Seven years later and I still don't get Intel's naming convention

[pantsaregood]

So, we're currently waiting for Broadwell-E supposedly. Skylake is out and Broadwell-E seems to still be pretty far off, yet they're expected to be labeled as part of the same generation.

Speaking of which, we all know that Intel's Core i-series CPUs are represented by a four-digit model number, right? Core i7-2600K implies a second-generation CPU and Core i7-6700K implies a sixth-generation CPU based on the first digit of the model number.

The Core i7-4960X and Core i7-4770K share the same "generation" number despite being built off of completely different architectures.

Also, since when does a die-shrink count as a generational leap? On the CPU front, Ivy Bridge and Broadwell were little more than die-shrinks, yet they were declared a new generation of CPUs.

Speaking of which, what happened to Westmere? Of all of the die-shrinks Intel has been through since 2008, Westmere would appear to be the one most deserving of being declared a new "generation." Westmere added AES instructions! Doesn't that count for a little more than Broadwell?

Come on Intel, get it together.

 

[Ken g6]

The Core i7-4960X and Core i7-4770K share the same "generation" number despite being built off of completely different architectures.

Yeah, they've kind of gotten out-of-sync. The big cores are something like 18 months behind now. :\

Also, since when does a die-shrink count as a generational leap? On the CPU front, Ivy Bridge and Broadwell were little more than die-shrinks, yet they were declared a new generation of CPUs.

I guess maybe since they started updating graphics on the die shrinks?

Westmere added AES instructions! Doesn't that count for a little more than Broadwell?

That's AES (a few specialized instructions for encryption/decryption), not AVX (a vector register twice as wide as SSE.)

 

[MrTeal]

It's a weird scheme. The naming could be that the HEDT parts might perform as well as the next generation of desktop parts, so Sandy HEDT gets a 3xxx part number like Ivy, but I would link it would make a lot more sense to just keep the same generation numbers and make an i9 or similar.

As for IVB and Broadwell, even though the IPC advancements are smaller than at a tock, the die shrink and uarch changes are enough that I think a major revision change makes sense.

 

[pantsaregood]

I'm aware that AES is not AVX, but AES is far more significant than the change from Sandy Bridge to Ivy Bridge or Haswell to Broadwell.

Ivy Bridge got a slightly faster memory controller. Broadwell got L4 cache. These aren't even modifications to the CPU core, unlike the addition of AES.

 

[MrTeal]

I'm aware that AES is not AVX, but AES is far more significant than the change from Sandy Bridge to Ivy Bridge or Haswell to Broadwell.

Ivy Bridge got a slightly faster memory controller. Broadwell got L4 cache. These aren't even modifications to the CPU core, unlike the addition of AES.

Ivy Bridge also moved from planer to 3D transistors, from a soldered heatspreader to a TIM interface, significantly improved the GPU, and had better performance/watt. It wasn't a big increase in IPC, but there was much more there than just a faster memory controller.

 

[alecmg]

Numbering scheme never made sense in the first place. Not when it was letter and 4 digits, then changed to 3 digits then back to 4 with iX in front.
At least at one point if name had Q in it it was quad core and E was dual core.
Now you could guess i3 was dual core with HT i5 was quad and i7 quad with HT. But no, the system gets destroyed by mobile and workstation chips that share same notation.
Forget getting any information on frequency or actual performance from the index.

 

[Arachnotronic]

Broadwell and Ivy Bridge brought CPU core uArch improvements over Haswell and Sandy Bridge, respectively. Westmere was a really boring Nehalem shrink w/ no uArch improvements aside from the inclusion of AES instructions.

 

[ninaholic37]

Seven years later and I still don't get Intel's naming convention

Come on Intel, get it together.

Sounds like you do 'get it', as in you understand it, but you're opposed to it. I was thinking this was going to be a Q&A thread, but it was just random complaining in disguise. Meh.


edit-

Core i7-2600K implies a second-generation CPU and Core i7-6700K implies a sixth-generation CPU based on the first digit of the model number.

I didn't even notice. Maybe just coincidence? Atom cores don't seem to do this.

 

[sm625]

The one good thing about intel's naming convention is that it makes it very easy to search on sites like slickdeals. If you search "i5" on slickdeals, 90% of the results are products that actually do contain one flavor or another of an intel i5. If you search for "A8" you will be lucky if 10% of the results are actually an AMD product, and of the ones that are, half or more are cat cores. lol. That's pretty bad. And now that some parts are labeled FX, it is even worse. FX is basically an unsearchable term.

 

[BigDaveX]

Broadwell and Ivy Bridge brought CPU core uArch improvements over Haswell and Sandy Bridge, respectively. Westmere was a really boring Nehalem shrink w/ no uArch improvements aside from the inclusion of AES instructions.

It wasn't a boring shrink for those who bought 6-core chips! :awe:

But yeah, outside of the highest-end chips Westmere never even really showed up on the desktop, unless you count the CPU cores in the hodgepodge first-gen i3 and i5 duallies.

 

[JM Popaleetus]

I think the HEDT parts should be moved to the "i9" nomenclature. Otherwise, the BMW-esque 3,5,7,etc. naming convention is pretty standard now, and it works for Intel.

Where things get confusing is within the series. I'm not sure there is a way to rectify that though. And honestly, AMD and Nvidia also suffer from SKU overload.

EDIT: Maybe go "German" all the way? E.g, i5-4x15HT? So Series-(Core Count)(Model)(Year)(Hyperthreaded/Unlocked/etc.) So the 6600K would be i5-4315K. The 6600, i5-4315HT. The 6500, i5-4215HT.

 

[ElFenix]

the problem with doing the same thing on mobile as you do on desktop is that 'best' mobile performance necessarily incorporates a battery life measure that desktop doesn't have. within a given tdp, an i7 should always have the highest benchmarks/most features. but, it may be the same benchmarks/features as an i3 at a higher TDP.



and the "german" method has been breaking down for over a decade.

 

[cytg111]

something about lawsuits, reverting to something that they no way in hell could be sued over...
I know Haswell is a US town with ~70 residents .. I still think its the coolest name they've chosen yet.

 

[cytg111]

Come on Intel, get it together.

You think they should accommodate some OCD like facist structure and stick to it or die trying? 🙂 .. Whatever it's spun as whaen we get it, you can be sure its in a frame thats supposed to somewhat maximize profit. Of course it is,- it'd be bananas if it wasnt.

 

[ninaholic37]

Whatever it's spun as whaen we get it, you can be sure its in a frame thats supposed to somewhat maximize profits. Of course it is,- it'd be bananas if it wasnt.

Yup. If they called i7-4960X an i7-3960X, it might seem more outdated at launch.