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

[FredFlint]

They need to rework Arrow Lake, remove the E-cores if need be for some parts of the lineup.

E-cores have been a disaster for Intel, the only benefit has been in lower power / more efficient mobile CPUs. Useless for desktop PCs (at least compared to larger / higher power cores).

Laptops is big market and power is a big selling point to many laptop buyers.

E-Cores will probably go the way AMD is going, high density low power cores that have the same functionality as the high power cores. Think phones are also going to go in that direction as it makes the software side easier.

 

[g67575]

Well the E-cores are getting better at least. But, I always found it strange that Intel wanted more E-cores than P-cores, even when the performance gap between the 2 types was quite large.

Low power cores would imply lower clock speeds, presumably?

If they can build P core like cores, that simply run at lower clock speeds, and use less physical die space, that doesn’t sound so bad. But there’s still an IPC gap with current E cores.

 

[WARburton]

e-cores are only on desktops because intel couldn't put any more p-cores in..

14900k is a hot CPU, think if it have 12 p-cores
what 600w with a TJ-Max of 170c

that had to find a way of adding more performance, e-cores are it.

 

[g67575]

I think Intel is always trying to push for higher clock speeds (because of the marketing focus) even with basically the same microarchitecture (Goldencove to Raptorlake) and Intel fabrication process.

If they limited their clock speeds to something like 4.5 to 5 Ghz, I’m sure that would reduce power consumption and heat on the P-cores, which would presumably allow a 12 core design. That’s because clock speeds closer to the ‘base clock’ typically use much lower voltage.

The E cores let Intel have it both ways, the figure on the box can say turbo 6ghz (the 14900K is marketted as 36M Cache, up to 6.00 GHz) while the e cores run below 5ghz (4.4 Ghz).

Generally, I think Intel’s desktop CPUs have been getting less power efficient over time, with lower base clocks. The 9900KS had a base clock of 4.0Ghz for example:
https://www.intel.com/content/www/u...-16m-cache-up-to-5-00-ghz/specifications.html

In comparison, the base clock on the 14900KS is just 3.2Ghz.

 

[jigger]

I think Intel is always trying to push for higher clock speeds (because of the marketing focus) even with basically the same microarchitecture (Goldencove to Raptorlake) and Intel fabrication process.

If they limited their clock speeds to something like 4.5 to 5 Ghz, I’m sure that would reduce power consumption and heat on the P-cores, which would presumably allow a 12 core design. That’s because clock speeds closer to the ‘base clock’ typically use much lower voltage.

The E cores let Intel have it both ways, the figure on the box can say turbo 6ghz (the 14900K is marketted as 36M Cache, up to 6.00 GHz) while the e cores run below 5ghz (4.4 Ghz).

Generally, I think Intel’s desktop CPUs have been getting less power efficient over time, with lower base clocks. The 9900KS had a base clock of 4.0Ghz for example:
https://www.intel.com/content/www/u...-16m-cache-up-to-5-00-ghz/specifications.html

In comparison, the base clock on the 14900KS is just 3.2Ghz.

For Intel to scale CPU core count they have to drop the Ringbus design or reduce the performance of each significantly, at which point they might as well just use an all E core design like those seen in the Xeon 6 parts.

We’ve already seen the peak P core scaling with the 10900k.

 

[g67575]

I think they could just copy AMDs chiplets design with 8 CPU cores per CCX ( Introduced with the Zen 2 lineup, which featured a ‘chiplet-based design for consumer CPUs featuring two compute dies or CCDs and one I/O die’). It’s clearly possible, supposedly AMD has a 16 core CCX design in the works as well, at which point I think Intel will be struggling even more. Even gamers will be interested in the 12 and 16 core parts.

E cores should be an afterthought / additional feature, not a core part of Intel’s design. It has not helped their Arrow Lake CPUs at all, the design has set them back significantly.

In general, I think e cores only made sense for mobile CPUs… That’s where Intel has gone wrong, by not designing the lineups separately to play to their strengths.

 

[Rroff]

I think they could just copy AMDs chiplets design with 8 CPU cores per CCX. It’s clearly possible, supposedly AMD has a 16 core CCX design in the works as well, at which point I think Intel will be struggling even more.

E cores should be an afterthought / additional feature, not a core part of Intel’s design. It has not helped their Arrow Lake CPUs at all, the design has set them back significantly.

Intel have their tile designs, etc. which are similar. The rumoured desktop Bartlett CPU would supposedly according to some sources use 2x 6P core tiles with hyper-threading for 24 threads and a hybrid bus configuration with some kind of "high bandwidth switch".

Intel's current desktop designs support up to 12 ring stops, though there is a bit of a latency penalty beyond 10 performance cores and then there is the whole "rentable units" approach they are working on which somewhat alleviates ring bus issues.

 

[g67575]

Advancements in TSMC’s fabrication process (3nm, and eventually 2nm) will probably allow a denser CCX design with more cores.

Presumably, Intel will find it difficult to compete without a proper improvement to their own fabrication process, which has been delayed / cancelled in 2024 (Intel 20a). Fingers crossed for 18a, because I think their company could really use a break.

I think they need a plan to get the 18a process ready for desktop CPUs, 18a hasn’t been confirmed yet for a desktop lineup.

Meanwhile, AMD CPUs are becoming more widespread in PCs and laptops:
https://www.techspot.com/news/106184-steam-survey-highlights-amd-amazing-year-almost-1.html

 

[SpellowHouse]

No idea what you want to see based on what you've typed. The 9800x3D is beat CPU for gaming at any resolution as long as you've got at least a mid level GPU to run it and you want best performance.

If you are looking at price to performance then take a look at 9700x at £290. Compared to the current price of the 9800x3D at £529, you'll be loosing 30% performance but course cost to price is better. It is similar performance to 14700k/14900k but gives option to get an x3D chip if you need later. All assuming this is about gaming.

You are missing the point. And the point is that it's not the best choice for everyone. If you play in 1080P and want limitless FPS, and don't care about things like idle power consumption, then yes, it is. However, if you don't meet that criteria then it is not clear which is the right CPU for people to buy. I think the media are presenting is as the only gaming CPU and that's simply not true. They are prioritising FPS above everything else when gamers may not actually need what's on offer.

I remember hearing one reviewer on youtube saying that "well, there's no point in presenting 4K data because all the graphs would be the same". My point is there is every reason to do so! Because that they are the same is important information, even if it's not what the youtubers like to present. Nope, sorry, there is way too much hype surrounding the 9800X3D.

As someone suggested in this thread, I am waiting for the 9950X3D to arrive, not because I am only interested in the 9950X3D but because it should be accompanied with a slew of broader tests.

I might add that the promised presentations and improvements to intel don't seem to have happened, so I have just about given up on that.

 

[g67575]

The Ryzen 7700X works well in games for me. I managed to undervolt it more after a recent BIOS update too, the clockrate is decent.

Protip - BIOS screenshots can make it much easier to update the firmware, and keep your important settings.

The Ryzen 9000 series has been unimpressive, except for the expensive 9800X3D.

AMD has been aiming at the premium, gamer market, that is likely to continue with the next gen also.

 

[Armageus]

You are missing the point.

It sounds more like you are.

They are prioritising FPS above everything else when gamers may not actually need what's on offer.

FPS is the number one priority for any gamer. Not sure why you would think they would recommend the "best gaming cpu" on any other merit.



Regardless, if you want to continue to discuss not buying a 9800X3D for whatever reason, then you need to open your own thread, not continue to discuss it in this one which is about Intel Arrow Lake CPUs.

 

[Pottsey]

It sounds more like you are.


FPS is the number one priority for any gamer. Not sure why you would think they would recommend the "best gaming cpu" on any other merit.



Regardless, if you want to continue to discuss not buying a 9800X3D for whatever reason, then you need to open your own thread, not continue to discuss it in this one which is about Intel Arrow Lake CPUs.

FPS is not the number one priority for any gamer as there are other merits to look at. For exmple like a lot of gamers I have an MMO on my right hand screen and my main game on the left hand screen multitasking matters more then pure FPS in this case. Plus the games I play like Factorio and Stellaris is not about FPS, its about UPS that is the key measurement and a growing number of reviews are looking at UPS that's why so many are adding in Factorio style benchmarks.

Though I do agree on your 9800X3D comments. While I am interested in that CPU its presence in this thread is overly heavy.

 

[Curlyriff]

You are missing the point. And the point is that it's not the best choice for everyone. If you play in 1080P and want limitless FPS, and don't care about things like idle power consumption, then yes, it is. However, if you don't meet that criteria then it is not clear which is the right CPU for people to buy. I think the media are presenting is as the only gaming CPU and that's simply not true. They are prioritising FPS above everything else when gamers may not actually need what's on offer.

I remember hearing one reviewer on youtube saying that "well, there's no point in presenting 4K data because all the graphs would be the same". My point is there is every reason to do so! Because that they are the same is important information, even if it's not what the youtubers like to present. Nope, sorry, there is way too much hype surrounding the 9800X3D.

As someone suggested in this thread, I am waiting for the 9950X3D to arrive, not because I am only interested in the 9950X3D but because it should be accompanied with a slew of broader tests.

I might add that the promised presentations and improvements to intel don't seem to have happened, so I have just about given up on that.

