2021 - link
i think under the old "regime" marketing did out rank engineering so that the old CEO listened more to marketing than engineering (of course the CEO makes the decision but he takes input from various camps and that is what i mean with marketing "were still in charge"). A non-engineering educated CEO is particularly influenceable by marketing (especially if he/she has MBA with marketing specialization like the old CEO's). Hence the messaging decisions to "fuse it off" was likely heavily influenced by marketing who i think finally won over Engering. Pat had to inherit this decision but could not change i for windows 11 launch - it was too late.Oxford Guy - Saturday, November 6, 2021 - link
Of course he could change it. He’s the CEO.mode_13h - Saturday, November 6, 2021 - link
> so that the old CEO listened more to marketing than engineeringIn this case, the issue wouldn't be who the CEO listens to, but who gets to define the products. Again, the issue of AVX-512 in Alder Lake is something that would probably never rise to the attention of the CEO, in a company with $75B annual revenue, tens of thousands of employees at hundreds of sites, and many thousands of products in hundreds of different markets. OG apparently has no concept of what these CEOs deal with, on a day to day basis.
mode_13h - Saturday, November 6, 2021 - link
> Gelsinger wasn’t the CEO when Alder Lake was released?So what was he supposed to do? Do you think they run all PR material by the CEO? How would he have any time to make the important decisions, like about running the company and stuff?
It seems to me like you're just playing agent provocateur. I haven't even seen you make a good case for why this matters so much.