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

[Ajay]

Ahh. Its like amd bios switch in the 290 gpu. They cant have high perf and low noise at the same time. You have to choose.
The meaning of this cpu is very much connected to the avx perf. You dont gimp on it. Then you are better of with either a cl for gaming or tr for productivity.
It is 4.4 tops.
All this water cooling is crazy for this kind of perf imo. And then on a cpu with this tim issue. Stupid. Like a racecar with 14 inch tires. Makes no sense.

One would only need custom water for more extreme overclocks, and with a delidded CPU an AIO would be sufficient. The delidding part is unfortunate and was a bone head move by Intel, the negative press can't have been worth it. IIRC, x299 BIOSes allow one to set an AVX offset to reduce clock speeds when running said loads, for those few who will need it (and most who really need it won't be doing extreme overclocks). HEDT has never been about the best gaming performance, it's about the best overall performance particularly in multithreaded tasks. X299 is for those who want something closer to a professional system, while still having excellent gaming performance and for those who want a real challenge when it comes to overclocking and benchmarks. Things are different with AMD back in the game - Intel has really been unchallenged in this space for a long time. It's silly to bash those who want X299 for its real and perceived flaws just as it was pointless to bash those who built AMD systems over the past five years. People want what they want, impugning others motives is pure folly.

 

[pantsaregood]

Has anyone seen power consumption for a delidded Skylake-X running at more modest temperatures? I'm wondering if the ridiculous power consumption is a result of leakage caused by the high temperatures - which are caused by the IHS' poor contact to the die.

 

[crashtech]

Has Intel abandoned indium solder altogether, or do LGA 3647 CPUs still get the old method?

 

[lolfail9001]

Largest Broadwell-E Xeon die is around 450 sq. mm, and that was for 16+ cores clocked at 2.6GHz or below. Their rated TDP is for base clock operation in an "Intel-specified" workload, who knows what that means. So it's no surprise that they can claim to 'satisfy' their TDP target with these requirements.

Once again, Intel had a semi-custom 20C 200W TDP SKU with 3.2Ghz base clock. Since SKL-X visibly has better voltage/freq curve than BDW-E, i would be very surprised if they did not manage to fit 18C 3-3.2Ghz (their typical base clock for highest end SKU in the past) into 165W, especially with AVX-512 offset in play.

The meaning of this cpu is very much connected to the avx perf.

It is not, really. And please, do not bring up TR until you see how it performs with software that does not know how to act with dual socket systems.

I would like to see what Intel can do with UEFI/firmware updates to improve the situation. The reported current draw of 200 amps @ 3.8 GHz with 1v vcore is disturbing.

I'll repeat it: it is in LinX with AVX-512 enabled on 7900X. Move to AVX2-only and it is like 120-140W at 3.8Ghz, more efficient than 6950X ever was.

Has Intel abandoned indium solder altogether, or do LGA 3647 CPUs still get the old method?

Xeon Phis are using same TIM as Skylake-X as far as i know. I have no clue about Skylake-SP, but seeing as Xeon Phis have higher TDP, i will be very surprised if they are using solder.

 

[tamz_msc]

Once again, Intel had a semi-custom 20C 200W TDP SKU with 3.2Ghz base clock. Since SKL-X visibly has better voltage/freq curve than BDW-E, i would be very surprised if they did not manage to fit 18C 3-3.2Ghz (their typical base clock for highest end SKU in the past) into 165W, especially with AVX-512 offset in play.

Running the 18C SKL-X at 3-3.2GHz base clock and 165W TDP would result in abysmal Turbo 2.0. The highest Turbo 2.0 has usually been ~1GHz over the base clock in the case of Broadwell-E, and the 10C 7900X. Can't have that especially when there are reports that the Xeon Gold 6150 at 165W TDP has an all core turbo of 3.4GHz.

 

[TheGiant]

Why is AVX512 so power hungry?

Is it really more efficient - throughput/power wise?

Who is the customer for AVX512 workload? google, amazon ?

 

[lolfail9001]

Running the 18C SKL-X at 3-3.2GHz base clock and 165W TDP would result in abysmal Turbo 2.0.

