Who's buying Skylake-X? (You may now change your vote)

[Hail The Brain Slug]

It looks like there isn't enough room on that board physically for 6 DIMMs per socket. If you notice, that is an ATX form factor board, which is rather compact for a dual socket server.

It's definitely a space constraint. Those sockets are gigantic, I'm surprised they fit two on an ATX board to begin with.

 

[Valantar]

So Hexus.net is reporting a SiSoft Sandra database listing of an "i7" (probably a software misidentification) 7900X with some interesting differences and similarities to previous rumors:

• 4GHz base clock - boost remains 4.5 like previous leaks/rumors
• 175W TDP. Uh oh?
• Confirmation of 13.75MB L3, 1MB L2 per core

I have to admit I would have been surprised if Intel managed to squeeze the 12c24t monster into a 140W TDP. I suppose a 500MHz bump in both base and boost clocks on the same process (or does SKL-X use 14nm+?) when compared to the 6950X ought to drive up power consumption quite a bit too. Still, a 10c20t CPU that's as fast per core as a 7700K? Hot damn.

 

[blue11]

So Hexus.net is reporting a SiSoft Sandra database listing of an "i7" (probably a software misidentification) 7900X with some interesting differences and similarities to previous rumors:

• 4GHz base clock - boost remains 4.5 like previous leaks/rumors
• 175W TDP. Uh oh?
• Confirmation of 13.75MB L3, 1MB L2 per core

I have to admit I would have been surprised if Intel managed to squeeze the 12c24t monster into a 140W TDP. I suppose a 500MHz bump in both base and boost clocks on the same process (or does SKL-X use 14nm+?) when compared to the 6950X ought to drive up power consumption quite a bit too. Still, a 10c20t CPU that's as fast per core as a 7700K? Hot damn.

These figures are all BIOS readings. They are probably parameters set for a test run. In particular, the "TDP" (not labeled as such on SiSoft) probably refers to the BIOS long-term power limit, which throttles the CPU if it draws too much power.

 

[LTC8K6]

That Sisoft listing for SL-X was reported a few days ago in the other thread here.
The overclocked 6900 chips report 140W in Sisoft, which is probably not accurate.
Not sure 175W is anything at all to be concerned about, even if we assume it's accurate.

 

[mikk]

These figures are all BIOS readings. They are probably parameters set for a test run. In particular, the "TDP" (not labeled as such on SiSoft) probably refers to the BIOS long-term power limit, which throttles the CPU if it draws too much power.

Yes exactly. The Bios might be tweaked to allow for higher TDP than Intels. It's nothing special about that on a OC board for desktop. And TDP is not even meaningful without a power consumption test, even though power draw might be high in this test with AVX 512 support (if it supports AVX 512).

 

[LTC8K6]

https://twitter.com/WikiChip/status/863874789622022145

That has 160W for the 12C chip...

 

[TheF34RChannel]

https://twitter.com/WikiChip/status/863874789622022145

That has 160W for the 12C chip...

That makes more sense than 175W for the 10C (which was BIOS readings anyway as a few have mentioned earlier).

 

[aigomorla]

Taken from his tweeter....
https://twitter.com/WikiChip/status/863874789622022145

dang... 160W TDP.... that sucker is either gonna be a tiny nuclear furnace or have a obnoxiously large die.

 

[mikk]

https://twitter.com/WikiChip/status/863874789622022145

That has 160W for the 12C chip...

Nothing new there. It's the same leak/rumor we already know from two weeks ago.

 

[LTC8K6]

Nothing new there. It's the same leak/rumor we already know from two weeks ago.

I know. That tweet has the clocks wrong, anyway. But it was new to me, and I had not seen it posted here before.

There's no reason to think the Sisoft entry is fake, but it may be inaccurate due to the new chip.

Do we even know if SL-X is on 14nm+? I think I read that it is.

 

[tential]

Taken from his tweeter....
https://twitter.com/WikiChip/status/863874789622022145

dang... 160W TDP.... that sucker is either gonna be a tiny nuclear furnace or have a obnoxiously large die.

I want a 140W i9-7280X for a server that essentially emulates 2 i7s (or what used to be the i7 back when it was a 4c/8th processor). Getting 2 CPUs worth in a 160-140 W package is pretty good ya know?

I think that'd be fun. Then I throw 2 GPUs in it.

I just have no reason to do this.... I just want to. So I'll probably not get it this generation.

As long as my i7-4770k delivers good framerates at 4K gaming I'm ok. Second that fails, I will probably get the best HEDT chip I can just to do stupid things with the cores.

My real server will be an AMD processor. Because that thing will probably be for the general purpose machines. Weak sauce. I pity whatever poor soul has to watch a movie using one of those cores. O wait that will be me probably at some point.... still........ Always great to use an i7.

i7, you can replace it when you want, it's always good.

 

[TheF34RChannel]

Posted this in the other thread as well, but it fits here too and finally allowed me to cast my vote.

I'd promised myself that when CFL-S required a new board, I'd go with SKL-X instead. Either upgrade would require the same number of parts to be purchased (except RAM) so it makes sense to go with the better platform.

 

[moonbogg]

If that 8 core comes in at $500 and can hit 4.6ghz, I will suffer from temptation. I can wait though. I'm not in the mood to drop $900 on something that won't provide any gaming benefit over what I got. A faster GPU would tug at me a lot harder though. There's always room for more GPU power, especially at my res and refresh rate. However, I am eager for more info from computex. If Intel really wants to rain on AMD's parade, they will come in with some offers that the enthusiast simply can't refuse, and in doing that they will prevent AMD from gaining a significant foothold in the market. At this point its just wait and see. I still can't see myself dumping my platform for Skylake. I don't really envy 7700K performance, so I don't think I'll be bothered too much by Sky-X performance. A 6 core at $300 though? That would complicate things for me.

