MikOnTopGGHmmm, they changed the naming... Again?

Marketing Manager: "our products have been doing poorly since we went to the 3-number system, what can we do?"
Marketing person 1: "we could go back to 4 numbers, but we've broken the sequence, so we'll have to start over"
Marketing person 2: "yeah, that's great, but we can't go back to 1000, these were supposed to be the "4th" gen of "core ultra", so we should just add a 0 to the 3-digit numbers we were going to use"
Marketing person 3: "aren't people sick of us changing the numbering system regularly? Shouldn't we just stick with what we have and focus on the product improvements?"
Marketing Manager: "1 and 2, I love this energy, great stuff! person 3...turn in your badge and gun and get the hell out of here"

JustBenchingCan't wait to see the name for the 52 core blc. Is it going to be VBFC? (Very Big Fu $)

It will he interesting to see how it's configured and if it even gets the extra $

This is fake. 6C listed as Ultra 9, model numbers do not even line up with core counts or capabilities.

Everyone is running out of numbers.

AMD and Intel should try naming them like cars with a trim level, something normal people can remember. Instead of using these cryptic SKU numbers out of a parts list.

freeagentSweet!

Moving back to Intel :)

Really? I'll take your am5 mobo for zen6 which should be a massive uplift from previous gen, wool platform support too.

docnorthTo be more precise, backwards compatible with Sandy Bridge... :D. This naming scheme is Haswell.


This means competitive, they have to surpass AMD on every sector, even on bad naming scheme:slap:.

Unless they get a 33% st uplift they aren't matching amd in gaming.

I appreciate they left LPE out of core count. Since those aimed at background OS tasks, not normally available for user apps, they should not count towards the number of cores any more than the neural processors do.

To be honest, there isn't really a way to make this lineup simple. You have models with 1 compute tile, 2 compute tiles, 1 compute tiles with BLLC, 2 compute tiles with BLLC, and a host of graphics options (if the rumored NVL-H APU is actually coming). For exmple, there will probably be F SKUs, 2 Xe core SKUs, 12 Xe core SKUs, and everything else. Intel has gone wild with the tiles and it's put up quite a challenge for the naming scheme (which was already not particularly beginner-friendly).

randomdude9182Really? I'll take your am5 mobo for zen6 which should be a massive uplift from previous gen, wool platform support too.

Why can’t I have both? 🤔

People dreaming zen6 beating nova lake when in reality watt for watt it won't even beat arrow lake. Mark my words, the 200 and 300$ zen 6 offerings are going to get pummeled by the 250k / 270k at iso watt.

RandallFlaggEveryone is running out of numbers.

AMD and Intel should try naming them like cars with a trim level, something normal people can remember. Instead of using these cryptic SKU numbers out of a parts list.

They tried with Core and Core Ultra. It didn't work.

AntiX3DGlazerTo be honest, there isn't really a way to make this lineup simple. You have models with 1 compute tile, 2 compute tiles, 1 compute tiles with BLLC, 2 compute tiles with BLLC, and a host of graphics options (if the rumored NVL-H APU is actually coming). For exmple, there will probably be F SKUs, 2 Xe core SKUs, 12 Xe core SKUs, and everything else. Intel has gone wild with the tiles and it's put up quite a challenge for the naming scheme (which was already not particularly beginner-friendly).

Yeah, modern CPUs are inherently complex. There's no sane way to map that complexity onto a series of numbers that you can easily order.
Maybe instead of model numbers, CPUs should get bar codes directly. You can scan those and get all the relevant info :P

bugYeah, modern CPUs are inherently complex. There's no sane way to map that complexity onto a series of numbers that you can easily order.
Maybe instead of model numbers, CPUs should get bar codes directly. You can scan those and get all the relevant info :P

They're not that complex. You have 1 or 2 chips (single tile / dual tile), and 1 or 2 variations of them (x number of cores fused off). That easily fits on a numerical scale from 0 to 9. Add another number for the generation, and you have two digits. Add a letter or another number for special bins, and that's 3 digits max. No big deal.

Most of the characters in modern CPU names is useless junk. The 0s and Xs for example. Or the Core "9" or Ryzen "9" number.

Now that is one stupid naming scheme if based on reality

AusWolfThey're not that complex. You have 1 or 2 chips (single tile / dual tile), and 1 or 2 variations of them (x number of cores fused off). That easily fits on a numerical scale from 0 to 9. Add another number for the generation, and you have two digits. Add a letter or another number for special bins, and that's 3 digits max. No big deal.

