Intel Comet Lake Thread

[Ajay]

Are you crazy enough to try overclocking one of those fireballs? If the answer is, "no", then congratulations. You pass the sanity test. OCing a 2990WX to 4 GHz (or less) is pretty pedestrian and sane compared to the W3175, despite the 2990WX having four more cores. Similarly, it's unlikely that Intel would try schlepping the W3175 (or something similar) to the HEDT space when their 18c offerings are already borderline. Hopefully Cascade Lake-X will offer an 18c that is a bit easier to manage. For Intel's sake.

Maybe 100-200MHz higher base and boosts, plus hardware mitigating for some (all?) of the CPU based security risks.

 

[DrMrLordX]

Maybe 100-200MHz higher base and boosts, plus hardware mitigating for some (all?) of the CPU based security risks.

Really they just need something better-binned that can reach X clockspeed without pulling any more power/producing any more heat. If they can squeeze another 200 MHz out of it then maybe someone, somewhere will upgrade. It'll still face a lot of stiff competition. I don't think I need to say from whom.

 

[Ajay]

Really they just need something better-binned that can reach X clockspeed without pulling any more power/producing any more heat. If they can squeeze another 200 MHz out of it then maybe someone, somewhere will upgrade. It'll still face a lot of stiff competition. I don't think I need to say from whom.

Well, in the DIY market I expect AMD to gain massive share. In the OEM market, it'll be a tough slog despite AMD's perf/$ advantages just because Intel is so entrenched on many levels (customers, supply chain and OEM support orientation, etc.).

 

[DrMrLordX]

Well, in the DIY market I expect AMD to gain massive share. In the OEM market, it'll be a tough slog despite AMD's perf/$ advantages just because Intel is so entrenched on many levels (customers, supply chain and OEM support orientation, etc.).

How big is the Intel HEDT OEM market? Most of them are hype-driven, so if someone has an objectively better product getting a lot of attention, you better believe that Alienware et al will have . . . something else available.

 

[aigomorla]

Really they just need something better-binned that can reach X clockspeed without pulling any more power/producing any more heat.

Thats really not possible...
the faster a clock is on a processor the more heat it will deliver.
Your chasing your tail for a better analogy.

Even if its better bin'd, you will still hit thermal wall as you ramp the processor up, and with that many cores its compounded by 28.
The easiest yet stupidest solution to it, is the make the die larger, so you can spread the heat out over a larger surface area, and have a better cooling solution.

Or you can invent unobtainum which has a massive impossible C/W ratio, and is also a super conductor and make a die off that.

 

[DrMrLordX]

Thats really not possible...

Intel has been doing it with 14nm for awhile. There are limits. I'm pretty sure they can make a few tweaks to their 18c HEDT chips. But I honestly do not expect 28c from them in the HEDT space for the reasons you articulated. The heat is just too ridiculous for anything other than Xeon-like clockspeeds.

That's why it hurts so much for Intel. Any time some apologist asks, "does the market really want 10nm?" the answer is "YES". Intel can't do anything meaningful to update their CPU lineup. Intel has already refreshed Skylake-X once. Cascade Lake-X is going to look just like Skylake-X with some clockspeed adjustments.

 

[aigomorla]

Intel has been doing it with 14nm for awhile. There are limits. I'm pretty sure they can make a few tweaks to their 18c HEDT chips. But I honestly do not expect 28c from them in the HEDT space for the reasons you articulated. The heat is just too ridiculous for anything other than Xeon-like clockspeeds.

well i have stated this many times...
id honestly rather want a 14c x 2 platform then a single 28c.

Its the stupid core wars i tell you.

Another thing i find halarious is how AMD called intel Core 2 Quad a Fake Quadcore because of how it was essentially 2 dies on 1 pcb.
Yet Liza did the same exact thing, and intel is now calling TR on that.
Sigh...feels like they are both little kids pointing fingers at each other going
"your it.. no your it.. no your it.. "

 

[DrMrLordX]

id honestly rather want a 14c x 2 platform then a single 28c.

Be careful what you ask for. You might get Cascade Lake-AP. That's not socket-compatible with the current HEDT lineup though, and it has some other issues.

 

[BigDaveX]

Another thing i find halarious is how AMD called intel Core 2 Quad a Fake Quadcore because of how it was essentially 2 dies on 1 pcb.
Yet Liza did the same exact thing, and intel is now calling TR on that.

Not really a valid comparison, since Rome/TR3 - and for that matter, all of AMD's MCM processors with the possible exception of the early Socket G34 Opterons - have been designed from the ground-up for that kind of configuration, whereas Intel's 2000s-era MCM chips had clearly been hastily stitched together to head off competition from AMD.

That said, there were probably some sour grapes from AMD in those comments, seeing how at the time the original Phenom was well behind the Core 2 Quad on the desktop.

