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

Page 13 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

jpiniero

Lifer
Oct 1, 2010
14,843
5,457
136
Probably not server parts but Skylake-W (1P workstation). Can't imagine it would have been that difficult to productize an X model.
 

DrMrLordX

Lifer
Apr 27, 2000
21,813
11,168
136
I think it's more likely that Intel held the 18c option in reserve, in case AMD did something to provoke a response. After all, they're doing little more than rebranding workstation Xeons here, so it's not like they needed much time to prepare.
 
Mar 10, 2006
11,715
2,012
126
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.

AMD has said Threadripper is a 16C/32T part, not 32C/64T.
 

blue11

Member
May 11, 2017
151
77
51
AMD has said Threadripper is a 16C/32T part, not 32C/64T.
It shares the same socket as Naples, so AMD can just release Naples as "ThreadRipper 2" or whatever. It looks like Intel won't be able to win the marketing game, since they only have up to 28 cores. After all, this farce isn't about performance or value or anything else. Both companies just want to be able to claim to have the most cores for PR reasons.
 
Last edited:

IndyColtsFan

Lifer
Sep 22, 2007
33,656
687
126
Well, with the leaks over the past few days, I think I've decided to go SKL-X. I think the 8 and 10 core models are probably the best combination of ST speed and MT performance, but the 8 core will likely have more OC headroom so that will likely be my choice.
 
Reactions: TheF34RChannel

blue11

Member
May 11, 2017
151
77
51
Well, with the leaks over the past few days, I think I've decided to go SKL-X. I think the 8 and 10 core models are probably the best combination of ST speed and MT performance, but the 8 core will likely have more OC headroom so that will likely be my choice.
The price leaks cemented the HEDT series as dead to me. Just like the previous generation, where you could get 14 cores for the same price as the i7-6950X or 10 cores for the same price as the i7-6900K, the smart shopper is going back to Xeons again. That said, the R7-1700 at 4 GHz is a hard proposition to turn down, and it looks like Intel continues to have no answer to it in terms of value.
 
Reactions: ddogg
Mar 10, 2006
11,715
2,012
126
The price leaks cemented the HEDT series as dead to me. Just like the previous generation, where you could get 14 cores for the same price as the i7-6950X or 10 cores for the same price as the i7-6900K, the smart shopper is going back to Xeons again. That said, the R7-1700 at 4 GHz is a hard proposition to turn down, and it looks like Intel continues to have no answer to it in terms of value.

You don't even know the pricing on the Skylake-X parts yet, perhaps you should wait for the official announcement before drawing such conclusions?
 

Ajay

Lifer
Jan 8, 2001
16,094
8,106
136
That said, the R7-1700 at 4 GHz is a hard proposition to turn down, and it looks like Intel continues to have no answer to it in terms of value.

If they all did 4GHz easily, that would be more compelling. I don't know if the new stepping (due this summer?) will help at all.
As to the rest, as @Arachnotronic notes, we really need to see actual pricing from Intel, and we get that this week
 
Reactions: Drazick

ddogg

Golden Member
May 4, 2005
1,864
361
136
I agree, if these leaked prices are accurate they make no sense when you can get a 1700 8 core CPU for close to 300. There's no way I'm paying 650-700 for a 6900k replacement. I'm hoping AMD thread-ripper CPUs are aggressively priced as I will likely pick one up.

Sent from my Nexus 6P using Tapatalk
 

wildhorse2k

Member
May 12, 2017
180
83
71
I agree, if these leaked prices are accurate they make no sense when you can get a 1700 8 core CPU for close to 300. There's no way I'm paying 650-700 for a 6900k replacement. I'm hoping AMD thread-ripper CPUs are aggressively priced as I will likely pick one up.

Sent from my Nexus 6P using Tapatalk

It makes more sense to those who were considering or bought a 1800X. Those are people who don't necessarily want to OC and have the money to afford a good 8 core. Those looking for 1700 and OC will hardly be interested in any Intel CPU as they historically tend to be more expensive.

It could happen that some of 1800X owners will be selling them, unhappy with gaming and single thread performance and getting i9 7820X instead.

Well, with the leaks over the past few days, I think I've decided to go SKL-X. I think the 8 and 10 core models are probably the best combination of ST speed and MT performance, but the 8 core will likely have more OC headroom so that will likely be my choice.

