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

[TheF34RChannel]

From what I've seen so far, it reinforces my opinion to hold off on an upgrade until the fall and see what's best. I believe that once TR is released, there will be a price war. At any rate, I am very concerned with the temps I've seen and hope the reviews next week put my mind at ease.

Between what will you be deciding? I was hell-bent on getting a 7820X or maybe even a 7900X, however I've just taken two steps back, you know, to let sanity get back in, to see what Monday brings and let myself get back to where I started my quest: wanting a (Intel) more than 4C chip with HT for gaming (=high frequency) and it may very well be CFL-S that's the best fit for my needs. Maybe Intel will surprise us with an 8C part?

8C is the new 4C for sure. That's why I am hoping that Intel will release an 8C/16T Coffee Lake-S part with a nice big fat L3 cache (15MB would be nice), it would also benefit their mainstream market seeing as the competition has just that. So it's hoping that, but doubting they will. It would make for a more logical segmentation as well: leave HEDT for the prosumers and the mainstream for gamers and below, with the 8C/16T chip they deserve.

 

[Topweasel]

Between what will you be deciding? I was hell-bent on getting a 7820X or maybe even a 7900X, however I've just taken two steps back, you know, to let sanity get back in, to see what Monday brings and let myself get back to where I started my quest: wanting a (Intel) more than 4C chip with HT for gaming (=high frequency) and it may very well be CFL-S that's the best fit for my needs. Maybe Intel will surprise us with an 8C part?

8C is the new 4C for sure. That's why I am hoping that Intel will release an 8C/16T Coffee Lake-S part with a nice big fat L3 cache (15MB would be nice), it would also benefit their mainstream market seeing as the competition has just that. So it's hoping that, but doubting they will. It would make for a more logical segmentation as well: leave HEDT for the prosumers and the mainstream for gamers and below, with the 8C/16T chip they deserve.

It is way to late for coffee lake.

 

[jpiniero]

8C is the new 4C for sure. That's why I am hoping that Intel will release an 8C/16T Coffee Lake-S part with a nice big fat L3 cache (15MB would be nice), it would also benefit their mainstream market seeing as the competition has just that. So it's hoping that, but doubting they will. It would make for a more logical segmentation as well: leave HEDT for the prosumers and the mainstream for gamers and below, with the 8C/16T chip they deserve.

The extra cores aren't going to help. Seriously, people are going to be disapointed in 6C12T Coffee Lake in gaming unless it delivers on clocks. Despite that, It'll still be the #1 gaming processor though.

 

[IndyColtsFan]

Between what will you be deciding? I was hell-bent on getting a 7820X or maybe even a 7900X, however I've just taken two steps back, you know, to let sanity get back in, to see what Monday brings and let myself get back to where I started my quest: wanting a (Intel) more than 4C chip with HT for gaming (=high frequency) and it may very well be CFL-S that's the best fit for my needs. Maybe Intel will surprise us with an 8C part?

8C is the new 4C for sure. That's why I am hoping that Intel will release an 8C/16T Coffee Lake-S part with a nice big fat L3 cache (15MB would be nice), it would also benefit their mainstream market seeing as the competition has just that. So it's hoping that, but doubting they will. It would make for a more logical segmentation as well: leave HEDT for the prosumers and the mainstream for gamers and below, with the 8C/16T chip they deserve.

For me, it will be between SKL-X, TR, and CFL. I may even wait until closer to Christmas because Frys and Microcenter have had good sales on the HEDT CPUs the past couple of Christmases. Between SKL-X and TR, I have a feeling SKL-X will be faster but TR will be fast enough and cheap enough that the SKL-X premium is going to be super hard to justify. That's why I believe Intel is going to engage in a price war - a 16 core TR is likely going to be (if rumors old) cheaper than the i9-7900x and will likely trounce it in MT. Both TR and SKL-X should handily beat my 2600k in ST as well. And CFL is an unknown at this stage. That's why I'm going to try to hold out because I have no urgent need to upgrade at the moment (other than my upgrade itch) and the more I save, the more I can beef up other components. Also, I jumped on Ryzen early and have gotten bitten by some platform issues, so I want to give both x299 and x399 a few months to mature before committing.