 

[Ajay]

How big is the Intel HEDT OEM market? Most of them are hype-driven, so if someone has an objectively better product getting a lot of attention, you better believe that Alienware et al will have . . . something else available.

Interesting question. I don't know, but they are probably cheaper than W series workstations.

 

[DrMrLordX]

Interesting question. I don't know, but they are probably cheaper than W series workstations.

Heh! Just about anything outside of server-series Xeons are cheaper than that thing.

 

[jpiniero]

https://twitter.com/x/status/1160935769965355011

Looks like Comet Lake U is going to get announced today. Lenovo jumped the gun it looks.

 

[repoman27]

Well, Comet Lake U 6+2 is here. And oh look, Amber Lake Y 4+2 finally turned up, but Intel wants us to call it Comet Lake Y now.

I guess this partly explains why AML-Y 2+2 was such an odd duck. It would appear that Amber Lake was supposed to be the 7W Y version of the 14nm WHL-U 4+2 die paired with the 22nm SPT-LP PCH in an FCBGA1377 package. For whatever reason (e.g. 14nm capacity constraints, difficulty binning 7W 4+2 parts, packaging issues, trying to clear out existing inventory) Intel simply released a KBL-Y 2+2 refresh with higher TDP and clocks. Now we get the real AML-Y 4+2 (including SIPP SKUs) on the SIPP release schedule, but Intel is calling it Comet Lake Y. Except CML-Y was already on the roadmaps, and ostensibly it was a similar 4+2 die and BGA1377 package but with a 14nm 400 series CMP-LP PCH. It seems like that's unlikely to see the light of day now, so this version of Comet Lake Y will just have to hang in there with nothing better than 5 Gbps USB 3.0 through most of 2020. <shakes head>

Meanwhile, Comet Lake U appears to be coming through with LPDDR4X support out of the gate, which is a pleasant surprise indeed. Also, I was expecting the base TDP to get bumped to 25W for the 6C, so kudos to Intel for even claiming 15W. Although I'm guessing we should probably expect a lot of 10th Gen 15W U-Series processors, both ICL and CML, to ship in 25W cTDP-up mode.

 

[EXCellR8]

I'm really having a hard time getting past this "iX-12345" 5-digit naming scheme... seems like they could have gone another more abbreviated direction but you know, Intel.

 

[jpiniero]

I'm really having a hard time getting past this "iX-12345" 5-digit naming scheme... seems like they could have gone another more abbreviated direction but you know, Intel.

Guessing Marketing wanted the 10 in the model number. So we should expect Tiger Lake/Rocket Lake to be 11xxx.

 

[EXCellR8]

Yea I get why, just looks and reads a tad awkward to me. Instead of saying "intel i7 twenty-five hundred" we have to say "intel iX ten twenty-five hundred", respectively.

I guess I could get used to it, and probably won't have a choice, but I've always pondered how companies continue SKU naming in the past. Adding a digit just seems like a worst case scenario lol

 

[ehume]

It's hard to read. Better if they had picked iX 10-yyy instead of iX 10yyy.

 

[Amol S.]

Is Intel no longer going to make Core-i7 X500U or X550U series? Last one for the X500U was the 7500U and for the X550U was the 8550U.

 

[jpiniero]

Is Intel no longer going to make Core-i7 X500U or X550U series? Last one for the X500U was the 7500U and for the X550U was the 8550U.

Comet Lake is 10xxxU. They just skipped 9xxxU.

 

[ehume]

Comet Lake is 10xxxU. They just skipped 9xxxU.

. . . and they should have put a hyphen after the generation# Five digits is too many.

 

[Amol S.]

Comet Lake is 10xxxU. They just skipped 9xxxU.

X in my post meant any number 1 digit or more. But ok i7 XX500U or XX550U?

 

[repoman27]

I'm going to go ahead and take back what I said previously about CML-U having LPDDR4 support out of the gate. It clearly doesn't support it right now, even if the silicon might. It looks like we probably won't see devices with LPDDR4 until Q2 2020 after all.

