Intel Readies Core Ultra 3 205, Brings E-cores to the "3" Tier

[btarunr]

Intel may have debuted its Hybrid (heterogeneous multicore) architecture for the desktop with the 12th Gen Core "Alder Lake-S," but the value-ended Core i3 series SKUs throughout the 12th, 13th, and 14th Gen Core processors have remained 4-core/8-thread traditional multicore chips, with just four P-cores. Intel is about to change this with the Core Ultra 200 series "Arrow Lake-S." According to OneRaichu, a reliable source with Intel leaks, the company is giving finishing touches to a pair of Core Ultra 3 series desktop processor SKUs based on the "Arrow Lake" microarchitecture. These will be 8-core chips, a doubling in core-count form the past generations, but the nature of these 8 cores is not yet known.

Among the SKUs in the leak are the Core Ultra 3 205, and the Core Ultra 3 215, both of which are 8-core chips. The two are probably differentiated in a similar manner to past generations of Intel Core i3 desktop did, using cache sizes (eg: Core i3-10100 and i3-10300). The chips probably feature a 4P+4E core configuration, as a "2P+6E" configuration might not be possible, as the E-core clusters are indivisible, although we don't know if the same rule applies to the "Skymont" E-core clusters. The dedicated L2 caches of both the P-cores and E-core clusters could be smaller than on Core Ultra 5 and above SKUs. The Core Ultra 200V "Lunar Lake" processor uses "Lion Cove" P-cores with 2.5 MB of L2 cache per core, while the Core Ultra 9 285K probably has "Lion Cove" P-cores with 3 MB of L2 cache per core.



View at TechPowerUp Main Site | Source

 

[Vayra86]

My god this naming scheme for their products is atrocious. They really are actively seeking to kill any semblance of logic to this, for good reasons obviously, gotta repackage turds proper.

 

[Tomgang]

That's properly also needed. Because with ht gone on the p-cores, there is insufficient threads with out ht and only p-cores.

So this move make sense to me. With 8 cores, you still have those 8 threads former i3 had.

My guess will be a 4p + 4e core configuration. 2p cores seems to little and 6p + 2e cores is to good to be true knowing intel.

 

[Daven]

Could these be E-core only chips?

 

[chrcoluk]

To me the N100 has proven how well this could work for lower end.

Using Windows on a N100 with 4 physical E cores is better than an older I3 or older I5 that is 2 cores with HT.

 

[P4-630]

gotta repackage turds proper.

A turd as your previous i7 8700K you mean?...

To me the N100 has proven how well this could work for lower end.

Using Windows on a N100 with 4 physical E cores is better than an older I3 or older I5 that is 2 cores with HT.

Like my little mobile i3-1315U runs wonderfully with just 2P-4E (8-threads)...

 

[AGlezB]

Could these be E-core only chips?

P-cores generally having higher clock speeds than AMD's is a major marketing point for Intel and if the remove the P-cores they loose in every metric to AMD except maybe power efficiency, which is only relevant at the same level of perfomance. A CPU with a ton of E-cores could be perfectly fine for some workloads but I don't think they can do it, at least within the Intel Core brand.

 

[DudeBeFishing]

P-cores generally having higher clock speeds than AMD's is a major marketing point for Intel and if the remove the P-cores they loose in every metric to AMD except maybe power efficiency, which is only relevant at the same level of perfomance. A CPU with a ton of E-cores could be perfectly fine for some workloads but I don't think they can do it, at least within the Intel Core brand.

What if they glued a stupid amount of cache onto the E-cores?

 

[TheinsanegamerN]

What if they glued a stupid amount of cache onto the E-cores?

Would dramatically increase costs and power use, and would still lose to AMD, since the X3d chips have not just stupid amounts, but perhaps BRAIN DEAD amounts of cache

 

[AGlezB]

What if they glued a stupid amount of cache onto the E-cores?

If the E-cores could reach the performance of P-cores they wouldn't be E-cores.
It is possible such a CPU could exist but, like I said, never under the "Intel Core" brand.

 

[Onasi]

What if they glued a stupid amount of cache onto the E-cores?

That would require a separate silicon (so not just a B0 with deactivated parts) and a drastically different one at that. Which would be costly to produce. Doing something like that for the low-margin 3 line makes absolutely no sense. Reminder that for 13-14 gens Intel just reused H0 Alder Lake dies for lower tier models.

 

[persondb]

That's properly also needed. Because with ht gone on the p-cores, there is insufficient threads with out ht and only p-cores.

I don't think it not having HT will have big impact in multithreading, and it's not like HT is a silver bullet that doubles the multithreading capabilities. It can improve it by 10~20% depending on the workload, with some it being detrimental.