I saw a rumor (which I am highly skeptical of given the source) that Intel is planning to release soldered versions of the HEDT chips in Q4 or Q1 18. I think what is more likely is that they may have an HEDT refresh next spring/summer and maybe release higher core counts on the KBL-X architecture. At any rate, I doubt those would be soldered but I could be mistaken.

 

[TheF34RChannel]

The extra cores aren't going to help. Seriously, people are going to be disapointed in 6C12T Coffee Lake in gaming unless it delivers on clocks. Despite that, It'll still be the #1 gaming processor though.

There is way too little known at the moment to make any such assumption if you ask me. I definitely agree with you if we are looking at today's game with the exception here and there (anything Frostbite, Destiny 2), however if I'm taking a plunge wherever, I may as well be prepared I figure - and, like you said, it will be probably be the #1 gaming CPU (to which I add) with a putative high frequency @ 14nm++ and decent L3 cache increase (12MB if I remember correctly).

For me, it will be between SKL-X, TR, and CFL. I may even wait until closer to Christmas because Frys and Microcenter have had good sales on the HEDT CPUs the past couple of Christmases. Between SKL-X and TR, I have a feeling SKL-X will be faster but TR will be fast enough and cheap enough that the SKL-X premium is going to be super hard to justify. That's why I believe Intel is going to engage in a price war - a 16 core TR is likely going to be (if rumors old) cheaper than the i9-7900x and will likely trounce it in MT. Both TR and SKL-X should handily beat my 2600k in ST as well. And CFL is an unknown at this stage. That's why I'm going to try to hold out because I have no urgent need to upgrade at the moment (other than my upgrade itch) and the more I save, the more I can beef up other components. Also, I jumped on Ryzen early and have gotten bitten by some platform issues, so I want to give both x299 and x399 a few months to mature before committing.

I saw a rumor (which I am highly skeptical of given the source) that Intel is planning to release soldered versions of the HEDT chips in Q4 or Q1 18. I think what is more likely is that they may have an HEDT refresh next spring/summer and maybe release higher core counts on the KBL-X architecture. At any rate, I doubt those would be soldered but I could be mistaken.

Your choices: Giving either platform time will benefit you for sure; it's the way it's always been, practically. Good that have you time and are willing to use it, well, usefully. Whatever you end up with will be an improvement over your current setup, like you said. Don't expect a night and day difference though; I came from the same CPU as yours to this and although it is better, it's not a revolutionary difference in my experience. That said, they are beginning to show their age, aren't they. We should say more Coffee Lake-S leaks this month and July, August and I am eager to see them and give the idea of getting a chance.

$850 TR rumor: No way in hell they will sell a 16C chip for that money unless it's seriously flawed. Didn't a 32C Epyc part price leak this week for $4000? I'm not sure. The $850, I believe to remember, is also a story originating from an obscure website and subsequently the 'I want to believe' (for all you Mulder fans out there ) mindsets took off. I'm not buying it.

Solder rumor: Don't buy into that rumor, please; it's someone's musing on a bad site turned into news. The article itself isn't worthy of the name article. See: https://forums.anandtech.com/thread...core-i9-7980xe.2428363/page-495#post-38942772

 

[moonbogg]

For me, it will be between SKL-X, TR, and CFL. I may even wait until closer to Christmas because Frys and Microcenter have had good sales on the HEDT CPUs the past couple of Christmases. Between SKL-X and TR, I have a feeling SKL-X will be faster but TR will be fast enough and cheap enough that the SKL-X premium is going to be super hard to justify. That's why I believe Intel is going to engage in a price war - a 16 core TR is likely going to be (if rumors old) cheaper than the i9-7900x and will likely trounce it in MT. Both TR and SKL-X should handily beat my 2600k in ST as well. And CFL is an unknown at this stage. That's why I'm going to try to hold out because I have no urgent need to upgrade at the moment (other than my upgrade itch) and the more I save, the more I can beef up other components. Also, I jumped on Ryzen early and have gotten bitten by some platform issues, so I want to give both x299 and x399 a few months to mature before committing.

I saw a rumor (which I am highly skeptical of given the source) that Intel is planning to release soldered versions of the HEDT chips in Q4 or Q1 18. I think what is more likely is that they may have an HEDT refresh next spring/summer and maybe release higher core counts on the KBL-X architecture. At any rate, I doubt those would be soldered but I could be mistaken.

I thought the same thing, but there's a reason I have a 6800K right now instead of the "something else" that I returned to the store. I don't need the best single thread performance in the world, but I wasn't willing to go backwards on it, which is exactly what would have happened coming from a 4.6ghz Sandy Bridge. So I got a 6800K instead. Make sure to watch the single thread performance of whatever you buy. If its important to you, its easy to assume one thing and end up with a downgrade.
Multithread performance, well that's a whole different ball game of course.

 

[Arachnotronic]

Hope no one is done X299 testing & taking weekend off! Newest BIOes from last 48hrs provide significant perf uplift. Time to retest.....

https://twitter.com/hardwarecanucks/status/876162292139909120

 

[Arachnotronic]

Between what will you be deciding? I was hell-bent on getting a 7820X or maybe even a 7900X, however I've just taken two steps back, you know, to let sanity get back in, to see what Monday brings and let myself get back to where I started my quest: wanting a (Intel) more than 4C chip with HT for gaming (=high frequency) and it may very well be CFL-S that's the best fit for my needs. Maybe Intel will surprise us with an 8C part?

8C is the new 4C for sure. That's why I am hoping that Intel will release an 8C/16T Coffee Lake-S part with a nice big fat L3 cache (15MB would be nice), it would also benefit their mainstream market seeing as the competition has just that. So it's hoping that, but doubting they will. It would make for a more logical segmentation as well: leave HEDT for the prosumers and the mainstream for gamers and below, with the 8C/16T chip they deserve.

CFL-H and the top CFL-S dies are 6 core/12 thread. Intel's original plan was to release Cannon Lake-S with 4/6/8 core options, but this project was killed due to issues with 10nm and replaced with the 14nm++ Coffee Lake.

I think we will get an 8 core mainstream product out of Intel with Ice Lake-S.

 

[jpiniero]

CFL-H and the top CFL-S dies are 6 core/12 thread. Intel's original plan was to release Cannon Lake-S with 4/6/8 core options, but this project was killed due to issues with 10nm and replaced with the 14nm++ Coffee Lake.

I think we will get an 8 core mainstream product out of Intel with Ice Lake-S.

IIRC, it was 2, 4 and 8 core dies. As for Icelake I would think it would be the same although if it's using EMIB all bets are off.

 

[Ajay]

IIRC, it was 2, 4 and 8 core dies. As for Icelake I would think it would be the same although if it's using EMIB all bets are off.

I would be surprised if Intel use EMIB for client CPUs. @ 10nm even the 8 core is going to be a small die.

 

[dlerious]

Late reply, but I voted no. Built a Ryzen system already. I'm seriously considering Threadripper though. It depends on whether or not I want to spend the money. I could use the cores for some beefy VMs, the PCIe lanes for capture cards, 10GB ethernet, possibly storage. ECC memory is nice, but wallet empties pretty quick. RGB I can live without.

 

[Ajay]

https://twitter.com/hardwarecanucks/status/876162292139909120

Really looking forward to Monday now

 

[jpiniero]

I would be surprised if Intel use EMIB for client CPUs. @ 10nm even the 8 core is going to be a small die.

They need the practice for Tigerlake-SP.

 

[dlerious]

$850 TR rumor: No way in hell they will sell a 16C chip for that money unless it's seriously flawed. Didn't a 32C Epyc part price leak this week for $4000? I'm not sure. The $850, I believe to remember, is also a story originating from an obscure website and subsequently the 'I want to believe' (for all you Mulder fans out there ) mindsets took off. I'm not buying it.