Most of the characters in modern CPU names is useless junk. The 0s and Xs for example. Or the Core "9" or Ryzen "9" number.

Yes, real easy. Now add in heterogeneous cores, low power versions/various TDP targets, different number of IGP cores. Gets pretty trivial real fast.

AusWolfThey're not that complex. You have 1 or 2 chips (single tile / dual tile), and 1 or 2 variations of them (x number of cores fused off). That easily fits on a numerical scale from 0 to 9. Add another number for the generation, and you have two digits. Add a letter or another number for special bins, and that's 3 digits max. No big deal.

Most of the characters in modern CPU names is useless junk. The 0s and Xs for example. Or the Core "9" or Ryzen "9" number.

I would not call the 3, 5, 7, 9 numbers "useless junk". They serve an important function which is to make sure the consumer knows vaguely what they're recieving. This is a function that it is generally successful at (you can find examples of it not working, but they tend to be the exception rather than the rule). 0s allow for incremented editions of parts (for example, if AMD wanted to release a higher-tier version of an already released processor or GPU, they can simply increment the 0 digit by a set amount-- RDNA 2 refresh is an example of this). To be honest, I can't really defend the Xs lol, they're pretty useless most of the time outside of brand virtue signalling.

randomdude9182Really? I'll take your am5 mobo for zen6 which should be a massive uplift from previous gen, wool platform support too.

I mean hey, just remember that new platforms have benefits too! Faster memory and far better expansion (Z890 is a vastly superior chipset in that regard to X870e or whatever AMD has rebranded Prom21 to this time). Generally, I agree that platform support is important to gamers and should be a priority, but it's important to remember that there are certian benefits.

By all indicators, they should be more or less ready to go.
Come on, release it already!

Darmok N JaladI guess it probably is to differentiate from Haswell when someone does a versus search

How's that gonna help if someone searches for Ivy Bridge-E parts?

Core i7-4960X, i7-4930K, and i7-4820K... aren't these going to look like they're in the same series if you don't take the i7 or the X into account?

ApocalypseeWell Intel why don't you bring HEDT platform back? Yeah you have absurd number of cores but bring back quad channel memory.

Intel's current HEDT is Granite Rapids-W e.g. 658X, it has octa channel memory and you want quad channel memory back???

bugYes, real easy. Now add in heterogeneous cores, low power versions/various TDP targets, different number of IGP cores. Gets pretty trivial real fast.

C = Large L3 cache (consistent with the i7-5775C having L4)
D = Dual compute tiles + 175 W TDP (ignore the fact that the i7-5500DU and i7-6498DU exists)
P = P Core-only (Alder Lake-P is gone, and don't they all feature LPE cores?)
G = 12 Xe3 cores (weren't they shelved for desktop?)
F = No IGP
S = Special edition (e.g. 12900KS, 13900KS, 14900KS)
K = 125 W TDP + Unlocked, 65 W specifically for the 6P Ultra 9 and 175 W for the dual compute tiles
T = 35 W TDP
E = Embedded
L = LGA package for U and H parts (e.g. 120HL, 120UL)
A = Who knows?



If the Processor 300, 310, and the 300T doesn't get a successor, then the Core Ultra 1 wouldn't or shouldn't exist. The last digit according to present Intel is usually a refresh marker like how it was done in Comet Lake-S (e.g. i3-10105F). Now it's either nonsensical (e.g. Core Ultra 5 230F), a refresh (e.g. Core 5 250K Plus), a marker for better graphics (e.g. Core Ultra 5 338H), or a marker for less processor cores and graphics cores (e.g. Core Ultra 5 322). I wonder how is that going to work out?

T550pC = Large L3 cache (consistent with the i7-5775C having L4)
D = Dual compute tiles + 175 W TDP (ignore the fact that the i7-5500DU and i7-6498DU exists)
P = P Core-only (Alder Lake-P is gone, and don't they all feature LPE cores?)
G = 12 Xe3 cores (weren't they shelved for desktop?)
F = No IGP
S = Special edition (e.g. 12900KS, 13900KS, 14900KS)
K = 125 W TDP + Unlocked, 65 W specifically for the 6P Ultra 9 and 175 W for the dual compute tiles
T = 35 W TDP
E = Embedded
L = LGA package for U and H parts (e.g. 120HL, 120UL)
A = Who knows?

