News Core Ultra 9 285 65W CPU beats Core i9-14900K in early Geekbench benchmark — Arrow Lake chip shows 15% better-multithreaded performance than predec...

[Admin]

Initial results at Geekbench indicate that the Core Ultra 9 285 65W CPU can handily match the i9-14900K.

Core Ultra 9 285 65W CPU beats Core i9-14900K in early Geekbench benchmark — Arrow Lake chip shows 15% better-multithreaded performance than predec... : Read more

 

[TheHerald]

True, the 35w T lineup will be the go to for efficiency oriented people. Every gen Intel sets the efficiency bar higher and higher, excited to see what they managed with arrow lake. Will they increase their lead, will they still lose to the 14900t, it's gonna be interesting. .

 

[NightLight]

Impressive stuff indeed. It's amazing how a little crisis can focus people.

 

[yahrightthere]

Not sure about this being true, the 14900K results isn't not showing this to be true, or am i missing something?
https://browser.geekbench.com/search?q=14900k

Gigabyte Technology Co., Ltd. Z890 UD - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. Z890 UD with an Intel Core Ultra 9 285 processor.

browser.geekbench.com

 

[YSCCC]

Not sure about this being true, the 14900K results isn't not showing this to be true, or am i missing something?
https://browser.geekbench.com/search?q=14900k

All early leaks can be some sort of propaganda one way or the other, so better wait till the mass reviews come out before becoming overly optimistic or vice versa

 

[Gururu]

I dont understand. For months we've been told otherwise. What the heck is going on?

 

[jasonf2]

With the 14++++++ node stick Intel just kept figuring out how crank every mhz out by bumping up TDP. It is nice to see that Intel appears to recognize that if they don't get power efficiency back to par they will ultimately lose the whole game to AMD/ARM/RISC-V.

 

[TheHerald]

Not sure about this being true, the 14900K results isn't not showing this to be true, or am i missing something?
https://browser.geekbench.com/search?q=14900k

Gigabyte Technology Co., Ltd. Z890 UD - Geekbench

Benchmark results for a Gigabyte Technology Co., Ltd. Z890 UD with an Intel Core Ultra 9 285 processor.

browser.geekbench.com

It is true if you compare stock vs stock. The results you are linking might have oced memory. My 12900k can hit 20k at stock just with fast memory.

In any case, geekbench is largely useless anyways. Only the ST portion is somewhat okayish, the MT part just doesn't scale with cores.

 

[Alvar "Miles" Udell]

50-75% faster than my 5950X. A few years ago before the GPU became more expensive than the motherboard and CPU combined, I might be tempted to upgrade.

 

[diminishedfifth]

Excuse the ignorance but why would you want power efficient desktop CPU?
I built a gaming PC about a year ago, so that's my baseline for things. I always assumed the purpose getting a desktop was to push maximum values, with little to no care for efficiency?

 

[Pigpig]

looks like this version is non k version, wonder why the RAM speed set to 5600mhz? suppose non k version support up to 7200mhz ram speed , at least should use 6400mhz in stead of using 5600mhz

 

[TheHerald]

Excuse the ignorance but why would you want power efficient desktop CPU?
I built a gaming PC about a year ago, so that's my baseline for things. I always assumed the purpose getting a desktop was to push maximum values, with little to no care for efficiency?

Costs of electricity, excessive heat / noise, enviromental reasons.

I mean if you want to achieve maximum performance, an efficient design is the way to get there. You can't get an absurdly inefficient design and push it to high perforrmance cause you won't be able to cool it. The 14900k is a prime example of it, the reason it can get so high in performance (breaking 40k CBR23 score is because it is incredibly efficient. Take an inefficient CPU (like the 7700x for example) and try to push it to 40k, it will consume over 800 watts.

 

[JasonovichClone]

Costs of electricity, excessive heat / noise, enviromental reasons.

I mean if you want to achieve maximum performance, an efficient design is the way to get there. You can't get an absurdly inefficient design and push it to high perforrmance cause you won't be able to cool it. The 14900k is a prime example of it, the reason it can get so high in performance (breaking 40k CBR23 score is because it is incredibly efficient. Take an inefficient CPU (like the 7700x for example) and try to push it to 40k, it will consume over 800 watts.

Are you serious?
Since when has Ryzen 7000 series been less efficient than the Rapture Lake toaster?
Thanks for the chuckle 🤭 🤭

 

[Alvar "Miles" Udell]

Excuse the ignorance but why would you want power efficient desktop CPU?
I built a gaming PC about a year ago, so that's my baseline for things. I always assumed the purpose getting a desktop was to push maximum values, with little to no care for efficiency?

Turbo speeds for one, as well as overclocking headroom potential, but also with an increasing drive for smaller, more compact systems that fit in an mATX case, or smaller, which can't fit a beefy 360mm liquid cooler, having a smaller thermal load is beneficial. Also think about systems like the Mac Mini and how Apple silicon allowed them to pack a huge punch in a tiny format.

 

[TheHerald]

Are you serious?
Since when has Ryzen 7000 series been less efficient than the Rapture Lake toaster?
Thanks for the chuckle 🤭 🤭

I didn't say they are. 7000 series includes a lot of CPUs, some of them like the 7950x / 7950x 3d are incredibly efficient in MT workloads (still bad at ST) . The rest are kinda bad though, due to lack of cores they need a lot of power to get half decent performance in MT.