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

Can't have that especially when there are reports that the Xeon Gold 6150 at 165W TDP has an all core turbo of 3.4GHz.

That fits perfectly with what i expect 7980X to have, you know.

Why is AVX512 so power hungry?

Because it has more throughput on 7900X. AVX512 and AVX2 should use similar amounts of power on 7820X/7800X for example. And yeah, since uncore is constant it should technically be slightly more efficient throughput/watt wise.

 

[AdamK47]

Went to the local Micro Center today. No Skylake-X or X299. Whats the deal? I thought it was releasing today?

 

[Edrick]

Its killing me that I have had my 7820X since Saturday and my MB wont be here until Wednesday.

 

[krumme]

It is not, really. And please, do not bring up TR until you see how it performs with software that does not know how to act with dual socket systems.

We heard that story 10 years ago for kensfield. And yet over a simple dog slow fsb it only had minor impact. On 10 year old software and os.
So until the data proves otherwise its safe to assume where its going.
Nice try but fail.
Lol.

 

[krumme]

A 7900x is using 70 watts more stock in cb r15 than 6950. Load system. Pcper review.

238w at stock for a normal render task. It is hedt but the perf simply doesnt match power.
Lets be frank. Nobody expected that bad efficiency level.
https://www.pcper.com/reviews/Proce...X-Processor-Review/Power-Perf-Dollar-Conclusi

 

[lolfail9001]

We heard that story 10 years ago for kensfield.

Kentsfield is different. If you want that story, look up magny cours.

And yet over a simple dog slow fsb it only had minor impact.

Precisely: Core 2 did not have IMC, you had 1 memory controller for every core even if you put 4 dies on a package. (or rather 2 packages with 2 dies, see Skulltrail).

A 7900x is using 70 watts more stock in cb r15 than 6950. Load system. Pcper review.

At what clocks? Go, try and guess the clocks, i'll wait.

 

[Ajay]

Lets be frank. Nobody expected that bad efficiency level.

That much is true! I wonder just how much power those large L2$ are using. It must pay off in some common server tasks, or Intel really blew it.
I am wondering, a bit, if something is off with the X299 platform. It was rushed, like AM4 was. It will be interesting to see how things shake out by the time the i9 7920X is available.

 

[Guru]

I'm not a 24/7 gamer so for me I go for mid range parts, I think the value is pretty much always better on mid range parts and then upgrade every 2 years to maintain 60fps on average.

Right now I'm looking at a 6 core Ryzen processor, I've set my sights on the 1600 since it offers the best value and most can be overclocked to 3.9GHz on all cores, so getting a 6 core for about 200 euros that oc's to almost 4GHz, it also comes with a fan attached, unlike the x version.

Though I'm nor buying too soon, so if 8 cores become even cheaper, I might go for that.

 

[krumme]

At what clocks? Go, try and guess the clocks, i'll wait.

What do i care?
Aprox 15% more perf for 50% more power.
Its a record.
I dont understand why nobody didnt call them out on this more brutally.

 

[moonbogg]

What do i care?
Aprox 15% more perf for 50% more power.
Its a record.
I dont understand why nobody didnt call them out on this more brutally.

I think maybe at least a few people are just kind of stunned about the whole thing and don't know what to make of it. Maybe they are waiting for BIOS revisions to somehow sort things out before making any harsh declarations. The greatest sin above all IMO is the thermal paste. I'll never be able to understand why they would do this. I believe it reflects an obvious lack of prioritization regarding the consumer and especially enthusiast market. Maybe the paste makes sense for them when taking into account their whole product range because most people might not care, such as server customers where clocks are low anyway and maybe they want to just use the same process for all their chips to simplify things.
I think the ridiculous power consumption really caught people off guard. Taking into account the much higher power consumption, hilarious temps funny enough to make one cry with laughter as well as lower game performance in some titles despite the huge clock speed advantage compared to Broadwell, It doesn't surprise me to have seen some call this release Intel's 2017 netburst architecture or others even call it the new Bulldozer. I feel pretty good about my 6800K at the moment.

 

[IndyColtsFan]

Went to the local Micro Center today. No Skylake-X or X299. Whats the deal? I thought it was releasing today?

People already have the CPUs, but the boards (at least the preorders) were all set to ship tomorrow or later AFAIK.

 

[Ajay]

What do i care?
Aprox 15% more perf for 50% more power.
Its a record.
I dont understand why nobody didnt call them out on this more brutally.

Some have. I've seen a number of youtuber's being anything from perplexed to irate.

 

[Ajay]

Went to the local Micro Center today. No Skylake-X or X299. Whats the deal? I thought it was releasing today?

Dunno, their site is down right now.

 

[krumme]

That much is true! I wonder just how much power those large L2$ are using. It must pay off in some common server tasks, or Intel really blew it.
I am wondering, a bit, if something is off with the X299 platform. It was rushed, like AM4 was. It will be interesting to see how things shake out by the time the i9 7920X is available.

X299 is certainly rushed but imo not like the am4. Do you eg expect to see Intel shave 6ns ram latency in 1 month time or release some ram oc bios that gives 400MHz+ 3 months down the line.

I still think there is perhaps process cost saving issues involved here. But we need proper testing at different freq and also temp levels. I guess most was running at above 90c.

 

[lolfail9001]

What do i care?

Exactly, what are you doing in this thread?

Its a record.

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

I'll never be able to understand why they would do this.

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.

 

[Timmah!]

I really need to see more reviews from actual people using and OCing 7900x, it really cant come fast enough. I have Corsair H105 and i would be content with just 4,4 or 4,5Ghz overclock, but from those various reviews it seems like even that might be too much without delidding ....

Speaking about power consumption, i am starting to wonder, if my 750W PSU would be actually enough? Luckiy its Seasonic X-Series Gold, but then again i have 2 1080s...

Finally, when i put this rig together with the help of my friend, he put me the the liquid metal stuff instead of regular thermal paste between the CPU and the heatsink/water pump. I hope thats removable as much as the TIM is, right? I hope it was not some kind of permanent solution , i have no experience with it whatsoever and i sanctioned its use, cause i did not really think i would actually ever want to change CPU.

 

[krumme]

I think maybe at least a few people are just kind of stunned about the whole thing and don't know what to make of it. Maybe they are waiting for BIOS revisions to somehow sort things out before making any harsh declarations. The greatest sin above all IMO is the thermal paste. I'll never be able to understand why they would do this. I believe it reflects an obvious lack of prioritization regarding the consumer and especially enthusiast market. Maybe the paste makes sense for them when taking into account their whole product range because most people might not care, such as server customers where clocks are low anyway and maybe they want to just use the same process for all their chips to simplify things.
I think the ridiculous power consumption really caught people off guard. Taking into account the much higher power consumption, hilarious temps funny enough to make one cry with laughter as well as lower game performance in some titles despite the huge clock speed advantage compared to Broadwell, It doesn't surprise me to have seen some call this release Intel's 2017 netburst architecture or others even call it the new Bulldozer. I feel pretty good about my 6800K at the moment.

The 6800 was always a very fine cpu.
Nobody just never realized it.
Because they were looking at arma 3 bm with a titan x at 1080 on a 7700k or vs 1800x in cb.

For skl x The disaster about the tim is imo many will run into throtling at stock and therefore get unpredictable performance. Stock.

I mean when we heard about the tim we knew there would be problems for oc but the problem is actually worse than expected. Because its there at stock for reasonably air coolers.

No. Intel didnt invent a better solution.
No. The space is not less.

I follow Ajay a bit here. It looks like something went wrong. And not as intended. But it seems to me the problems is many.

 

[formulav8]

I wonder just how much power those large L2$ are using

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

 

[formulav8]

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

They cared enough to rush the new enthusiast product line before AMD could launch theirs and 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 performnce in their hedt offerings. So they still love the money and ego that the hedt line brings.

I do agree cheeping out using paste shows they don't care about certain aspects of enthusiasts in a certain sense. Even so people should just be ok with them doing things like that?