• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Intel Core Ultra 9 285k 'Arrow Lake' Discussion/News ("15th gen") on LGA-1851

Mr Jack said:
I'm not sure this is really true. Apart from E-cores being useful to keep background stuff out of the way even if the game you're playing doesn't directly benefit from them, in terms of silicon you get four E-cores for the price of one P-core. Individually they're not offering as much, but four of them probably outweigh a single P-core a lot of the time.

That said, with Arrow Lake dropping Hyper Threading presumably the ratio of transistors P-Core to E-Core is much lower, but you're also getting less threads running on the P-Cores so... who knows?
Click to expand...
I guess we'll never really know..., unless Intel produce dies with them removed, but from the benches I've seen, having the E-Cores disabled on 12-14th gen is mostly within margin of error when benched in games and for secondary tasks, the FPS loss is usually only a few percent on a 6-core CPU, but that's synthetic testing (e.g. always opening the same browser tabs and the same apps like Discord) so may not be representative of real world usage.

Some games showed large increases with them enabled (Civ VI was one of those and DOTA 2, for some reason).

I'd say that they're only there to keep the CPU competitive in multithreading, Intel wouldn't want mainstream benchmarks like the Xeon E versus Epyc 4004 ones, since they'd make it look like a pretty bad CPU overall.
 
Cyber Ghost said:
Click to expand...


As per other videos have shown the same thing - stick 8400mhz+ cudimm ram in, overclock it, overclock the 285k and then you get decent performance, mostly, but high power draw. It is very weird that a ram overclock can increase cinebench score for the 285k, I've never seen cinebench respond to ram overclocks. It still makes the 285k an expensive system - having to buy the 285k, a new board, new ram and then overclocking everything and it's not a universal fix - some games still don't respond, the 285k still loses by 20% to the 14900k in HZD


Also as Jay says right at the end of the video, even if you like what you see from these 285k tweaks, his words not mine, he says all of these Intel numbers are about to look small once the 9800x3d is added to the charts next week
 
Last edited:
Aren’t CUDIMM mobos extremely expensive? So CUDIMM seems to make a significant change in some scenarios, overclock doesn’t look worth it.

Edit: ignore me, confused CUDIMM with CAMM2
 
Last edited:
Grim5 said:
even if you like what you see from these 285k tweaks, his words not mine, he says all of these Intel numbers are about to look small once the 9800x3d is added to the charts next week
Click to expand...
The 7800X3D already does that.

I wonder what the prices will be like for the 7800X3D, once the 8000 series version launches?
 
Grim5 said:
Also as Jay says right at the end of the video, even if you like what you see from these 285k tweaks, his words not mine, he says all of these Intel numbers are about to look small once the 9800x3d is added to the charts next week
Click to expand...
It would be nice if these kind of changes could be made easily 'out of the box'. E.g. just a single BIOS preset tweak.

Intel needs a gaming optimised preset, available for all motherboards.
 
Last edited:
Grim5 said:
As per other videos have shown the same thing - stick 8400mhz+ cudimm ram in, overclock it, overclock the 285k and then you get decent performance, mostly, but high power draw. It is very weird that a ram overclock can increase cinebench score for the 285k, I've never seen cinebench respond to ram overclocks. It still makes the 285k an expensive system - having to buy the 285k, a new board, new ram and then overclocking everything and it's not a universal fix - some games still don't respond, the 285k still loses by 20% to the 14900k in HZD
Click to expand...

As an experiment, run R23 at base speeds and record the score, now run the same test again with XMP and you will often see an uplift. Using my 7950X3D as an example, on default setting scored around 35K and then with XMP enabled it was then scoring 36K. While it's a small gain, it's still a gain of just changing one setting same has often been true on my Intel systems.

I can't comment on the CU DIMMs as I don't have any yet to test with but to get 9000 on normal DIMMs (24GB M-Die) isn't too hard you are limited by the IMC on the 13/14th Gen then it is a case of trimming out the voltages as they sweet spot hard at those levels.

Once I have all my bits, I can run the same tests but measure power from the wall and see how it does.

OCF and 285K's are still pending :(
 
g67575 said:
It would be nice if these kind of changes could be made easily 'out of the box'. E.g. just a single BIOS preset tweak.

Intel needs a gaming optimised preset, available for all motherboards.
Click to expand...

Intel has been the same way for ages now, which is often why it needs some fafo at times.
 
  • Like
Reactions: G J
I imagine we'll get the 16th generation in October 2025.

I don't think there's much chance of it being a 20a/18a based series. I think I read somewhere that the 18a process isn't going to go into high volume production until 2026.