$850 sounds low, but it's believable, 2 1800X is less than $1000. The Epyc prices, I don't know. They're server parts usually at a premium. The 24 core Intel Xeon E7-8894 v4 is $8898.

 

[Shivansps]

$850 sounds low, but it's believable, 2 1800X is less than $1000. The Epyc prices, I don't know. They're server parts usually at a premium. The 24 core Intel Xeon E7-8894 v4 is $8898.

Well leaked EPYC prices shows +$600 for 16C versions, so $850 is certanly possible. BUT those are kinda low clock versions, TR needs to be well over 3.3ghz ACT to be competitive here, otherwise it may lose to a 10C SKL-X or the 12C at least.

They also dont want to eat their own EPYC sales with TR. The higher clock 16C EPYC is at +$1100 and i still dont think that clock is enoght.

 

[TheF34RChannel]

CFL-H and the top CFL-S dies are 6 core/12 thread. Intel's original plan was to release Cannon Lake-S with 4/6/8 core options, but this project was killed due to issues with 10nm and replaced with the 14nm++ Coffee Lake.

I think we will get an 8 core mainstream product out of Intel with Ice Lake-S.

Quite possibly right. I'd hate it (and get it or its HEDT platform, regardless) if true. We want 8C mainstream parts sooner rather than later. One the one hand I can see it happening with Ice Lake-S, on the other hand Intel hates being left behind (refer to 18C SKL-X) so they may be preparing one for CFL-S. Who knows, time will tell.

 

[Shivansps]

Quite possibly right. I'd hate it (and get it or its HEDT platform, regardless) if true. We want 8C mainstream parts sooner rather than later. One the one hand I can see it happening with Ice Lake-S, on the other hand Intel hates being left behind (refer to 18C SKL-X) so they may be preparing one for CFL-S. Who knows, time will tell.

As far as i can recall Intel doesn't do preemptive strikes (in reference to the 18C SKU). The bad thing about 8C in Icelake is that im going to get bored in the following years hearing about it is thanks to AMD Ryzen, when the 8C was already planned to Cannonlake and it got cancelled and replaced by CFL.

 

[jpiniero]

on the other hand Intel hates being left behind (refer to 18C SKL-X) so they may be preparing one for CFL-S. Who knows, time will tell.

I don't think Intel feels like they are being left behind with CFL, they just needed something to sell since they cancelled the higher watt Cannonlake. The 18C is happening because they do want to keep the e-penis crowd and it was easy since they were already planning on doing MCC Xeons on LGA 2066.

 

[Arachnotronic]

I would be surprised if Intel use EMIB for client CPUs. @ 10nm even the 8 core is going to be a small die.

I suspect even if we see EMIB on client dies, the CPU cores will be in their own "tile".

 

[IndyColtsFan]

I thought the same thing, but there's a reason I have a 6800K right now instead of the "something else" that I returned to the store. I don't need the best single thread performance in the world, but I wasn't willing to go backwards on it, which is exactly what would have happened coming from a 4.6ghz Sandy Bridge. So I got a 6800K instead. Make sure to watch the single thread performance of whatever you buy. If its important to you, its easy to assume one thing and end up with a downgrade.
Multithread performance, well that's a whole different ball game of course.

That's the great debate, isn't it? Over the past decade, my philosophy has always been to pick faster cores rather than more cores. It has generally worked out well for me. And you do bring up a good point about clock speed - I have a Ryzen 1500x and it would beat my 2600k at equal clocks, but I haven't gotten the Ryzen above 3.9 Ghz with stability though it boots at 4 Ghz. Therefore, the 2600k in my sig is faster.

This time, however, I've come to the realization that if I wait for a CPU which is 50% faster than my 2600k at the same clock, I'm probably going to be waiting another 5 years. As I mentioned in a thread earlier this year, it is almost to the point where you're upgrading more for platform enhancements than anything else. I also believe that while more cores will be important in the future, we probably won't see 12+ threads as a firm requirement for several years.