Looking at shipping / announced / leaked Comet Lake products, there's no LPDDR4/X anywhere in the mix:

Dell
Inspiron 14 3000 (3490) DDR4-2666 / Optane
Inspiron 15 3000 (3590) DDR4-2666 / Optane
Inspiron 17 3000 (3790) DDR4-2666 / Optane
Inspiron 13 5000 (5391) LPDDR3-2133
Inspiron 14 5000 (5494) DDR4-2666 / Optane
Inspiron 15 5000 (5594) DDR4-2666 / Optane
Inspiron 13 7000 (7391) LPDDR3-2133
Inspiron 13 7000 2-in-1 (7391 2n1) LPDDR3-2133
Inspiron 15 7000 2-in-1 (7591 2n1) DDR4-2666
Inspiron 17 7000 2-in-1 (7791 2n1) DDR4-2666
Vostro 13 5000 (5391) LPDDR3-2133
XPS 13 (7390) LPDDR3-2133

HP
ENVY - 13t (6VC07AV) DDR4 (memory down)
ENVY - 17t (6VC02AV) DDR4-2666
ENVY x360 - 15t (6WW67AV) DDR4-2666
Pavilion x360 - 14t (6XL50AV) DDR4-2666
Pavilion x360 - 15t (6XJ11AV) DDR4-2666

Intel
NUC “Frost Canyon i3” (NUC9i3FNH/K) DDR4-2666
NUC “Frost Canyon i5” (NUC9i5FNH/K) DDR4-2666
NUC “Frost Canyon i7” (NUC9i7FNH) DDR4-2666

Lenovo
ThinkPad X1 Carbon Gen 7 LPDDR3
ThinkPad X1 Yoga Gen 4 LPDDR3
ThinkPad X390
ThinkPad T490

MSI
Modern 14 DDR4-2666
Prestige 14 LPDDR3-2133
Prestige 15 DDR4-2666

 

[NTMBK]

I'm going to go ahead and take back what I said previously about CML-U having LPDDR4 support out of the gate. It clearly doesn't support it right now, even if the silicon might. It looks like we probably won't see devices with LPDDR4 until Q2 2020 after all.

Looking at shipping / announced / leaked Comet Lake products, there's no LPDDR4/X anywhere in the mix:

Dell
Inspiron 14 3000 (3490) DDR4-2666 / Optane
Inspiron 15 3000 (3590) DDR4-2666 / Optane
Inspiron 17 3000 (3790) DDR4-2666 / Optane
Inspiron 13 5000 (5391) LPDDR3-2133
Inspiron 14 5000 (5494) DDR4-2666 / Optane
Inspiron 15 5000 (5594) DDR4-2666 / Optane
Inspiron 13 7000 (7391) LPDDR3-2133
Inspiron 13 7000 2-in-1 (7391 2n1) LPDDR3-2133
Inspiron 15 7000 2-in-1 (7591 2n1) DDR4-2666
Inspiron 17 7000 2-in-1 (7791 2n1) DDR4-2666
Vostro 13 5000 (5391) LPDDR3-2133
XPS 13 (7390) LPDDR3-2133

HP
ENVY - 13t (6VC07AV) DDR4 (memory down)
ENVY - 17t (6VC02AV) DDR4-2666
ENVY x360 - 15t (6WW67AV) DDR4-2666
Pavilion x360 - 14t (6XL50AV) DDR4-2666
Pavilion x360 - 15t (6XJ11AV) DDR4-2666

Intel
NUC “Frost Canyon i3” (NUC9i3FNH/K) DDR4-2666
NUC “Frost Canyon i5” (NUC9i5FNH/K) DDR4-2666
NUC “Frost Canyon i7” (NUC9i7FNH) DDR4-2666

Lenovo
ThinkPad X1 Carbon Gen 7 LPDDR3
ThinkPad X1 Yoga Gen 4 LPDDR3
ThinkPad X390
ThinkPad T490

MSI
Modern 14 DDR4-2666
Prestige 14 LPDDR3-2133
Prestige 15 DDR4-2666

It might have LPDDR4X support, but that no doubt requires a redesign of the motherboard to support it. Whereas these models have no doubt just taken their existing Whisky Lake boards and slapped in a new CPU.

 

[repoman27]

It might have LPDDR4X support, but that no doubt requires a redesign of the motherboard to support it. Whereas these models have no doubt just taken their existing Whisky Lake boards and slapped in a new CPU.

Using DDR4 SODIMMs is a completely legitimate design decision, and Comet Lake-Y (a.k.a. Amber Lake) only supports up to LPDDR3-2133 anyway, but I think it's pretty clear that zero announced products with LPDDR4/X at this point means that it's highly unlikely that we'll see any before April 2020. I have a hard time believing that every OEM was willing to forgo the power and performance benefits of LPDDR4/X in memory down configurations entirely out of complacency.

All of the leaks pointed to the LPDDR4 variants of Comet Lake arriving later, and ARK still doesn't list the current Comet Lake parts as supporting it. The official Comet Lake product brief is also pretty cagey on the topic. It seems far more likely that either these parts really don't support LPDDR4, or they are still awaiting platform validation for use with LPDDR4/X-2933. It looks a lot like Intel is intentionally misleading the media, and by extension customers, in regards to the true capabilities of Comet Lake at this time.

 

[jpiniero]

My feeling is that the LPDDR4(X) support isn't there until an updated stepping which won't be available until like March.