Intel launches Kaby Lake Xeons

[Armakar]

Am I missing something? Why is a 4c/8t cpu with a higher tdp and lower clockspeed double the price of the 7700K?

[raphidy]

Just now, Armakar said:

Am I missing something? Why is a 4c/8t cpu with a higher tdp and lower clockspeed double the price of the 7700K?

They are server parts, xeon.

[Armakar]

Just now, raphidy said:

They are server parts, xeon.

sooo they have higher IPC or what? what makes them better for server use if they have the same threads, better platform?

[mikat]

2 hours ago, Jacobsheeler said:

Can someone explain why the 1280 is 600$?

Probably because it's the single best non overclocked single core xeon performance king?

Server market is crazy man, but I could be wrong because even for huge companies a 100 mhz increase for almost double seems a bit off

[mikat]

2 minutes ago, Armakar said:

sooo they have higher IPC or what? what makes them better for server use if they have the same threads, better platform?

ECC memory

[NumLock21]

6 minutes ago, zMeul said:

and?! since when Micron started manufacturing 3DXpoint ? 2015 ... coincidence?! 

in August Intel announced Optane: http://www.anandtech.com/show/9541/intel-announces-optane-storage-brand-for-3d-xpoint-products

Intel can disable or enable features when they feel like it.

[MandelFrac]

3 hours ago, Princess Cadence said:

While I agree with it, wouldn't you also trade the iGPU even for just an extra 1c/2t on the 7700k?

Nope, I'll take the emergency failover graphics. Not to mention the iGPU is fully programmable, and it would be far more useful than more CPU cores for getting the heavier physics/compute tasks done CPU side before the results are handed off to the dGPU.

[MandelFrac]

1 hour ago, NumLock21 said:

Intel can disable or enable features when they feel like it.

Depending on how the chipset was originally designed and how much extra EPROM there is. There's usually just enough to load microcode fixes and add new CPUID strings/configurations.

[SteveGrabowski0]

6 hours ago, Princess Cadence said:

Because 95% of i7 7700k owners have dedicated graphics cards?

So what? Aren't 7700k down to Celeron the same chips with different binning and different things disabled after the binning to segment into different products? If you want chips designed without an IGP wasting die space you have the enthusiast line.

[MandelFrac]

4 hours ago, Armakar said:

Am I missing something? Why is a 4c/8t cpu with a higher tdp and lower clockspeed double the price of the 7700K?

Lower TDP and lower clockspeed* also, it's because it's validated for servers and has ECC enabled for both the CPU and iGPU. Beyond this, if you look back historically and see benchmarks of xeons against their Core I counterparts, you'll find xeons edge them out consistently. Intel reduces the nominal clockspeed but also tends to tweak internal timings to get certain server-critical instructions ticking a bit faster. A CPU nominally running at 4GHz actually has some internal components running at clocks 5x that rate or more.

[MandelFrac]

5 minutes ago, SteveGrabowski0 said:

So what? Aren't 7700k down to Celeron the same chips with different binning and different things disabled after the binning to segment into different products? If you want chips designed without an IGP wasting die space you have the enthusiast line.

Celerons are always taken from I3 as I recall or are built from Atom processors. Has anyone seen a 4-core Celeron? Even if you tear an I5 of the same generation down to the same clocks and just use benchmarks that can fit in L1 and L2 cache, the performance difference can be very significant against a quad-core Pentium.

[SteveGrabowski0]

1 minute ago, MandelFrac said:

Celerons are always taken from I3 as I recall or are built from Atom processors. Has anyone seen a 4-core Celeron? Even if you tear an I5 of the same generation down to the same clocks and just use benchmarks that can fit in L1 and L2 cache, the performance difference can be very significant against a quad-core Pentium.

I don't think they pattern an entire wafer of 7700k and then an entire wafer of G4560 and so on though. AFAIK they're all the same chips and the ones that test best become i7-HQ mobile processors, the next best unlocked i7 desktop, maybe the ones that aren't efficient enough to hit that 65W TDP as i7 or i5 get two cores disabled so they can be sold as i3 or Pentium, etc.

[MandelFrac]

Just now, SteveGrabowski0 said:

I don't think they pattern an entire wafer of 7700k and then an entire wafer of G4560 and so on though. AFAIK they're all the same chips and the ones that test best become i7-HQ mobile processors, the next best unlocked i7 desktop, maybe the ones that aren't efficient enough to hit that 65W TDP as i7 or i5 get two cores disabled so they can be sold as i3 or Pentium, etc.

No, I3s are their own die. Plenty of delidding has been done since the Ivy Bridge days showing that. I3/Pentium/Celeron and the dual-core mobile I3/I5/I7 are all one die (apart from Atom-based Celerons). Mainstream Core I5/I7 and E3 Xeons are one die where the Xeons are tuned differently. HEDT shares a die with some E5/E7 Xeon variants, and then there's usually 1 more die between HEDT and the very biggest E5/E7 Xeons.

 

Intel per mainstream architecture generation makes 5 different dies. And then it comes to binning for efficiency and all that.

[Ezio Auditore]

7 hours ago, zMeul said:

what does X have to do with Y?

 

I was talking about Zen as a SoC and what I read is bunch of unrelated nonsense

 

Ryzens have full blown chipset (Zeppelin) in them, on die - that's the problem I can't get over

and who designed that chipset? fucking AsMedia .... the utter most shittiest chipset manufacturer out there, they produce shit SATA controllers and USB controllers

what happens if any of those has an issue that can't be fixed? you're fucked

I thought that by saying SoC you meant the double ccx issue. Does double ccx mean SoC? 

[MandelFrac]

13 minutes ago, Ezio Auditore said:

I thought that by saying SoC you meant the double ccx issue. Does double ccx mean SoC? 

No, it's an SOC because it has chipset functions on the processor itself. It's somewhat incomplete because there's no iGPU, but otherwise Ryzen is most of an SOC.

 

To be fair, this is why Intel actually coined the term MPU or Multi-Processor Unit because its CPUs, even before they got iGPUs, had taken on the northbridge and several utilities which used to be found as dedicated chips on the motherboard.

[leadeater]

5 hours ago, SteveGrabowski0 said:

I don't think they pattern an entire wafer of 7700k and then an entire wafer of G4560 and so on though. AFAIK they're all the same chips and the ones that test best become i7-HQ mobile processors, the next best unlocked i7 desktop, maybe the ones that aren't efficient enough to hit that 65W TDP as i7 or i5 get two cores disabled so they can be sold as i3 or Pentium, etc.

 

5 hours ago, MandelFrac said:

No, I3s are their own die. Plenty of delidding has been done since the Ivy Bridge days showing that. I3/Pentium/Celeron and the dual-core mobile I3/I5/I7 are all one die (apart from Atom-based Celerons). Mainstream Core I5/I7 and E3 Xeons are one die where the Xeons are tuned differently. HEDT shares a die with some E5/E7 Xeon variants, and then there's usually 1 more die between HEDT and the very biggest E5/E7 Xeons.

 

Intel per mainstream architecture generation makes 5 different dies. And then it comes to binning for efficiency and all that.

 

They used to be the same die, I think Intel started using a dedicated 2 core die at Haswell or after.

[MandelFrac]

8 minutes ago, leadeater said:

 

 

They used to be the same die, I think Intel started using a dedicated 2 core die at Haswell or after.

I'm going to bet Ivy based on memory, but we'd have to gather up die shots over the past few years.