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

[lolfail9001]

They cared enough to rush the new enthusiast product line before AMD

Yes, their marketing does care to be the first, the largest, the newest et cetera et cetera. But the important part for them is to formally satisfy those adjectives, for money and ego. Providing product an enthusiast would value in every regard is not even on the list.

to announce a higher than 10 core part that was apparently not intended to ever be released on hedt to keep the highest core count and performance in their hedt offerings.

Technically 18 core part was always planned to be on Skylake-W if for nothing else but Apple's offerings. It's the same story as 1680v2 SKU.

Even so people should just be ok with them doing things like that?

Yes. You either agree to play on their terms buying their product or wait for something better [in one's eyes]. Capitalism (c).

 

[krumme]

Intel. Does. Not. Care. About. Enthusiasts.

It looks so. And they will have to learn again how to value their customers before shareholders and own agenda.

But please lets not get carried away by feelings here shall we. Its perfectly possible launch freq is 200MHz above when the cheap paste was planned. It would explain some of the horrific temp and throtling we see. Although the efficiency looks to be worse than bwe even at same low freq so its not that simple. But as said we need more test for that.

 

[formulav8]

You either agree to play on their terms buying their product

I'm talking about people upset whether they do or don't buy. Otherwise you're obviously correct. For instance some who wants their product is upset over their cheeping out on paste. I mean, you may have to accept it, but doesn't mean you have to like that.

 

[frozentundra123456]

Exactly, what are you doing in this thread?

It's not, Ryzen stack has set similar record before. In Ryzen's case inferior perf/watt costs more money though.


Intel. Does. Not. Care. About. Enthusiasts. I don't know why i have to repeat it every time everyone pretends to be baffled about Intel doing something enthusiasts can't like.

Don't feed him. Just use the ignore button or report threadcrapping posts.

 

[Ajay]

Intel. Does. Not. Care. About. Enthusiasts. I don't know why i have to repeat it every time everyone pretends to be baffled about Intel doing something enthusiasts can't like.

LOLZ! Missed this before. Yeah, SKX is really a server CPU - first and foremost. Intel just cares about our money - hence SKX and X299.

Speaking of cache, has it been revealed what the l2/l3 latency is compared to previous arch?

I saw some stuff on core-core pings (with no idea what the code was actually doing). I think allot of utilities were broken at first vis-a-vis latency, maybe they are fixed now (sort of what happened to Ryzen). Now you've got me wondering and I'll need to look around.

 

[formulav8]

Now you've got me wondering and I'll need to look around.

If you find out please post. I'm sure they increased L2 latency a bit (maybe a large bit even) but also could have reduced L3 latency a bit.

 

[tamz_msc]

6950X had 3.5Ghz Turbo 2.0 single core and 3.3Ghz all core, last time i checked.

The last time I checked, we were talking about 18C CPUs. And while you're at it use Tom's Hardware's Cinebench R15 power consumption vs frequency graph and extrapolate to 4 GHz, assuming linearity in the 4.2-4.4GHz range. The 7900X consumes just about it's rated TDP, which the Anandtech review rightly pointed out has never happened before.

 

[Grep_Linux]

The 6800 was always a very fine cpu.
Nobody just never realized it.

Actually, I don't know who didn't realize it? It was a great desktop experience when it came out. The clocks were a little lower but the price wasn't that bad and offered great multitasking and desktop experience as well as decent gaming.

Is there any specs on core voltage on these guys yet? Any chance of a undervolt?

 

[krumme]

Actually, I don't know who didn't realize it?

The customers. When i checked sale number for one of the biggest german retailers 2 months ago the r7 and especially 7700k was selling a magnitude more than the 6800.

 

[krumme]

The last time I checked, we were talking about 18C CPUs. And while you're at it use Tom's Hardware's Cinebench R15 power consumption vs frequency graph and extrapolate to 4 GHz, assuming linearity in the 4.2-4.4GHz range. The 7900X consumes just about it's rated TDP, which the Anandtech review rightly pointed out has never happened before.

Note: Cb r15 doesnt use avx as i know so the 7900x 15% perf uplift from 6950 at the cost of 50% extra power is not tied to avx load. And obviously neither is thg total power meassuring. Right?

 

[tamz_msc]

Note: Cb r15 doesnt use avx as i know so the 7900x 15% perf uplift from 6950 at the cost of 50% extra power is not tied to avx load. And obviously neither is thg total power meassuring. Right?

Yes, Cinebench R15 doesn't use AVX, and given the fact that THG's power consumption measuring equipment is pretty extensive, and the fact that they state this:

Power consumption is measured after the voltage converters and CPU, using points on the motherboard.

I'm inclined to believe that this is CPU-only power draw, even if they don't state it explicitly.

 

[tamz_msc]

Well what d'ya know, Computerbase results are out, they test three versions of Prime95, and with version 26.6, which is non-AVX, the results are(using NH-U12S with two fans)

6950X
177W(63C package, 58C core)
3.4GHz @1.07V

7900X
281W(98C package, 94C core)
4.0GHz@1.08V

Worse efficiency on every metric - Cinebench, Blender, h.264, h.265.

Those clocks obviously come at a price, and no matter how much process V-f curves you tout, it won't change this reality..

 

[DrMrLordX]

That . . . is a heavy price. I wish they had downclocked Skylake-X to 3.4 GHz for a better apples-to-apples comparison there.

 

[krumme]

That . . . is a heavy price. I wish they had downclocked Skylake-X to 3.4 GHz for a better apples-to-apples comparison there.

Yes or downclock it to a level where it matches 6950 in perf and then look at power usage and temp.

We could use perf vs power/temp curves as comparison here.

 

[MarkPost]

Well what d'ya know, Computerbase results are out, they test three versions of Prime95, and with version 26.6, which is non-AVX, the results are(using NH-U12S with two fans)

6950X
177W(63C package, 58C core)
3.4GHz @1.07V

7900X
281W(98C package, 94C core)
4.0GHz@1.08V

Worse efficiency on every metric - Cinebench, Blender, h.264, h.265.

Those clocks obviously come at a price, and no matter how much process V-f curves you tout, it won't change this reality..

Honestly this Skylake-X thing, is a bad joke

 

[ZGR]

Does SKY-X need significantly faster memory (DDR4-4000+) in benchmarks and games to pull ahead of its predecessor at similar clockspeeds? I gotta say, this power consumption problem is gonna be a big deal for a lot of builds out there.

I'm wondering if Intel should have waited with enlarging the L2 cache for 10nm or just have been more conservative with the clockspeeds.

Any benchmarks of the 7900X with higher speed DDR4 out there?

 

[JoeRambo]

That . . . is a heavy price. I wish they had downclocked Skylake-X to 3.4 GHz for a better apples-to-apples comparison there.

And even better, use delidded CPU to see 3.4Ghz and 3.4Ghz with undervolt, 4Ghz.

As forum giant Idontcare has shown us, delidding can have massive impact on voltages required and resulting power usage is less due to massively decreased leakage.

EDIT: ofc those Skylake-X results are nothing short of disaster. 4.5Ghz AVX offset 7700K @ ~1.3V is using just ~25w per IA core in same workload. So 10 cores definitely should not burn so much @4ghz and just 1.08V.

 

[JoeRambo]

Delete plz

 

[DrMrLordX]

And even better, use delidded CPU to see 3.4Ghz and 3.4Ghz with undervolt, 4Ghz.

As forum giant Idontcare has shown us, delidding can have massive impact on voltages required and resulting power usage is less due to massively decreased leakage.

Delidding Broadwell-E is a bit of a task, so getting an apples-to-apples comparison there could be a bit risky. Silicon Lottery doesn't even seem to sell delidded Broadwell-E as a legacy product. But if we could trust the solder on BDW-E to be roughly equal to a delid/relid + CLU (or similar) then that could be done.

EDIT: ofc those Skylake-X results are nothing short of disaster. 4.5Ghz AVX offset 7700K @ ~1.3V is using just ~25w per IA core in same workload. So 10 cores definitely should not burn so much @4ghz and just 1.08V.

You know I was thinking, why not compare Skylake-X's power consumption to Kabylake's on a core-per-core basis, but I didn't have good numbers for Kabylake's power numbers. That does put things in some perspective.

 

[Vaporizer]

Seems like their plans in reducing caches did not pan out as intended and now the have to move clocks up in order to be better than the broadwell E. What you get is a hot powerhog.
Reminds me of bulldozer vs phenom...

 

[.vodka]

Delidding Broadwell-E is a bit of a task, so getting an apples-to-apples comparison there could be a bit risky. Silicon Lottery doesn't even seem to sell delidded Broadwell-E as a legacy product. But if we could trust the solder on BDW-E to be roughly equal to a delid/relid + CLU (or similar) then that could be done.

Results start at around 12m15s. CLU is better by about 5°C than Intel's solder, at least for this particular 6950x. Delidded 7900x is definitely comparable to a soldered/stock 6950x.



This is a disaster. I don't see how they're going to fit the 12-18C models in this platform, are they going to include a 1000W titanium class PSU with these once average joe decides he wants to overclock his shiny new CPU? How are they even going to cool 18 of these toasty cores under the slightest load? The 12-18C die is turned into locked Xeons for a reason.

That rumor of revised 2066 boards with stronger power capability and soldered CPUs incoming
seems more and more plausible by the minute.

 

[pantsaregood]

Hardware Unboxed managed to cool a Core i9-7900X down a bit by using a gigantic radiator in a custom loop. Prior to using this huge radiator, they were pulling about 420W from the wall at 4.6 GHz - post custom loop (and lower temperatures), they were down to about 370W from the wall at 4.7 GHz. That's somewhere around 40W that the CPU's power consumption actually dropped.

Sounds like a huge issue with leakage. Delidding may actually curb the power consumption significantly due to dropping temperatures. I'm not entirely sure how Intel plans to push 12C-18C parts with this sort of thermal performance. Throwing larger coolers at this issue is relatively ineffective due to the fact that the coolers themselves aren't generally the limiting factor - the poor contact between the die and IHS is.

 

[Vaporizer]

It really seems like a well balanced product which is worth a few bucks (1000$)

 

[TheGiant]

Hardware Unboxed managed to cool a Core i9-7900X down a bit by using a gigantic radiator in a custom loop. Prior to using this huge radiator, they were pulling about 420W from the wall at 4.6 GHz - post custom loop (and lower temperatures), they were down to about 370W from the wall at 4.7 GHz. That's somewhere around 40W that the CPU's power consumption actually dropped.

Sounds like a huge issue with leakage. Delidding may actually curb the power consumption significantly due to dropping temperatures. I'm not entirely sure how Intel plans to push 12C-18C parts with this sort of thermal performance. Throwing larger coolers at this issue is relatively ineffective due to the fact that the coolers themselves aren't generally the limiting factor - the poor contact between the die and IHS is.

This is what I was looking for. So there is a fine chance that a properly cooled delidded SKL-X will have an acceptable power consumption.
I wonder what is the thermal performance of 28 core Skylake Xeon

 

[ManyThreads]

TechSpot got the 7900X to 73-74C maximum under load with a larger AIO cooler at 4.7 GHz which gave about a 15% performance increase. That seemed pretty good to me considering the less than ideal IHS:

https://www.techspot.com/review/1437-overclocking-core-i9/

Previously, I was only able to stabilize the 7900X at 4.6GHz, but with the upgraded cooler I managed 4.7GHz and 4.8 even looked to be possible with a bit more voltage tuning. At 4.7GHz the CPU spiked to 60 degrees as soon as it was placed under load -- much better than the 87 degrees seen before.

I expected temps to continue rising rapidly but to my surprise they didn't, at least not to the extent that I had anticipated. Before long we were in the mid 60s, then 70 degrees, but soon the system maxed out at 73 degrees, occasionally peaking at 74 degrees for a second before returning to 73 degrees.

With temps staying well within acceptable parameters there was no risk of throttling and Thermaltake's 120mm fans even remained quiet. As a result, power consumption was also reduced by 7%, down from 402 watts previously at 4.6 GHz with the AIO liquid cooler to 375 watts. Granted, that still makes the 7900X super power hungry, but it's a better result for a 10-core CPU running all cores at 4.7GHz.

I'm hoping the 7820K can do the same with a slightly smaller cooler or a NH-D15 if I'm lucky. Also sometimes I wonder how consistent reviewers are with their cooler install, thermal paste application, etc. but who knows.