I'm too convinced the 8 core model will be an excellent all-around CPU and OC well. Good choice for those who are willing to pay a little more than 1800X. Motherboards are much better. My choice is 10 core as it will be the most efficient monolithic multi-core CPU and lot of PCIe lanes. 12 core will OC less, suffer from latencies, memory could run at lower clocks too.
 

Rngwn

Member
Dec 17, 2015
143
24
36
Already with with R7 1700 since last month. Have no choice but "no" for at least 3-4 years to come.
 

blue11

Member
May 11, 2017
151
77
51
So we now have actual pricing information. The low-end 6-10C SKUs, particularly the 10C, are still offering terrible value in the face of $300 R7-1700. On the other hand, the linear pricing of additional cores from 12-18C is surprisingly good value from Intel.

Probably the most frustrating thing is the AVX-512 status, which is still unknown. Intel claims that the 18C SKU will have "teraflop" performance, but a full speed AVX-512 implementation should be reaching 1.7 TFLOP/s double-precision and 3.4 TFLOP/s single precision at 3 GHz.

512-bit vector / 64-bit word * 2 IPC * 2 = 32 FLOP/cycle
32 FLOP/cycle * 3 GHz * 18 = 1728 GFLOP/s

If AVX-512 is full speed, a quad-channel memory interface will not be enough bandwidth.
 
Last edited:

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
Probably the most frustrating thing is the AVX-512 status, which is still unknown.
It is literally stated on Intel's slide deck, Skylake-X has AVX512, Kaby Lake-Rip Off Extreme does not.
but a full speed AVX-512 implementation should be reaching 1.7 TFLOP/s double-precision and 3.4 TFLOP/s single precision at 3 GHz.
Why would it? For all i know full-speed AVX512 can easily run you the same speed as AVX2 on Haswell/Skylake does currently, at 1 AVX512 FMA per clock.
 

blue11

Member
May 11, 2017
151
77
51
It is literally stated on Intel's slide deck, Skylake-X has AVX512, Kaby Lake-Rip Off Extreme does not.

Why would it? For all i know full-speed AVX512 can easily run you the same speed as AVX2 on Haswell/Skylake does currently, at 1 AVX512 FMA per clock.
Of course the slide says it has AVX-512. What people actually want to know is if it supports AVX-512 at 2 IPC. Sure, Intel can make a gimped version that only does 1 IPC like AMD with Zen. It'll get the exact same (actually worse, because of latency) performance as regular AVX, so it'll be useless.
 

CHADBOGA

Platinum Member
Mar 31, 2009
2,135
832
136
Now that Intel has gone Core Crazy too, perhaps now we can have honest discussions about how few desktop applications actually benefit meaningfully from more than 4 cores.
 

lolfail9001

Golden Member
Sep 9, 2016
1,056
353
96
It'll get the exact same (actually worse, because of latency) performance as regular AVX, so it'll be useless.
I'll let a certain guy running AVX512 software on Skylake-SP chime in
Depending on the application much more than that. I'm running a few Skylake-EP instances on GCP as we speak and they are magical.

Intel CPUs were always cache starved and AVX-512 with CDI is a game changer since it allows you to vectorize code that otherwise would not be possible to vectorize at all due to memory conflicts.

ImageMagick recompiled with AVX-512 support gets 5 fold increase in historgraming, blur/beblur and many other image processing functions. FFTW which is used in a metric tons of things also gets a similar performance boost.

I need to take a deeper look at handbreak and a few other things in the future.

Did you get the hint? If you did not, let me spell it out: who cares about throughput when you can't use it. AVX512 by it's own existence allows to use the throughput Skylake/Haswell had often underutilized outside of burn-in tests.
 
Reactions: Arachnotronic

dullard

Elite Member
May 21, 2001
25,214
3,632
126
Now that Intel has gone Core Crazy too, perhaps now we can have honest discussions about how few desktop applications actually benefit meaningfully from more than 4 cores.
As a programmer (for part of my job), it is extremely difficult to properly sub-divide the problem into exactly 2, 4, 6, 8, 10, etc parts. Especially when you have to consider many different types of computers that your software will be used on. So, when you are already working 60+ hour weeks just to meet the deadlines, you don't have the time to sit back and optimize everything for every end-use. The default for me, unless there is a known terrible bottleneck, is just to divide the software into chunks that are logical to program--not into chunks that are ideal for every possible CPU.

For example, if I were to program a game, I would have one thread for the player's actions, one for the map and other stationary items, one for the movable items (enemies/NPCs/other characters) and one for communication with the game server. There you go, four cores done. To then go back and decide well, in this room of that map with this many other things going on then I'll program enemy #3 to use 2 cores instead of 1, but if that enemy goes into the next room free up that core so that the user's actions have 2 cores, would be very, very challenging and lead to nearly no gains.
 