So 4k results right (which there a good dozen reviews on if you want not 1080p numbers) can be found. The idle power and the non-gaming performance are the two knowns which for both if you want those you pick Intel. And with that it would be the 14700k then for price/performance numbers. It similar to the 9700X, a little more monies but does better in idle and some non-gaming side things. Only issue it is dead upgrade path which why more would recommend the AMD side. And yes in terms of purely gaming is still a chunk slower then the 9800X3D whatever resolution you want it to be. Gaming performance is just that. People are saying it is the best gaming chip because it offers the highest FPS numbers at any resolution and that is all that counts for the large majority of gamers.

Edit: and you keep saying there is other criteria with no demonstration of what you want then. What would you think most gamers out there are looking for. From all accounts generally 1080p screens are still the highest % used and most gamers at least from forums etc seem to be chasing the 120fps + numbers as more monitors at all resolutions are able to do higher Hz. Like the new 5K2K screens from LG even if you on high end can do 165Hz, good luck getting a current gen machine managing to do that even without some DLSS Upscaling or similar.

Here is a link to one such review showing the 4k numbers. It still shows the CPU to be 17% & 21% faster average over 14 games compared to the other two chips shown. Noting the shown chip being the 15th Gen Ultra 285K being outdone in gaming compared to a 7700x even at both 1080p and 4k so Intel latest certainly shouldn't be way to go for gaming. £260 chip vs £600 has got to sting for Intel honestly. Their 14700k being their best option still tbh. But as above not really ideal now.

Is the Ryzen 9800X3D Truly Faster for Real-World 4K Gaming?

Misconceptions in tech are endless, but myths about low-res CPU benchmarks top our list. Let's revisit this hot topic and test the new 9800X3D at 4K to...

www.techspot.com

 

[g67575]

I think the 14700f is quite good in terms of value, at around £290. Still decent for gaming.

Probably spending more than that on an Intel CPU is a waste of £££.

Or, you can get something slightly better 2nd hand, assuming you don’t buy a CPU that has degraded…

They fixed the crashes / voltage issues by now, I assume?

From that perspective, the 245K could really use a pricecut.

But, overall the AM5 platform just seems like a better option, and the firmware is pretty decent now, and stable.

 

[Vimes]

Going back to November last year there was a video, featuring Robert Hallock, noting the intent of Intel to, hopefully, resolve a number of issues that they had recognised were compromising the performance of this platform, beyond what they had expected.
With the firmware / BIOS updates that are currently available do they represent all of what Intel are able to do to try and improve the acknowledged performance concerns...?

Maybe I have missed updated follow-up information or news. Unless what we see today, with the updated microcode and other fixes, they represents all what is possible to achieve..?
Or, is it still a job in progress..?

 

[Shaolin Chicken]

Going back to November last year there was a video, featuring Robert Hallock, noting the intent of Intel to, hopefully, resolve a number of issues that they had recognised were compromising the performance of this platform, beyond what they had expected.
With the firmware / BIOS updates that are currently available do they represent all of what Intel are able to do to try and improve the acknowledged performance concerns...?

Maybe I have missed updated follow-up information or news. Unless what we see today, with the updated microcode and other fixes, they represents all what is possible to achieve..?
Or, is it still a job in progress..?

I don't know if I anymore fixes are going to be released, but I have read they are going to do a keynote at CES. Hopefully there is some news about Arrowlake.

 

[Vimes]

I don't know if I anymore fixes are going to be released, but I have read they are going to do a keynote at CES. Hopefully there is some news about Arrowlake.

Thanks. There seemed to be a focus on what was wrong, and then how what could be addressed would then be resolved. Since then it seems to have been somewhat quiet....

 

[Dave2150]

I don't know if I anymore fixes are going to be released, but I have read they are going to do a keynote at CES. Hopefully there is some news about Arrowlake.

Best hope for Intel at CES is a sneak peak of Nova Lake (16th gen). There's no fixing Arrow Lake - it'll never get close to the competition (Zen5X3D).

 

[TrixP10]

Best hope for Intel at CES is a sneak peak of Nova Lake (16th gen). There's no fixing Arrow Lake - it'll never get close to the competition (Zen5X3D).

IMHO the laptop reveal at CES is a mess

Intel fills out Core Ultra 200 laptop CPUs with hodgepodge of CPUs, GPUs, and NPUs

Core Ultra 200U, 200H, and 200HX will all fall short of Copilot+ compatibility.

arstechnica.com

 

[g67575]

It looks like Lunar Lake CPUs are a better option than Arrow Lake if people want all the latest AI features.

But it’s mobile only . So little reason to buy an expensive ARL CPU on desktop or mobile.