I think it will be interesting to compare Raptor Cove cores to Lion Cove cores as you can generally disable the E-cores in the bios and see how a new arch without HT goes against the old arch with HT. I suspect it won't make as much of a difference as people expect.

 

[Wirko]

There have always been processors below the i3 too, Intel won't just abandon them. 4 P cores and no E cores for the new Pentium Ultra (or 300 Ultra)?

 

[dirtyferret]

My god this naming scheme for their products is atrocious. They really are actively seeking to kill any semblance of logic to this, for good reasons obviously, gotta repackage turds proper.

what? it's easy. It goes in this simple to follow naming scheme of performance tiers

Core Ultra 3
Core Ultra 5
Core Ultra 7
Super Ultra
Ultra Boy
Captain Ultra
Tide Ultra Oxi
and finally
Dawn Ultra Platinum Advance Power! AMD 7800X3D can eat it's grease fighting powers dust!

 

[Darmok N Jalad]

A turd as your previous i7 8700K you mean?...



Like my little mobile i3-1315U runs wonderfully with just 2P-4E (8-threads)...

View attachment 359680

I have the i7 variant, which is 2P+8E. In real life, there are too many scheduling issues for me to appreciate it. I'd take 6C/12T over it easily. I'd even take 4C/8T, TBH, as I find Windows scheduler dumping all the work on the P cores way too often, and even skipping the HT cores.

 

[Rowsol]

I'm just glad to see the low end see an update.

 

[P4-630]

I have the i7 variant, which is 2P+8E. In real life, there are too many scheduling issues for me to appreciate it. I'd take 6C/12T over it easily. I'd even take 4C/8T, TBH, as I find Windows scheduler dumping all the work on the P cores way too often, and even skipping the HT cores.

Using Chrome OS here.

 

[Darmok N Jalad]

Using Chrome OS here.

I really do think it's Windows scheduler not using the resources correctly. It's just like with the Zen 5 reviews--under Linux we see more performance gains over Zen4 than we do under Windows.

 

[P4-630]

I really do think it's Windows scheduler not using the resources correctly. It's just like with the Zen 5 reviews--under Linux we see more performance gains over Zen4 than we do under Windows.

The thing that I do know about my little i3 on Chrome OS is that it´s stays responsive , even with 30 browser tabs open, and that with just 8GB ram,
it´s ram is doing some magical things because it never gets full it seems...

 

[AGlezB]

The thing that I do know about my little i3 on Chrome OS is that it´s stays responsive , even with 30 browser tabs open, and that with just 8GB ram,
it´s ram is doing some magical things because it never gets full it seems...

https://en.wikipedia.org/wiki/Memory_paging
I don't know if memory management can be called magic outside Clarke's third law.

 

[P4-630]

https://en.wikipedia.org/wiki/Memory_paging
I don't know if memory management can be called magic outside Clarke's third law.

Gemini said:

Chrome OS uses a technique called zRAM to compress infrequently used data in memory.
This effectively creates more free space, allowing more tabs and apps to remain active without causing slowdowns.

 

[Darmok N Jalad]

The thing that I do know about my little i3 on Chrome OS is that it´s stays responsive , even with 30 browser tabs open, and that with just 8GB ram,
it´s ram is doing some magical things because it never gets full it seems...

macOS does the same thing. You can really push the memory on it, but it comes at the cost of offloading to the NMVE. In practice, the OS doesn’t seem to miss a beat, but I’m not a fan of hitting the solid state storage like that, even if it can handle years of rw work. More RAM is definitely the long term solution.

 

[AGlezB]

Gemini said:

Chrome OS uses a technique called zRAM to compress infrequently used data in memory.
This effectively creates more free space, allowing more tabs and apps to remain active without causing slowdowns.

Yup. I'd say "without causing noticeable slowdowns as long as there is RAM available" but thats the gist of it. Compressing the swapped pages is particularly important when moving them to a slower storage since the compression times tend to be a lot lower than the write+seek+read times.

 

[Speedyblupi]

A CPU with a ton of E-cores could be perfectly fine for some workloads but I don't think they can do it, at least within the Intel Core brand.

They already did, with the Core i3-N305. That's 8 E-cores and 0 P-cores.
Though arguably it doesn't count, because it's a laptop CPU. They've never done anything similar on desktop, and I don't expect they will any time soon.

 

[AGlezB]

They already did, with the Core i3-N305. That's 8 E-cores and 0 P-cores.

They just never say it and a while a savy buyer might notice a x86 CPU with 8 cores and a 15W TDP makes no sense unless the cores are cut down to nothing, the average person looking for a cheap laptop will not.
To be clear when I said they can't do it, I meant it from a commercial/marketing POV. At normal (for x86) TDP levels they have to compete with AMD's "all our cores are P-cores" and below that is ARM territory. Plus reviewers will eat them alive if they try it on a desktop platform.