• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Intel Core Ultra 7 265K

The focus on lower power consumption is a step in the right direction for Intel, but this feels like a dejavu when you compared this with the launch of Alder Lake. Alder Lake started off fairly efficient as a new CPU design, and by the next gen, the power consumption went wild. While Alder Lake sees an overall improvement in performance as compared to Comet Lake and even Tiger Lake (on the mobile form), Arrow Lake sees performance regression or parity in a number of tests/ workload. So I do think the results are disappointing. Not to mention, you need fast DDR5 to get the most out of Arrow Lake CPUs, which actually increases the cost since anything above DDR5 6000 or 6400 will get more expensive.
 
W1zzard said:
24H2 = red bar:
View attachment 368881
Click to expand...
Do you happen to have more data with 24h2? Currently it seems broken on all Intel chips (alderlake, raptorlake and apparently with what you just posted core ultra). Win 10 is > 10% faster than win 11 24h2 - not in all games to be fair. Is it the same case with zen 5?
 
I see that on the CUDIMM scaling page there are only productivity tests. Does way faster RAM not make much of a difference in gaming with this CPU?

I got the 265KF from Amazon for $209 and I'm wondering if it's worth selling the 6400 cl32 RAM that I already have and getting a 7600 CL36 or 8000 CL38 kit. But I don't know if that's a complete waste or not.
 
Gaming wise not a whole lot of difference.

I have this kit

www.techpowerup.com

Team Group T-Force Delta RGB DDR5-7200 CL34 2x 16 GB Review

Team Group is ready for the next generation of high performance DDR5. Using the popular T-Force Delta RGB once again, Team Group offers a DDR5-7200 kit for hardcore gamers and PC enthusiasts alike. Follow along as we benchmark, compare and overclock to see what all the fuss is about!
www.techpowerup.com www.techpowerup.com

Haven't overclocked it. Using it just at XMP speeds with my 285k
 
Maybe test with faster RAM next time? ARL does really poorly at lower memory speeds (as used here).
 
AntiX3DGlazer said:
Maybe test with faster RAM next time? ARL does really poorly at lower memory speeds (as used here).
Click to expand...
But then people will say I'm giving Intel and AMD unfair treatment by using mismatched memory speeds? How to approach?
 
W1zzard said:
But then people will say I'm giving Intel and AMD unfair treatment by using mismatched memory speeds? How to approach?
Click to expand...
Id say highest ram speeds achievable with a midrange mobo just by turning on xmp with no further tinkering.
 
JustBenching said:
Id say highest ram speeds achievable with a midrange mobo just by turning on xmp with no further tinkering.
Click to expand...
That will hurt AMD a lot, because of the UCLK divider that's activated at higher memory freq
 
W1zzard said:
That will hurt AMD a lot, because of the divider that's activated at higher memory freq
Click to expand...
Nah you keep those at 6000, since the higher ram speeds dont actually improve performance.
 
JustBenching said:
Nah you keep those at 6000, since the higher ram speeds dont actually improve performance.
Click to expand...
But you said "highest speeds achieved with a midrange mobo" ?
 
W1zzard said:
But you said "highest speeds achieved with a midrange mobo" ?
Click to expand...
Ok - let me rephrase - highest speeds that also offer a performance benefit. I mean the only reason youd even want to buy faster ram is to get more performance, if you are not getting more performance whats the point
 
JustBenching said:
Ok - let me rephrase - highest speeds that also offer a performance benefit. I mean the only reason youd even want to buy faster ram is to get more performance, if you are not getting more performance whats the point
Click to expand...
Aha.. how to measure "performance", i.e. which benchmark
 
W1zzard said:
Aha.. how to measure "performance", i.e. which benchmark
Click to expand...
I guess that would depend on the scope of the review. If its gaming focused well its obvious. If its a more general performance review then it gets tricky.

Maybe you could test amd with the normal 6000 ram and then just include a couple of productivity tests that are also tested with 8k ram. Something like 7zip and similar stuff that scale with ram.
 
JustBenching said:
I guess that would depend on the scope of the review. If its gaming focused well its obvious. If its a more general performance review then it gets tricky.

Maybe you could test amd with the normal 6000 ram and then just include a couple of productivity tests that are also tested with 8k ram. Something like 7zip and similar stuff that scale with ram.
Click to expand...
It's a CPU review, let's say for Intel. What do I run all the Zen 5 comparison CPUs at? I can't have multiple results for each of them, because it's like 10+ models.

Sorry for dragging this along, but I want you to understand my dilemma.