So this time around, knowing that whatever CPU I choose for my next build could likely last me the better part of a decade, what do I do? Do I jump on what will likely be the fastest ST processor of the bunch, Coffee Lake, and just resign myself to the fact that if I am wrong on the timeline for games taking advantage of more than 6 cores and it comes sooner than I thought, I'll just have to upgrade again sooner? Or do I go for a massive multicore chip (10+ cores), sacrifice a little of the single thread performance, and live well for 5-10 years? TR intrigues me but objectively speaking, SKL-X is likely the only platform where I can have better ST performance than my 2600k @4.5 Ghz. The temps REALLY concern me though.....

You know it is bad when your BIOS battery will likely die before your CPU is obsolete and unusable.

 

[jpiniero]

I suspect even if we see EMIB on client dies, the CPU cores will be in their own "tile".

If they do, they could split it up like this:

PCH (and keep it at 14++)
Cores + Memory Controller + L3
GPU (optional of course)
Everything else (eg: Uncore)

A lot of this is dependent on EMIB being good enough. Which we don't know.

 

[TheF34RChannel]

As far as i can recall Intel doesn't do preemptive strikes (in reference to the 18C SKU). The bad thing about 8C in Icelake is that im going to get bored in the following years hearing about it is thanks to AMD Ryzen, when the 8C was already planned to Cannonlake and it got cancelled and replaced by CFL.

In this case it's not preemptive with AMD already offering several 8C mainstream SKUs.

That's the great debate, isn't it? Over the past decade, my philosophy has always been to pick faster cores rather than more cores. It has generally worked out well for me. And you do bring up a good point about clock speed - I have a Ryzen 1500x and it would beat my 2600k at equal clocks, but I haven't gotten the Ryzen above 3.9 Ghz with stability though it boots at 4 Ghz. Therefore, the 2600k in my sig is faster.

This time, however, I've come to the realization that if I wait for a CPU which is 50% faster than my 2600k at the same clock, I'm probably going to be waiting another 5 years. As I mentioned in a thread earlier this year, it is almost to the point where you're upgrading more for platform enhancements than anything else. I also believe that while more cores will be important in the future, we probably won't see 12+ threads as a firm requirement for several years.

So this time around, knowing that whatever CPU I choose for my next build could likely last me the better part of a decade, what do I do? Do I jump on what will likely be the fastest ST processor of the bunch, Coffee Lake, and just resign myself to the fact that if I am wrong on the timeline for games taking advantage of more than 6 cores and it comes sooner than I thought, I'll just have to upgrade again sooner? Or do I go for a massive multicore chip (10+ cores), sacrifice a little of the single thread performance, and live well for 5-10 years? TR intrigues me but objectively speaking, SKL-X is likely the only platform where I can have better ST performance than my 2600k @4.5 Ghz. The temps REALLY concern me though.....

You know it is bad when your BIOS battery will likely die before your CPU is obsolete and unusable.

Tough decision with each being based on expectations not making it easier. Waiting until all the facts are in makes your dilemma somewhat easier. Other than that, I can't tell you what to do.

 

[Ajay]

I suspect even if we see EMIB on client dies, the CPU cores will be in their own "tile".

I suppose, if Intel thinks that yield improvements with a split CPU/GPU are worth the extra packaging costs. EMIB is a low cost multi-chip packaging solutions, not no cost.

 

[Arachnotronic]

I suppose, if Intel thinks that yield improvements with a split CPU/GPU are worth the extra packaging costs. EMIB is a low cost multi-chip packaging solutions, not no cost.

I think they'll eat the extra packaging costs in exchange for the modularity and ability to build a wide proliferation of SKUs. Intel's modus operandi appears to be aggressive product segmentation and this looks like a great way to get there.

 

[TahoeDust]

Well...I pre-ordered my Asus Strix X299-E. It was my original cut instinct when all the boards were coming out at Computex. I figure if I start reading bad things about it, I will just cancel and go with something else.