EDIT - actual quote:
"We will start to ramp to higher volumes in 2026 for 18A." "The bulk of our wafers in 2025 are driven by Intel 7 and Intel 10," Gelsinger said.
Click to expand...

But, maybe a refreshed and optimised version of Arrow Lake will seem more worthwhile, especially for anyone looking to buy a new platform.

I guess there's a possibility of using an improved TSMC node for their Oct. 2025 series. If they can boost clock speeds vs Arrow Lake by say 10%, I think it will be easier to exceed the performance of the 14th gen.

Intel's foundries aren't doing well, considering that they do not have a mainstream CPU series that makes use of EUV for the CPU tile, except for Meteor Lake on Intel 4.

I still think Intel is trying to play for time at the moment, and probably the main reason for the problems with Arrow Lake, relates to using TSMC's 3nm, rather than their 20a process, as originally planned.
 
Last edited:
Tetras said:
The point is still valid though, a CPU without them would likely be just as effective for gaming in the vast majority of games and better value.
Click to expand...
Mr Jack said:
That said, with Arrow Lake dropping Hyper Threading presumably the ratio of transistors P-Core to E-Core is much lower, but you're also getting less threads running on the P-Cores so... who knows?
Click to expand...

this is really my point.. HT uses zero watts it uses the unused power from the main cores. but it is shown to have a very positive affect on gaming, 6c vs 6c/12t
but the Ecore's do use power and are shows to have little to no affect on gaming... so why they just keep strapping on more and more is just silly

yes they have a positive affect on productivity and bench marks. but as shown dropping HT as a negative affect on that stuff too.

the new flagship CPU is 8c/8t + 16Ec, the CPU is replacing was 8c/16t + 16Ec and it cant keep up that tells you HT works
and in gaming mode the CPU still pulls 300w..... with 8 less threads

lets says an Ecore pulls 10w(maybe more or less IDK) but there is 16 if you want to drop power get rid of 10...



Anyway is 2023 intel was talking about 3d stacked cache in CPU in the future so maybe the best is still to come
 
WARburton said:
this is really my point.. HT uses zero watts it uses the unused power from the main cores. but it is shown to have a very positive affect on gaming, 6c vs 6c/12t
but the Ecore's do use power and are shows to have little to no affect on gaming... so why they just keep strapping on more and more is just silly
Click to expand...

The energy cost of HT is small rather than strictly zero, but even if it was the draw of a single P-core is more than twice the draw of an E-core so running two E-cores is more energy efficient than hyperthreading a P-core. I don't know about a positive impact on gaming though, there's plenty of people reporting that turning off HT improved their frame rates on, e.g. the 14900K; although I rather suspect it depends on what exactly you're running. As for why they keep adding E-cores, well, because you can put 4 E-cores on for the price of 1 P-core. The question on E-cores isn't whether P-cores help more (they obviously do) but whether, e.g. having just 8 P-cores is better than having 6P/8E.

They seem to have gone off half-cocked with this release, but that doesn't mean that the idea of dropping HT is a bad idea in itself. I guess the main advantage is that it boosts cache efficiency, which HT is really bad for, and that in turn allows for better memory management - something they're rumoured to be making big changes to for Panther Lake and Nova Lake. There's also been a number of security issues with HT.
 
HT never seemed all that efficient to me, particularly on older CPUs. Significantly increased power usage and temps for a relatively small performance improvement. More of a thing when core counts were more limited.

The impression I got from Intel, is that HT wasn’t worth the die space on Arrow Lake.
 
Last edited:
Grim5 said:
As per other videos have shown the same thing - stick 8400mhz+ cudimm ram in, overclock it, overclock the 285k and then you get decent performance, mostly, but high power draw. It is very weird that a ram overclock can increase cinebench score for the 285k, I've never seen cinebench respond to ram overclocks. It still makes the 285k an expensive system - having to buy the 285k, a new board, new ram and then overclocking everything and it's not a universal fix - some games still don't respond, the 285k still loses by 20% to the 14900k in HZD


Also as Jay says right at the end of the video, even if you like what you see from these 285k tweaks, his words not mine, he says all of these Intel numbers are about to look small once the 9800x3d is added to the charts next week
Click to expand...
so the AMD Ryzen 9950X3D coming out next week?

this will beat the Ultra 9 285k in performance?

And i bet they will be the same price 6-700 quid
 
hobib_ali said:
so the AMD Ryzen 9950X3D coming out next week?

this will beat the Ultra 9 285k in performance?

And i bet they will be the same price 6-700 quid
Click to expand...
Its the 9800X3D that's coming out next week (7/11/24) not the 9950X3D. I think the 9950X3D is due out next year

EDIT - AMD released the 9800X3D early because of the new Intel processors that have just come out but by the looks of things they didn't really have anything to worry about
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom