Intel’s new Bartlett Lake flagship loses fight to a four-year-old CPU — Core 9 273PQE has 50% more P-cores but can't surpass Core i9-13900K in games

Bartlett lake is the same architecture and process node plus it isn’t even a “game” optimized chip. These are pre soldered boards for edge applications…

[-]

Exist50@reddit

These are pre soldered boards for edge applications…

No, these are socketed chips. It literally says LGA1700 in the article...

[-]

Sweet-Sale-7303@reddit

It uses the same socket BUT they had to modify the bios directly to get it to run.

[-]

Pillokun@reddit

he did not read nor looked up the parts.

[-]

Jumpy_Cauliflower410@reddit

This is 5600 c46 memory vs 6000 c28 memory. It's not a fair comparison.

[-]

Noreng@reddit

No, PCGH test with JEDEC supported memory. The 13900K is listed with 5600 MT/s 46-45-45

[-]

ImSpartacus811@reddit

E cores are a great idea if you have the chip design budget to develop two architectures at the game time.

That gets expensive and consumers still are confused that e cores are about saving power. So it's hard to justify keeping it going.

[-]

LAwLzaWU1A@reddit

E-Cores aren't just about power saving. It is also about getting more performance in highly threaded applications.

With the 285K, each P-core (Lion Cove) takes up almost as much die area as four E-cores (Skymont). So when designing a chip, the choice is between having one extra P-core or four extra E-cores.

According to Chip and cheese's test of SPEC2026, one Skymont core gets about 75% of the performance of a Lion Cove core. In other words, for the same die area you can get almost a 3x increase in multi-threaded performance by going with E-cores. (Please note that the 285K Chester tests with has some issues reaching the full boost frequency, but I've tried to account for a bit of that with my "75%" number).

I feel like having 4, 6 or 8 P-cores covers the "low to medium threaded" applications really well. If your workload scales so well that it can take advantage of more than 6 or 8 P-cores then chances are it scales well enough that four E-cores provides better performance than one more P-core. The E-cores are also more than powerful enough to handle a lot of background tasks so they don't have to steal CPU time from the P-cores.

I feel like people underestimate the E-cores. They are "efficiency cores", but they actually provide quite a lot of performance, especially when considering the die area they take up.

[-]

ImSpartacus811@reddit

E-Cores aren't just about power saving.

I agree. My earlier comment may have been poorly worded.

I see a lot of people mistakenly getting confused and saying that they don't need any e cores on their desktop because they don't care about saving power. Obviously e cores about maximizing MT (and to a lesser extent, ST) performance within a given die area constraint. They are not about saving power.

[-]

F9-0021@reddit

They kind of are about saving power when a similar MT performance chip with only P cores would have a much higher power draw due to the much larger die needed for that many P cores. Cost saving too, since such a big die would be incredibly expensive.

[-]

ResponsibleJudge3172@reddit

New E cores ARE more power efficient than P cores

[-]

dparks1234@reddit

They never were

[-]

mustafar0111@reddit

More e-core would not help it here.

[-]

Pillokun@reddit

I see some people are not that keen on what is happening here. 5600mt/s at cl46 on the Bartlett vs 6000 cl28 on the 13900k. that is a big perf penalty. I see big perf gaps between 6000 an 7600c34-7800c36 on my 12700k for instance.

[-]

Noreng@reddit

PCGH tests with JEDEC memory setup unless otherwise mentioned. 13900K ran at it's specified 5600 MT/s.

[-]

Sevastous-of-Caria@reddit

X3d being less ram sensitive is a gift in these trying times

[-]

faziten@reddit

Intel getting a taste of phenom / FX era for AMD.
Its amazing how history repeats itself.

[-]

Mindless_Hat_9672@reddit

what an embedded cpu vs desktop cpus the author demand... from a somewhat distorted intel lover

[-]

Limited_Distractions@reddit

I feel like there's just a very wild progression of thought to this article

Premise: People thought an LGA1700 design that uses only P-cores could be better at games due to the various considerations made around E-cores, but they were wrong because Bartlett Lake isn't strictly better

Conclusion: Well actually the reason Bartlett Lake isn't strictly better is mostly unrelated to that but we wrote this article anyway, oops

[-]

dparks1234@reddit

Generally speaking there aren’t a lot of games that can meaningfully leverage more than 8 cores/16 threads. I’ve watched enough Aliexpress X99 benchmarking videos to know that there are a few, but it isn’t the norm by any means.

My ultimate fantasy is for Intel to have an AM4 X3D moment by launching a final DDR4/DDR5 LGA 1700 CPU with BLLC.

[-]

AutoModerator@reddit

Hello gurugabrielpradipaka! Please double check that this submission is original reporting and is not an unverified rumor or repost that does not rise to the standards of /r/hardware. If this link is reporting on the work of another site/source or is an unverified rumor, please delete this submission. If this warning is in error, please report this comment and we will remove it.

I am a bot, and this action was performed automatically. Please contact the moderators of this subreddit if you have any questions or concerns.