> with the normal 6000 ram

Who defined that 6000 is "normal", and does he define 8000 as "normal" for Intel ARL, too? or should it be 6200 vs 8000? or 6000 vs 8800 ?
 
@W1zzard
Easy solution - just test without RAM at all. Can’t have “number bigger” issue without any numbers at all. EZ maffs.

No, but seriously, weren’t there already a community-wide discussion in a dedicated thread about this, no consensus was really reached and mostly everyone just accepted what you do already as good and representative enough?
 
Onasi said:
No, but seriously, weren’t there already a community-wide discussion in a dedicated thread about this, no consensus was really reached and mostly everyone just accepted what you do already as good and representative enough?
Click to expand...
Yeah, and it keeps coming up and nagging me, because I feel like there must be better solution
 
W1zzard said:
It's a CPU review, let's say for Intel. What do I run all the Zen 5 comparison CPUs at? I can't have multiple results for each of them, because it's like 10+ models.

Sorry for dragging this along, but I want you to understand my dilemma.

> with the normal 6000 ram

Who defined that 6000 is "normal", and does he define 8000 as "normal" for Intel ARL, too? or should it be 6200 vs 8000? or 6000 vs 8800 ?
Click to expand...
Oh i understand your dilemma, im not trying to downplay it. Im not suggesting you should test all zen 5 with multiple ram configs, just pick one model (9950x would be my bet) and run 3-4 tests that are memory bound with faster ram.

I consider normal everything that will work flawlessly out of the box with just xmp on "normal" mobos. So no 2 dimm ocer mobos, no tinkering with voltages etc.
 
W1zzard said:
But then people will say I'm giving Intel and AMD unfair treatment by using mismatched memory speeds? How to approach?
Click to expand...
W1zzard said:
That will hurt AMD a lot, because of the UCLK divider that's activated at higher memory freq
Click to expand...
It's AMD's problem not yours.
AMD made it so the Ryzen sweetspot is at 6000.
Zen4 is rated at 5200 and Zen5 at 5600. Anything over is overclock.

It's not Intel's fault that Arrow Lake is rated at 6400. Thus anything under is downclocked RAM.

Ryzen got this:
www.techpowerup.com

DDR5 Memory Performance Scaling with AMD Zen 5

Our AMD Zen 5 Memory Scaling Review examines six different DDR5 memory speeds, including DDR5-8000, DDR5-6400 1:1, DDR5-6400 CL28 and others. We looked into how these speeds impact performance in gaming at four resolutions and a wide range of application workloads.
www.techpowerup.com www.techpowerup.com
Arrow Lake didn't, but if it did and the conclusions would be in favor of higher speed RAM then I think most people (even AMD fans) would agree that you can have mismatched memory speed and still be fair.
W1zzard said:
It's a CPU review, let's say for Intel. What do I run all the Zen 5 comparison CPUs at? I can't have multiple results for each of them, because it's like 10+ models.

Sorry for dragging this along, but I want you to understand my dilemma.

> with the normal 6000 ram

Who defined that 6000 is "normal", and does he define 8000 as "normal" for Intel ARL, too? or should it be 6200 vs 8000? or 6000 vs 8800 ?
Click to expand...
8000 is a bit too extreme.
I don't see a reason to go over 7200. Because then the difference in price between the RAM kits won't be negligible and thus it would negatively impact the total platform cost.
With 200S Boost enabled 7200 is more than enough to knock some teeth out of Ryzen.
The slightly expanded suite of tests (the one used for the 9950X3D review) is missing a lot of CPUs, Zen4 as well not just the Arrow Lake lineup.
I am sure that Arrow Lake with 200S Boost enabled and 7200 RAM (UDIMM obviously) would put up a good fight. Not so much against the X3Ds as those are a special breed but against the vanilla Ryzens. The pricing also reflects this, the X3Ds are priced higher because they are special.
 
W1zzard said:
But then people will say I'm giving Intel and AMD unfair treatment by using mismatched memory speeds? How to approach?
Click to expand...
You're right. People took a lot of flak for using D5-7200 when testing ADL, but at least it showed the best-case scenario. Personally I'd rather see either: testing at supported JEDEC with correct timings for both platforms, or a best-case scenario in which all CPUs are running at the highest RAM speed that is a) gamer stable and b) relatively easy to achieve on a wide range of motherboards.
Testing at JEDEC is good because it levels the playing field to what the manufacturer intended to them to be used for, and it's also guaranteed to be near 99% stable. Testing at the max of both CPUs is also helpful because it shows the theoretical best-case scenario in which all CPUs are given their best chance to perform.
 
You must log in or register to reply here.
Back
Top