 

[coercitiv]

or does SKL-X use 14nm+?

SKL-X is on 14nm+.

 

[blue11]

It was revealed on the other thread that Skylake-X will in fact go to 18 cores, which would require the XCC die. We will have to wait and see if this means the top Skylake-X SKUs will support AVX-512. On the other hand, with only quad-channel memory, it will probably be bandwidth-limited. I am still leaning towards Skylake-SP/LGA-3647, particularly if the entry-level XCC die can be obtained for around $2000.

 

[Sweepr]

It was revealed on the other thread that Skylake-X will in fact go to 18 cores, which would require the XCC die. We will have to wait and see if this means the top Skylake-X SKUs will support AVX-512. On the other hand, with only quad-channel memory, it will probably be bandwidth-limited. I am still leaning towards Skylake-SP/LGA-3647, particularly if the entry-level XCC die can be obtained for around $2000.

LCC is 10C, MCC is 18C, HCC is 28C. They are using the MCC die for Core i9-7980XE.

 

[wildhorse2k]

On the other hand, with only quad-channel memory, it will probably be bandwidth-limited. I am still leaning towards Skylake-SP/LGA-3647, particularly if the entry-level XCC die can be obtained for around $2000.

Should be no big deal since Threadripper will also be only quad-channel. On the low end - 6/8 core it will probably be very difficult to fully utilize the quad-channel memory controller.

Only a fool could think Intel would let Threadripper go unchallenged and destroy its HEDT market.

 

[krumme]

Should be no big deal since Threadripper will also be only quad-channel. On the low end - 6/8 core it will probably be very difficult to fully utilize the quad-channel memory controller.

Only a fool could think Intel would let Threadripper go unchallenged and destroy its HEDT market.

Well now they have to fight a 32 core for the same market with a 18c die. Dont know if it hadnt been better with a 10c vs 16c. They entered a gunfight with a knife.

 

[IEC]

Well now they have to fight a 32 core for the same market with a 18c die. Dont know if it hadnt been better with a 10c vs 16c. They entered a gunfight with a knife.

It would be 18c/36t versus 16c/32t (top "Threadripper") chip. Naples/Epyc is the chip with 32c/64t but is meant for 1P and 2P servers.

 

[moonbogg]

I think the cores battle has gone from interesting to ridiculous in an instant. 8 cores was already a lot, then AMD shocked with 16, and Intel said NUT UH with 18 cores. Amazing how just a few weeks of competition has so drastically changed the entire CPU landscape. It was very recent that 8 cores was a big uplift and cost a grand. Now it only feels like a small step up from mainstream. Crazy. If Intel charges over $500 for their 8 core CPU, I'll just roll my eyes and move on.

 

[lolfail9001]

If Intel charges over $500 for their 8 core CPU, I'll just roll my eyes and move on.

Why should Intel charge less for their 8 core than AMD? It is their 8 core that is bound to be faster, you know.

 

[Shivansps]

I think the cores battle has gone from interesting to ridiculous in an instant. 8 cores was already a lot, then AMD shocked with 16, and Intel said NUT UH with 18 cores. Amazing how just a few weeks of competition has so drastically changed the entire CPU landscape. It was very recent that 8 cores was a big uplift and cost a grand. Now it only feels like a small step up from mainstream. Crazy. If Intel charges over $500 for their 8 core CPU, I'll just roll my eyes and move on.

AMD has nothing to do with ANY of this. In the past, X99 supported all Xeon E5 lineup, you can place a 22C Xeon E5-2699 v4 on ANY X99 2011-3 motherboard. Seriusly you belive Intel reacted to a yet unlaunched product this quickly?

 

[krumme]

It would be 18c/36t versus 16c/32t (top "Threadripper") chip. Naples/Epyc is the chip with 32c/64t but is meant for 1P and 2P servers.

Nope. I will eat my hat if tr16 and skl 18c have remotely the same price.

Everything above i7 7700k and 1800x is server rip off tech. I dont care if its labelled threadripper epyc i9 or xeon. Same tech with slightly different bins and tdp without some functionality.
Its about price vs perf. An 18c is probably competing with 24/32c or Intel is in trouble and thats not likely.

I find it great what is happening. Soon 6c is the new 2c pentium/i3 class.
Why not? Its like the 775 core 2 penryn 5200/7200.

Amd can take their 1800x and cut its 25% as a starter this summer. I wouldnt even dare talk about the pathetic i5 4 thread stone age line.

Moonbogg is right that the moar cores race is crazy but what is more insane is the zero development the last 6 years.

There is a lot to catch up. Intel reacted like they should. Now its next step.

 

[DrMrLordX]

AMD has nothing to do with ANY of this.

I think AMD did provoke Intel to release server parts on X299 with high(er) core counts to tamp down any enthusiasm over Threadripper. I don't think AMD has goaded Intel into lowering prices. Yet. Give it time.

 

[Shivansps]

I think AMD did provoke Intel to release server parts on X299 with high(er) core counts to tamp down any enthusiasm over Threadripper. I don't think AMD has goaded Intel into lowering prices. Yet. Give it time.

Well, unless the 14, 16 and 18C parts are released in like Q2 18 then i dont think so, there is just not enoght time for Intel to react to any AMD actions this quickly, this had to be planned a long time ago. Is like the KBL-X, makes 0 sence with Ryzen and CFL-S around the corner, yet, they were unable to cancel it. Intel response to AMD was to speed up SKL-X and CFL-S, thats all.

The 14, 16 and 18C core parts are not a big shocker to me as Server SKU are now segmented intro a expensive socket, X99 used to offer a option to have high core counts in a client motherboard, the 18C SKL-X is just a continuation of that.