Going by your signature you have a 7800x 3d, which is basically identical in MT to the 7700x i mentioned. That CPU needs what, ~90w for 20k CBR23? That's inefficient man. I have a 3 year old CPU that scores higher than that at 90w.

 

[SyCoREAPER]

Yeah well.. I'll let these sit in the wild for a while. New socket, new design. They couldn't even get the 13 and 14 Gen i9s right. Not about do dump money until they've proven themselves.

 

[Matrix8000]

I’m glad it’s faster, but 15% isn’t making me rush out to buy. I understand Apples chips needing efficiency in tablets and iMacs, but I’m just going to liquid cool on a 1200-w power supply. Are those that need killer performance focused on mATX designs?

 

[AnotherNobody]

Always skeptical of benchmarks that list incorrect specs. It's very unlikely the CPU has 4 memory channels. Certainly the 285K only has 2.

 

[halfcharlie]

looks like this version is non k version, wonder why the RAM speed set to 5600mhz? suppose non k version support up to 7200mhz ram speed , at least should use 6400mhz in stead of using 5600mhz

How about you read the article you're commenting on which specifically mentions this?

I’m glad it’s faster, but 15% isn’t making me rush out to buy. I understand Apples chips needing efficiency in tablets and iMacs, but I’m just going to liquid cool on a 1200-w power supply. Are those that need killer performance focused on mATX designs?

Only the rich needlessly upgrade every gen, it isn't for you. It's for the vast majority who upgrade every 3-5 years, and as has been mentioned better efficiency always leads to better performance, that's basic critical thinking. I'll be waiting for both this and 9000X3D chips to have proper benchmarks etc of course, but coming from my 11th gen chip it will be a huge upgrade no matter what.

 

[thestryker]

Always skeptical of benchmarks that list incorrect specs. It's very unlikely the CPU has 4 memory channels. Certainly the 285K only has 2.

Geekbench counts each DDR5 channel and there are technically two per module.

If you want to learn a bit more about the DDR5 specs: https://www.anandtech.com/show/1591...sed-setting-the-stage-for-ddr56400-and-beyond

looks like this version is non k version, wonder why the RAM speed set to 5600mhz? suppose non k version support up to 7200mhz ram speed , at least should use 6400mhz in stead of using 5600mhz

CUDIMMs don't appear to be shipping in volume so no OEMs are using them yet. To run DDR5-6400 (or higher) using JEDEC specifications CUDIMMs are mandatory.

 

[watzupken]

To be honest, Raptor Lakes were never known to scale well at lower power levels. I think there are already many reviews that proved this point. In this test, I am unsure what were the conditions that resulted in the results, and I don't want to speculate. Best to wait for independent reviews to confirm how Arrow Lake will actually perform. I think the only conclusion that I can draw is that Arrow Lake seems to be more frugal when it comes to power requirement.

 

[YSCCC]

To be honest, Raptor Lakes were never known to scale well at lower power levels. I think there are already many reviews that proved this point. In this test, I am unsure what were the conditions that resulted in the results, and I don't want to speculate. Best to wait for independent reviews to confirm how Arrow Lake will actually perform. I think the only conclusion that I can draw is that Arrow Lake seems to be more frugal when it comes to power requirement.

What can be seen is It’s should be more efficient than raptor lake, that’s it (but it’s really difficult to not be)

 

[TheHerald]

What can be seen is It’s should be more efficient than raptor lake, that’s it (but it’s really difficult to not be)

How is it not difficult? They need to offer better performance at same power, which is incredibly hard with the drop of HT. Intel just lowering the power limit and being "here you go, increased efficiency" isn't really an increase in efficiency, they are just running the chip at a better voltage curve.

 

[ingtar33]

50-75% faster than my 5950X. A few years ago before the GPU became more expensive than the motherboard and CPU combined, I might be tempted to upgrade.

you do know geek bench isn't valid to compare different chip structures right? it's barely useable to compare iterations like 12th to 13th gen. let alone a major redesign like the 285. scores on one structure do not directly compare to scores on a seperate structure. it was originally designed to score mobile chips; attempting to grade them across various designs and the like. and didn't really do that very well either. It does far worse with x86/64.

this is why THG isn't comparing these results to the recent ryzen chips released. you can't compare intel vs amd x86/64 chips with geekbench. as a benching tool it sometimes is accurate, sometimes isn't and it isnt until you see actual benches that represent real world performance do you know if the geekbench results were worth the time to make take them.

 

[JasonovichClone]

I didn't say they are. 7000 series includes a lot of CPUs, some of them like the 7950x / 7950x 3d are incredibly efficient in MT workloads (still bad at ST) . The rest are kinda bad though, due to lack of cores they need a lot of power to get half decent performance in MT.

Going by your signature you have a 7800x 3d, which is basically identical in MT to the 7700x i mentioned. That CPU needs what, ~90w for 20k CBR23? That's inefficient man. I have a 3 year old CPU that scores higher than that at 90w.

My CPU temperature is averaging 45 c from AIO cooling.
The noise levels from the case are reasonable.
Great gaming, OK on other stuff.
Can't complain 🙂