Intel Core Ultra 7 265K

[W1zzard]

The Intel Core Ultra 7 265K offers eight strong Lion Cove cores, just like the 285K, and merely four E-Cores fewer, for an almost 50% price difference. Despite the "Ultra 7" branding it offers plenty of application performance that can beat even the Ryzen 9 9900X. Things don't look so good in gaming though.

Show full review

 

[wheresmycar]

I was expecting better gaming performance, at least enough to outperform Raptor and Zen 4/5 non-X3D models. Credit where its due: AL makes up for it with its impressive low thermals and power consumption compared to previous generations.

The $400 price tag is more justifiable for multi-core workloads.

 

[chrcoluk]

Thank you for the review, very interesting.

My take is as expected HT is no loss on multi threaded productivity. E-cores have taken over that well enough.
Power efficiency has gone in a positive direction.
However there is some performance regressions compared to raptor lake, seemingly on emulators and gaming where its most visible. Curious if this needs scheduling improvements or if will just be a long term problem with the chip, which is something I thought might be an issue for these chips as its a shift of thread design.

Hopefully one day W1zzard or owners can do some playing with CPU affinity to see if performance is recoverable.

Given where the improvements are and my take on the Z890 motherboards, I dont see my self buying one of these, although I remain interested if Bartlett can pull out a 12 p-core chip for the Z690/Z790.

 

[_roman_]

I'm impressed by the impressive performance of this processor in comparison with my low value, half the costs, entry Ryzen 7600X in 1440p. It seems the graphic card matters the most

I'm interested in /proc/cpuinfo for the good and the worse cores.

I think pcgameshardware wrote that the processor has native Thunderbolt interface and therefore no need for an external usb 4 chip. This is a bonus when you need external devices.

 

[wheresmycar]

However there is some performance regressions compared to raptor lake, seemingly on emulators and gaming where its most visible. Curious if this needs scheduling improvements or if will just be a long term problem with the chip, which is something I thought might be an issue for these chips as its a shift of thread design.

In the value and conclusion section, one of the CONS being:

• Some games and applications aren't currently performing well at all

Perhaps additional optimization is needed for the scheduler and thread director, along with improvements in software or game hybrid threading. It could also be the added latency from moving the memory controller to a separate tile. I wonder how much impact the removal of HT has on certain games and applications.

 

[pavle]

Again - "7 265K" L2 cache performance is piss-poor compared to previous gen (i7 14700K) and the competition, and for a 3nm chip it sure is power-hungry.

 

[chrcoluk]

In the value and conclusion section, one of the CONS being:



Perhaps additional optimization is needed for the scheduler and thread director, along with improvements in software or game hybrid threading. It could also be the added latency from moving the memory controller to a separate tile. I wonder how much impact the removal of HT has on certain games and applications.

Good point about the memory controller.

 

[phints]

I'm... really not sure what to make of this CPU lol.

 

[Onyx Turbine]

I mentioned before this chip is a stepping stone towards a more efficient chip series. Thank you for this great review and updated performance tables, im now officially in love with my 12700K.

 

[Julhes]

 

[Onyx Turbine]

View attachment 368760

W

View attachment 368760

The delta differences presented here in your attachments are enormous in my opinion not equal , at least not correspond to real performance with TPs published article here.
What interests me most about this mid range cpu and the presented data from TP (Assume its correct) is following part which shows no significant improvement in 1440P Performance just 2,7% better than a 12700K. Also quickly viewed the power consumption which is not that impressively reduced.
Of course the chip itself would be for a new built with such a budget be a solid option, but the lga 1700 socket will compete against it also price wise

 

[Lew Zealand]

It's a Ryzen 9 9900X with slightly better pricing, power use, and memory support. I guess that's meh+ as the Ryzen 9000 series was meh?

 

[shk021051]

the leap is so small for rayzen and intel both are skippable this gen

 

[bgx]

Memory latency seems to be the big victim of the new architecture.

I just checked: ~82ns for arrowlake vs ~66ns for Zen 5 and raptor lake.
That explains why games have disapointing results, despite the Pcore with highest IPC of all (which should theoretically give very good game perf).

Zen1 had that problem. AMD was able to work on that. We can hope intel can also work on that issue with their disagregated architecture. 3D LLC could also be a stopgap solution.

Wizard, you could make the comparison in the AIDA64 cache and memory perf with other arch. Also, in the conclusion if you believe this is a big driver for what we are seeing. Not that it gives a fix for arrowlake, but an idea what Intel must work on for next iteration, and it explains things quite reasonably well!

