Raptor Lake P-Core only SKUs

[alex75871]

Intel has announced P-Core only Raptor Lake SKUs in the form of the Intel Xeon E-2400 series. This comes as no surprise as most virtual machine hosts, including VMware, do not support Intel's hybrid-core architecture. Unfortunately, being a Xeon series, they do not support overclocking. However, the top SKU features a base clock of 3.2 GHz and a boost of 5.6 GHz across all eight performance cores. Additionally, the E-2400 series offers support for 10 and 25 Gigabit Ethernet.

 

There are three 8-core SKUs to choose from, each with varying TDPs. Additionally, there are offerings with 4 and 6 cores.

 

E-2488: 8-cores, 3.2 GHz (base), 5.6 GHz (boost), 24MB cache, 48 combined PCIe lanes, 95w TDP, socket 1700

E-2478: 8-cores, 2.8 GHz (base), 5.2 GHz (boost, 24MB cache, 48 combined PCIe lanes, 80w TDP, socket 1700

E-2468: 8-cores, 2.6 GHz (base), 5.2 GHz (boost, 24MB cache, 48 combined PCIe lanes, 65w TDP, socket 1700

 

My thoughts

The Intel E-2xxx series and its predecessor, the E3-12xx, have always been interesting lineups of CPUs, offering affordable solutions for home and small business servers, workstations, and are sometimes compatible with certain client/consumer motherboards.

 

Sources

https://www.intel.com/content/www/us/en/products/docs/processors/xeon/e-2400-product-brief.html

 

[RejZoR]

Are there even any consumer models that just have 8 P-cores and no E-cores? Coz a 8c/16t running at 5.6GHz all core would be pretty badass gaming chip imo. Would also support AVX512 since only big cores.

[starsmine]

3 minutes ago, RejZoR said:

Are there even any consumer models that just have 8 P-cores and no E-cores? Coz a 8c/16t running at 5.6GHz all core would be pretty badass gaming chip imo. Would also support AVX512 since only big cores.

Given that the the most valid reason for these chips is proffesional software that you pay per core for, I don't see much of any reason for their existence in the consumer market. 

[alex75871]

1 minute ago, RejZoR said:

Are there even any consumer models that just have 8 P-cores and no E-cores? Coz a 8c/16t running at 5.6GHz all core would be pretty badass gaming chip imo. Would also support AVX512 since only big cores.

There are no 8 P-core only SKUs that I am aware of on the consumer side. The great thing about these entry-level Xeons is that they're typically compatible with some consumer motherboards. If not, workstation boards from the likes of Supermicro and ASUS are usually released not long after.

[RONOTHAN##]

I'd be curious about how they're priced, since I don't see many scenarios where it would make sense to buy one of these rather than a 13700K and disable the E-Cores. For prebuilt workstations/server boxes from Super Micro and the like, sure, but for the homelab community these aren't likely to make much sense unless they end up really cheap (same price as a 13600K). 

[RejZoR]

8 minutes ago, starsmine said:

Given that the the most valid reason for these chips is proffesional software that you pay per core for, I don't see much of any reason for their existence in the consumer market. 

Apart from the fact that Intel's hybrid design is highly unreliable and buggy. And same goes for AMD's hybrid gaming/work models that have 2 CCD's, one regular and one with Vcache. This crap just doesn't work. It's why I've gone for 5800X3D instead. Single CCD, single Vcache. Zero problems, max gaming performance. 7800X3D didn't quite exist back then yet, but if I were buying it today and deciding for new platform, I'd decide for 7800X3D any time over any Intel chip just for this fact (or 7950X3D for that matter).

[Crunchy Dragon]

36 minutes ago, alex75871 said:

Unfortunately, being a Xeon series, they do not support overclocking. However, the top SKU features a base clock of 3.2 GHz and a boost of 5.6 GHz across all eight performance cores. Additionally, the E-2400 series offers support for 10 and 25 Gigabit Ethernet.

Being a Xeon CPU, they're also not going to work on the consumer platform. Xeon support in consumer motherboards has been incredibly hit or miss since Skylake, and practically nonexistent since Kaby Lake. Intel's doc states support for C262 and C266 chipsets, which might mean these will appear in smaller servers and higher end workstations, akin to the older Dell PrecisionTowers and Lenovo ThinkStations that ran server chipsets and Xeons.

 

The boost clock is nice, but one thing a lot of people overlook with running a Xeon is that it tends to take more work for them to actually reach the maximum boost clock. I have a Xeon E5-2690v4(2.6Ghz base, 3.5Ghz max boost) and that topped out at 2.9Ghz running Cinebench R23 for the full 10 minutes.

 

Xeons are designed to be workstation and server powerhouses, so their behavior when you put them to work reflects that moreso than your typical consumer/enthusiast-grade CPU that's typically meant to run as fast as possible all the time. Occasionally, you can do some funny things to a Xeon by tinkering with BCLK, but that can also induce instability elsewhere in the system.

[RONOTHAN##]

4 minutes ago, Crunchy Dragon said:

The boost clock is nice, but one thing a lot of people overlook with running a Xeon is that it tends to take more work for them to actually reach the maximum boost clock. I have a Xeon E5-2690v4(2.6Ghz base, 3.5Ghz max boost) and that topped out at 2.9Ghz running Cinebench R23 for the full 10 minutes.

The other thing to look at is how they also lost cache. For workstation tasks it's not likely to matter that much, but for games the 6MB less cache might be a bit of a performance hit when compared to a 13700K at the same frequency and E cores disabled, to the point where they'll probably be about equal. Besides, 5.6GHz on a 13700K is not that hard of an overclock if the E cores are disabled, so the boost clock matters even less. 

 

Given you can get a 13700KF for $320 right now, I have a very hard time seeing a point for this since Xeons have historically been quite a bit more expensive than their consumer equivalents. If they offered a 12 P core option, that would make sense, but for just an 8 core chip they don't. ECC support isn't a factor since W680 is a thing, they have less cache, they have the same amount of cores, just no advantages unless they're significantly cheaper. 

[Crunchy Dragon]

4 minutes ago, RONOTHAN## said:

The other thing to look at is how they also lost cache. For workstation tasks it's not likely to matter that much, but for games the 6MB less cache might be a bit of a performance hit when compared to a 13700K at the same frequency and E cores disabled, to the point where they'll probably be about equal. Besides, 5.6GHz on a 13700K is not that hard of an overclock if the E cores are disabled, so the boost clock matters even less. 

 

Given you can get a 13700KF for $320 right now, I have a very hard time seeing a point for this since Xeons have historically been quite a bit more expensive than their consumer equivalents. If they offered a 12 P core option, that would make sense, but for just an 8 core chip they don't. ECC support isn't a factor since W680 is a thing, they have less cache, they have the same amount of cores, just no advantages unless they're significantly cheaper. 

Agreed.

 

Maybe if these had 12-16 core variants with more cache, they'd be decent bang for buck in 6-8 years on the used market, but outside of their targeted market of small businesses and essentially small cloud/server applications, most of us regular folks won't really have a use for these chips.

[alex75871]

48 minutes ago, RONOTHAN## said:

The other thing to look at is how they also lost cache. For workstation tasks it's not likely to matter that much, but for games the 6MB less cache might be a bit of a performance hit when compared to a 13700K at the same frequency and E cores disabled, to the point where they'll probably be about equal. Besides, 5.6GHz on a 13700K is not that hard of an overclock if the E cores are disabled, so the boost clock matters even less. 

 

Given you can get a 13700KF for $320 right now, I have a very hard time seeing a point for this since Xeons have historically been quite a bit more expensive than their consumer equivalents. If they offered a 12 P core option, that would make sense, but for just an 8 core chip they don't. ECC support isn't a factor since W680 is a thing, they have less cache, they have the same amount of cores, just no advantages unless they're significantly cheaper. 

The absent cache is probably the cache packaged with the E-cores? The P-cores should have the same amount of cache. I am not certain though.

[RONOTHAN##]

Just now, alex75871 said:

The absent cache is probably the cache packaged with the E-cores? The P-cores should have the same amount of cache. I am not certain though.

Yeah, that's the case. The thing is when you disable the E-cores, the P cores still have access to that cache (at least the L3 cache), so it is still worse overall. 

[alex75871]

23 minutes ago, RONOTHAN## said:

Yeah, that's the case. The thing is when you disable the E-cores, the P cores still have access to that cache (at least the L3 cache), so it is still worse overall. 

Ah I see - thanks for clarifying.

[Dabombinable]

Beyond AVX512 support (retroactively disabled in 12th gen due to scheduling problems - E-cores are a shitshow), will these actually prove that useful in servers?

[da na]

Just... eight? 
Just eight cores?

Come on, we were doing eight core Xeons in 2010...

[porina]

1 hour ago, Crunchy Dragon said:

The boost clock is nice, but one thing a lot of people overlook with running a Xeon is that it tends to take more work for them to actually reach the maximum boost clock. I have a Xeon E5-2690v4(2.6Ghz base, 3.5Ghz max boost) and that topped out at 2.9Ghz running Cinebench R23 for the full 10 minutes.