Segment	Core config.	Compute tiles	L3 cache	IGP	TDP	Examples	Unlikely
Core Ultra 9	16P+32E+4LPE 8P+16E+4LPE 6P+12E+4LPE	Dual Single Single	264 MB/288 MB 36 MB/144 MB 108 MB	2C Xe3 GT1 2C Xe3 GT1 2C Xe3 GT1	175 W 65 W/125 W 65 W	495DKC, 495DKCF, 495DKCS, 495DKCG 485KC, 485KFC, 485KCG, 485K, 485KF, 485 475KC, 475KFC	495PDKCGES 485P, 485G, 485F 475PKC, 475KCG
Core Ultra 7	8P+12E+4LPE 4P+8E+4LPE	Single Single	33 MB/132 MB 18 MB	2C Xe3 GT1 2C Xe3 GT1	65 W/125 W 35 W/65 W	465KC, 465KFC, 465K, 465KF, 465 455, 455T, 455E, 455F, 455TE	465P, 465G, 465E 455TF, 455TEF
Core Ultra 5	6P+12E+4LPE 4P+4E+4LPE 4P+0E+4LPE	Single Single Single	27 MB 15 MB 12 MB	2C Xe3 GT1 2C Xe3 GT1 2C Xe3 GT1	65 W/125 W 35 W/65 W 35 W/65 W	445K, 445KF, 445 435, 435T, 435F, 435E, 435TE, 435A, 435TA 425, 425T, 425F, 425E, 425TE	445F, 445T, 445P 435TF, 435TEF 425EF, 425TF
Core Ultra 3	2P+0E+4LPE	Single	6 MB	2C Xe3 GT1	35 W/65 W	415, 415T, 415E, 415TE, 415EF, 415TEF	415F, 415EF
Core Ultra 1	1P+0E+4LPE	Single	3 MB	1C Xe3 GT1	35 W/65 W	405, 405T, 405E, 405TE, 405EF, 405TEF	405UL, 405HL

If the Processor 300, 310, and the 300T doesn't get a successor, then the Core Ultra 1 wouldn't or shouldn't exist. The last digit according to present Intel is usually a refresh marker like how it was done in Comet Lake-S (e.g. i3-10105F). Now it's either nonsensical (e.g. Core Ultra 5 230F), a refresh (e.g. Core 5 250K Plus), a marker for better graphics (e.g. Core Ultra 5 338H), or a marker for less processor cores and graphics cores (e.g. Core Ultra 5 322). I wonder how is that going to work out?

This is awesome.

T550pHow's that gonna help if someone searches for Ivy Bridge-E parts?

Core i7-4960X, i7-4930K, and i7-4820K... aren't these going to look like they're in the same series if you don't take the i7 or the X into account?

Yeah, it certainly could still be a mess. I know when I search for old Intel stuff, I just do the 4 digits and whatever letter follows. Just doing a 4 digit number can get you lots of random results, but just "4770K," Intel is the first hit.

T550pHow's that gonna help if someone searches for Ivy Bridge-E parts?

Core i7-4960X, i7-4930K, and i7-4820K... aren't these going to look like they're in the same series if you don't take the i7 or the X into account?

Intel's current HEDT is Granite Rapids-W e.g. 658X, it has octa channel memory and you want quad channel memory back???

C = Large L3 cache (consistent with the i7-5775C having L4)
D = Dual compute tiles + 175 W TDP (ignore the fact that the i7-5500DU and i7-6498DU exists)
P = P Core-only (Alder Lake-P is gone, and don't they all feature LPE cores?)
G = 12 Xe3 cores (weren't they shelved for desktop?)
F = No IGP
S = Special edition (e.g. 12900KS, 13900KS, 14900KS)
K = 125 W TDP + Unlocked, 65 W specifically for the 6P Ultra 9 and 175 W for the dual compute tiles
T = 35 W TDP
E = Embedded
L = LGA package for U and H parts (e.g. 120HL, 120UL)
A = Who knows?