 

[Metroid]

Intel = step in the right direction, now I'm waiting for Intel 3dvcache and then we will have a better gaming competition x AMD. I like the idea that SMT/HT is dead for Intel desktop cpus, that was hacker's paradise, especially now they know how to take control using those virtual threads, more security for users = step in the right direction.

 

[N/A]

Memory latency seems to be the big victim of the new architecture.

The memory controller can be moved back to the CPU tile with a refresh, which is now cancelled. At least one SKU, the 225F originally intended to be monolithic, but 20A was canceled.

We can only hope that Nova lake will get back on track by being sensible.

 

[sLowEnd]

Credit where its due: AL makes up for it with its impressive low thermals and power consumption compared to previous generations.

Raptor Lake didn't exactly set a high bar to pass. Power consumption is still at about Alder Lake levels

 

[trsttte]

CUDIMM? They could at least standardize on ECC CUDIMM bringing ECC to the masses once and for all, or CAMM, or ECC CAMM. Will AMD support this? How many different types of memory are there now concurrently?

Perhaps additional optimization is needed for the scheduler and thread director, along with improvements in software or game hybrid threading.

How long will this charade of hybrid bigLittle architecture on the Desktop go until they realise it's unworkable? Windows is a mess and the decades of old software still in use are a mess, new software is not really doing anything to solve the problem and the Thread Director failed at an impossible task. AMD was smarter, they tested the waters with Zen 4 + 4c on some product lines and just built on top of something that already kind of worked with the preferred core system. Intel went all in with an heterogeneous architecture and it's simply not working out as they hoped and it's looking like no ammount of duct tape and marketing words will solve the base issue that is Windows is not ready for this.

I'm curious to see Linux benchmarks, probably better on somethings but just as much of a clusterfuck as windows on others.

 

[wheresmycar]

Raptor Lake didn't exactly set a high bar to pass. Power consumption is still at about Alder Lake levels

Can't deny drawing AL level juice and only ~4% perf advantage over a 3 year old chip doesn't look pretty. But its nice to see power/efficiency improvements over RL/RL-ref which was concerning (not to mention all the other problems)

How long will this charade of hybrid bigLittle architecture on the Desktop go until they realise it's unworkable?

As long as it takes! If Big Techs 'data-first' obsession is pushing P-e on us, the not-so interested mainstream consumer is feeling the burn anyway. But I agree, many older applications weren’t designed with heterogeneous computing in mind, which definitely doesn’t help. All we can do is hope that the OS and game/software devs will eventually make full use of this inevitable shift to our advantage. I also feel Intels mainstream heterogeneous nose-dive might have been either poorly timed or perhaps a necessary leap to kick-start the inevitable future. 'BIG changes in little increments' or just go "BIG.little", time will tell!

 

[bgx]

on power consumption: It seems to me that they pushed arrowlake to the limit on power consumption to extract every bit of perf they can, as it has deficit in clock vs Raptor lake and high latency too (so perf in game is not good).

So yes, efficiency @factory value is not impressive, but it does not mean the efficiency of the architecture is bad. 10nm vs 3(even fake TSMC) nm should still be very obvious.

A test @ 4 or 5Ghz and undervolted while being stable (for different arch, Zen4, 5, ARL, RL, AL) would be interesting to see the real efficiency of the underlying arch/process.

On a side note, on performance, Intel 7 is not that bad!

We can only hope intel 18A is viable next year.

 

[danc]

It's a Ryzen 9 9900X with slightly better pricing, power use, and memory support. I guess that's meh+ as the Ryzen 9000 series was meh?

It is a better 9900X since you are getting full fat 8 P-cores before crossing into CCD/Tiles latency

 

[TheDeeGee]

A very nice step in the right direction.

Despite no performance increase the power consumption and thermals are a winner.

 

[Pepamami]

A very nice step in the right direction.

Despite no performance increase the power consumption and thermals are a winner.

The worst start of the new platform.

Feels like I am looking at Bulldozer release.

Next step gonna be overclocking these 285k, and 385k will eat ton of power again, to keep up with competition.

 

[TheDeeGee]

The worst start of the new platform.

Feels like I am looking at Bulldozer release.

Next step gonna be overclocking these 285k, and 385k will eat ton of power again, to keep up with competition.

285K is almost twice as fast as my 11700, seems like a winner to me.

 

[Pepamami]

285K is almost twice as fast as my 11700, seems like a winner to me.

285K is almost 100 times as fast as my Pentium 4 2.0Ghz, some times its 1000 times, and so?

Right now new core ultra looks like an unfinished product, with advantages in very limited scenarios.
And ton of disadvantages, like buying new mother board and overpriced 285 and 245 models.

Maybe 240F will be interesting, with a low price point. And popular. People gonna buy it, just to sit and wait for the next 3xx Gen.