All core boost is usually much lower than single core max boost. Also do you know what power limit was enforced by the system? That can cut things back earlier compared to practically unlimited enthusiast mobos. Or even thermal limits.

 

 

On the others things:

 

https://www.intel.com/content/www/us/en/products/sku/236182/intel-xeon-e2488-processor-24m-cache-3-20-ghz/specifications.html

Doesn't list AVX-512 support there.

 

People getting this class product don't buy CPUs, they buy systems. The platform as a whole is the offering, not the CPU in isolation. Consumer tier gear likely fails in other areas. Basically don't say this sucks as a consumer product when it isn't a consumer product.

 

Not every task needs more than 8 cores. Other products exist if you genuinely need more.

[Crunchy Dragon]

20 minutes ago, porina said:

All core boost is usually much lower than single core max boost. Also do you know what power limit was enforced by the system? That can cut things back earlier compared to practically unlimited enthusiast mobos. Or even thermal limits.

Not a whole lot, actually. High performance in Windows, normal BIOS changes to let it run as far as it can. It was on a Gigabyte X99 board, so it's definitely not being held back by the C612 chipset on a server board.

[Mark Kaine]

1 hour ago, da na said:

Just... eight? 
Just eight cores?

Come on, we were doing eight core Xeons in 2010...

i am also super excited about new 8 core processors ...... in..... 2023........ 

[da na]

7 minutes ago, Mark Kaine said:

i am also super excited about new 8 core processors ...... in..... 2023........ 

I'd understand were these i5/i7 series but c'mon... huge step down from previous Xeon generations in terms of raw core count.

[alex75871]

4 minutes ago, da na said:

I'd understand were these i5/i7 series but c'mon... huge step down from previous Xeon generations in terms of raw core count.

 

12 minutes ago, Mark Kaine said:

i am also super excited about new 8 core processors ...... in..... 2023........ 

You do realize that those high-core-count Xeons use a completely different socket and platform, right? These are the desktop counterpart Xeons.

[da na]

17 minutes ago, alex75871 said:

 

You do realize that those high-core-count Xeons use a completely different socket and platform, right? These are the desktop counterpart Xeons.

Yes, I know that there have been lower-end Xeons on desktop socket since the 2010s

[Mark Kaine]

25 minutes ago, alex75871 said:

These are the desktop counterpart Xeons.

just doesn't seem to make a lot of sense... i appreciate they dropped the dreaded "e-cores" tho...

 

ps: basically this is the "budget variant" right? maybe there's a market for that, still super unexciting lol.

[leadeater]

4 hours ago, alex75871 said:

However, the top SKU features a base clock of 3.2 GHz and a boost of 5.6 GHz across all eight performance cores.

Ah sadly no, all core boost is not 5.6Ghz. It's for 2 cores but only if possible, thermal and application power load permitting. The 5.6Ghz is actually only guaranteed on a single core, thermals still permitting.

[SorryBella]

4 hours ago, RONOTHAN## said:

since I don't see many scenarios where it would make sense to buy one of these rather than a 13700K and disable the E-Cores.

Yep thats the issue, you can just disable the P side and get the same result. This is just bin cleaning done by Intel to me, so it being a Xeon makes more sense.

[leadeater]

4 hours ago, RONOTHAN## said:

The other thing to look at is how they also lost cache. For workstation tasks it's not likely to matter that much, but for games the 6MB less cache might be a bit of a performance hit when compared to a 13700K at the same frequency and E cores disabled, to the point where they'll probably be about equal. Besides, 5.6GHz on a 13700K is not that hard of an overclock if the E cores are disabled, so the boost clock matters even less. 

If you disable the E-Cores then the LLC/L3 cache also becomes unusable from what I would expect. The reduction of 6MB LLC/L3 cache is simply from not having the E-Core slices so there actually is no reduction and thus performance will not be impacted.

 

3 hours ago, RONOTHAN## said:

Yeah, that's the case. The thing is when you disable the E-cores, the P cores still have access to that cache (at least the L3 cache), so it is still worse overall. 

I don't think you can use that L3 if all E-Cores are turned off, could be wrong but I'd find that surprising. Not that it matters, fetching remote L3 cache is not optimal and it'll actually be migrated in to local L3 first anyway meaning that 6MB is of limited usefulness if usability is retained.

[leadeater]

3 hours ago, Dabombinable said:

Beyond AVX512 support (retroactively disabled in 12th gen due to scheduling problems - E-cores are a shitshow), will these actually prove that useful in servers?

In a small server for a small use case, yea probably. Flip side is most of that usage doesn't need AVX-512 either. The removal of E-Cores I think is just to widen compatibility for hypervisors and nothing else.