Segment	Core config.	Compute tiles	L3 cache	IGP	TDP	Examples	Unlikely
Core Ultra 9	16P+32E+4LPE 8P+16E+4LPE 6P+12E+4LPE	Dual Single Single	264 MB/288 MB 36 MB/144 MB 108 MB	2C Xe3 GT1 2C Xe3 GT1 2C Xe3 GT1	175 W 65 W/125 W 65 W	495DKC, 495DKCF, 495DKCS, 495DKCG 485KC, 485KFC, 485KCG, 485K, 485KF, 485 475KC, 475KFC	495PDKCGES 485P, 485G, 485F 475PKC, 475KCG
Core Ultra 7	8P+12E+4LPE 4P+8E+4LPE	Single Single	33 MB/132 MB 18 MB	2C Xe3 GT1 2C Xe3 GT1	65 W/125 W 35 W/65 W	465KC, 465KFC, 465K, 465KF, 465 455, 455T, 455E, 455F, 455TE	465P, 465G, 465E 455TF, 455TEF
Core Ultra 5	6P+12E+4LPE 4P+4E+4LPE 4P+0E+4LPE	Single Single Single	27 MB 15 MB 12 MB	2C Xe3 GT1 2C Xe3 GT1 2C Xe3 GT1	65 W/125 W 35 W/65 W 35 W/65 W	445K, 445KF, 445 435, 435T, 435F, 435E, 435TE, 435A, 435TA 425, 425T, 425F, 425E, 425TE	445F, 445T, 445P 435TF, 435TEF 425EF, 425TF
Core Ultra 3	2P+0E+4LPE	Single	6 MB	2C Xe3 GT1	35 W/65 W	415, 415T, 415E, 415TE, 415EF, 415TEF	415F, 415EF
Core Ultra 1	1P+0E+4LPE	Single	3 MB	1C Xe3 GT1	35 W/65 W	405, 405T, 405E, 405TE, 405EF, 405TEF	405UL, 405HL

If the Processor 300, 310, and the 300T doesn't get a successor, then the Core Ultra 1 wouldn't or shouldn't exist. The last digit according to present Intel is usually a refresh marker like how it was done in Comet Lake-S (e.g. i3-10105F). Now it's either nonsensical (e.g. Core Ultra 5 230F), a refresh (e.g. Core 5 250K Plus), a marker for better graphics (e.g. Core Ultra 5 338H), or a marker for less processor cores and graphics cores (e.g. Core Ultra 5 322). I wonder how is that going to work out?

So simple. And you only have to memorize half the alphabet to tell which is which :P

On a more positive note, I could live with weird, long numbers, if they found a way to keep that for all their generations, not having to learn a new scheme every 5 years or so. If you add how many times these have changed and recall the changes were always driven by "simplicity" and "more clarity", you'd think a first grader can read these today effortlessly.

Ultimately, the model numbers have just become somewhat logical SKUs that work in marketing. The really oddball names are typically not found in everyday consumer products. These days, I'm searching the model name and then reading up on the specs. I stopped trusting the "higher number is better" approach some time ago, especially since AMD and Intel each go about it differently.

I hope Intel ditches E-cores and LPE-cores in its future CPUs. Desktop applications have always been poorly optimized. And the trend has always been for app optimization to get progressively worse for commercial reasons and to make programming easier.

That is why we home users only need CPUs with P-cores.

NhonhoI hope Intel ditches E-cores and LPE-cores in its future CPUs.

Unfortunately, that is not an option. Therefore, we need to deactivate them. This creates some distance between local hotspots and improves latency between the cores. The trend is towards a unified architecture that includes both smaller cores with lower clock speeds and larger cores with higher frequencies and Hyper-Threading support.

Just do first number(s) is generation, and second and third are sku plus a letter monacher at the end for stuff like unlocked or BLLC. And only change with big architecture changes like the start of Core.

Darmok N JaladYeah, it certainly could still be a mess. I know when I search for old Intel stuff, I just do the 4 digits and whatever letter follows. Just doing a 4 digit number can get you lots of random results, but just "4770K," Intel is the first hit.

Yeah there will probably be some scams with it. Like a PC saying it has a 4770k, someone buying it thinking it’s a great deal on the newest cp only to be stuck with a 10 year old piece of crap.

NhonhoI hope Intel ditches E-cores and LPE-cores in its future CPUs. Desktop applications have always been poorly optimized. And the trend has always been for app optimization to get progressively worse for commercial reasons and to make programming easier.

That is why we home users only need CPUs with P-cores.

Could you explain which program isn't running well? From my perspective, that's not the case, and the e core performance is being utilized very well.