Reactions: richierich1212

TheF34RChannel

Senior member
May 18, 2017
786
309
136
Now that Intel has gone Core Crazy too, perhaps now we can have honest discussions about how few desktop applications actually benefit meaningfully from more than 4 cores.

For today you are mostly right. Tomorrow however... Besides, now with a plethora of many core CPUs devs may be moved into that direction as well.
 

swilli89

Golden Member
Mar 23, 2010
1,558
1,181
136
No one is crying foul that their 8c is $599 and then... a straight jump to $1000 for 2 more cores? How is anyone ok with that?

Kinda sucks that they are still going to be using TIM for their $2000 18C chips.. Same old Intel ha..
 

dullard

Elite Member
May 21, 2001
25,214
3,632
126
No one is crying foul that their 8c is $599 and then... a straight jump to $1000 for 2 more cores? How is anyone ok with that?

Kinda sucks that they are still going to be using TIM for their $2000 18C chips.. Same old Intel ha..
Why cry foul when the extreme edition level chips have been priced EXACTLY like that for years and years.

Skylake: 7820X $599, 7900X $999 (post-Ryzen, Intel went back to the old pricing)
Broadwell: 6850K: $617, 6900K $1089 (pre-Ryzen was a bit higher than their normal pricing)
Haswell: 5930K: $583, 5960X: $999
Ivy Bridge: 4930K: $583, 4960X: $999
Sandy Bridge: 3930K: $583, 3960X: $999
Westmere: 980: $583, 990X: $999
Nehalem (2009 version): 950: $562, 975: $999
Nehalem (2008 version): 940: $562, 965: $999

Same old Intel, you are correct.
 
Last edited:
Reactions: Sweepr

CHADBOGA

Platinum Member
Mar 31, 2009
2,135
832
136
For today you are mostly right. Tomorrow however... Besides, now with a plethora of many core CPUs devs may be moved into that direction as well.
$400+ CPU's will not bring about a multi-core/multi-threaded software revolution.
 

swilli89

Golden Member
Mar 23, 2010
1,558
1,181
136
Why cry foul when the extreme edition level chips have been priced EXACTLY like that for years and years.

Skylake: 7820X $599, 7900X $999 (post-Ryzen, Intel went back to the old pricing)
Broadwell: 6850K: $617, 6900K $1089 (pre-Ryzen was a bit higher than their normal pricing)
Haswell: 5930K: $583, 5960X: $999
Ivy Bridge: 4930K: $583, 4960X: $999
Sandy Bridge: 3930K: $583, 3960X: $999
Westmere: 980: $583, 990X: $999
Nehalem (2009 version): 950: $562, 975: $999
Nehalem (2008 version): 940: $562, 965: $999

Same old Intel, you are correct.

So you're saying that since we've had gouged margin products historically that its actually supposed to be ok that Intel charges $400 more for 2 additional cores on the same socket?

History has nothing to do with this. the 10 core chip is an absolutely awful value proposition and since we aren't talking about gaming core count/$ is everything.
 

blue11

Member
May 11, 2017
151
77
51
I'll let a certain guy running AVX512 software on Skylake-SP chime in


Did you get the hint? If you did not, let me spell it out: who cares about throughput when you can't use it. AVX512 by it's own existence allows to use the throughput Skylake/Haswell had often underutilized outside of burn-in tests.
I do get that you have no idea what you're talking about. Thanks for letting me know that I'm making "burn-in tests," instead of useful software. I'm sure my management will find that information useful.

Also, Skylake-SP has nothing to do with Skylake-X, which has crippled AVX-512.
 
Last edited:

IEC

Elite Member
Super Moderator
Jun 10, 2004
14,362
5,033
136
So you're saying that since we've had gouged margin products historically that its actually supposed to be ok that Intel charges $400 more for 2 additional cores on the same socket?

History has nothing to do with this. the 10 core chip is an absolutely awful value proposition and since we aren't talking about gaming core count/$ is everything.

Whether you think it is okay or not ultimately doesn't matter. If the market will pay it, they will price it accordingly.

Remember that there is the economics concept of "optimal pricing" - they are a multi-billion dollar corporation, and they didn't get to their current status as industry heavyweight by being a charity. Salaries and R&D are expensive.
 
Reactions: Drazick
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |