Intel Skylake / Kaby Lake

[TheF34RChannel]

Personally i have preference for Gigabyte boards. I guess its mostly down to the fact my last 3 boards were Gigabytes and all the rigs with them were pretty much without any major issues and instabilities. The only time i got annoyed was when i needed to borrow HW-E chip to flash my latest board for the BIOS supporting BW-E, cause it was not doable without it - that was pretty stupid mistake on Gigabyte´s part. Otherwise though, i am fine with my choice.

Then again, if i went Asus or MSI, i would probably equally satisfied right now. Still, if i decide to waste some more money this year on x299 (or x399 if i will go threadripper), then i will most likely choose Gigabyte again - say UD4 or preferably Aorus 3 board. Unless ofc MSI or Asus offers something better equipped and rated in that up to 250 range.

If Sweepr is correct and the TUF is 250 USD I'd get that for the quality it offers. I can almost assure you that you won't be disappointed.

 

[Ajay]

Its not about stability - why pay obscene amounts of money for a platform when you don't know where its performance will fall relative to the competition? (which will be available at approximately the same time).

Seems pretty stupid to me.

There are always early adopters, people who want to be on the cutting edge - and are willing to endure a few rough edges if necessary. It's just the way of the world. Just because you regard it as 'stupid' doesn't mean it's wrong.

 

[Timmah!]

If Sweepr is correct and the TUF is 250 USD I'd get that for the quality it offers. I can almost assure you that you won't be disappointed.

I will certainly take it into consideration, thank you.

BTW, it should be non-issue to replace my current X99 with BW-E CPU for X299 and SKL-X while keeping everything else intact, right? I mean, RAM, GPUs, drives, everything should be compatible?

 

[Edrick]

There are always early adopters, people who want to be on the cutting edge - and are willing to endure a few rough edges if necessary. It's just the way of the world. Just because you regard it as 'stupid' doesn't mean it's wrong.

Exactly. I bought into the X79 platform the day it was released and I had a very painful month trying to get a stable system (SATA drivers were the biggest pain). But after a few BIOS updates and Intel driver updates later, everything was great.

 

[ManyThreads]

I will certainly take it into consideration, thank you.

BTW, it should be non-issue to replace my current X99 with BW-E CPU for X299 and SKL-X while keeping everything else intact, right? I mean, RAM, GPUs, drives, everything should be compatible?

Yes, barring any weird unforeseen bios issues with RAM or something.

 

[AdamK47]

I thought this was funny.

I made up my mind last week not to buy an i9 and the X299 on the release date. I'll be going 16 core Threadripper if the 10 and 12 core i9 performance falls behind. If performance of Threadripper trails, it's the 18 core i9 and an Asus Rampage VI Extreme for me in October.

It's good to have AMD back in the game.

 

[twothreefive]

With Intel having higher overclocks and leading in IPC, I would not be surprised to see the I9 have a 25% to 35% advantage per core. I am glad AMD is back in the game though. Intel would have never released Skylake X so soon if Ryzen was not competitive with their current offerings.

 

[Markfw]

With Intel having higher overclocks and leading in IPC, I would not be surprised to see the I9 have a 25% to 35% advantage per core. I am glad AMD is back in the game though. Intel would have never released Skylake X so soon if Ryzen was not competitive with their current offerings.

I am not sure about the IPC lead. I have many Xeons E5 and Ryzens. They are equal, My E5 has more cores, my Ryzen has more clock speed, thus maybe more IPC. I will wait for benchmarks.

14 cores and 2.5 ghz vs 8 cores and 3.7 ghz. My gut says Ryzen has more IPC against this Xeon

 

[WhoBeDaPlaya]

I am not sure about the IPC lead. I have many Xeons E5 and Ryzens. They are equal, My E5 has more cores, my Ryzen has more clock speed, thus more IPC. I will wait for benchmarks.

While I get where you're coming from (IPC per chip), I think twothreefive was referring to IPC on a per core basis.
It certainly is impressive that an overclocked Ryzen 7 can match the MT performance of a pair of X5690s!

 

[Markfw]

While I get where you're coming from (IPC per chip), I think twothreefive was referring to IPC on a per core basis.
It certainly is impressive that an overclocked Ryzen 7 can match the MT performance of a pair of X5690s!

No, ONE E5-2683 14 core@2.5 turbo vs ONE 1700x 8 core@3.7 turbo. So its hard to tell the IPC difference. Too many variables. Where I disagree is that Intel has higher overclocks and higher IPC. At least based on current information and benchmarks.

Edit: And thats STOCK turbo, not overclocked on both CPU's/

 

[coercitiv]

With Intel having higher overclocks and leading in IPC, I would not be surprised to see the I9 have a 25% to 35% advantage per core.

When it comes to this many cores (10+) we should no longer discuss in terms of performance per core, clocks will vary quite a bit in Intel's SKUs from 10 cores to 14-18, so performance per core will go down as performance per chip increases to overtake TR in all benchmarks.

I think it's very interesting that this year we've gone from a period when Intel was outright dismissing Zen as potential challenger (at least in the eyes of the press & customers) to a period when Intel holds off specs for their 12C HEDT CPU in order to better meet the competition in August. Actually, this might tell us where Intel expects SKL-X and TR to get close enough that even a minor clock jump may matter.

 

[wildhorse2k]

When it comes to this many cores (10+) we should no longer discuss in terms of performance per core, clocks will vary quite a bit in Intel's SKUs from 10 cores to 14-18, so performance per core will go down as performance per chip increases to overtake TR in all benchmarks.

I think it's very interesting that this year we've gone from a period when Intel was outright dismissing Zen as potential challenger (at least in the eyes of the press & customers) to a period when Intel holds off specs for their 12C HEDT CPU in order to better meet the competition in August. Actually, this might tell us where Intel expects SKL-X and TR to get close enough that even a minor clock jump may matter.

They have 160W TDP limit on the 12C, they can't clock it higher even if it loses to competition. To determine the clocks I guess they need to manufacture perhaps thousand samples and see where they stand in TDP and frequency on average.
It's unnecessary to release this information now even if they had it as it could give an advantage to competitor.

 

[coercitiv]

They have 160W TDP limit on the 12C, they can't clock it higher even if it loses to competition.

How exactly do you reckon that considering there's binning involved?

 

[Zucker2k]

You, sir, speaketh the truth as I see it as well. Is that a 10C I see in your future?

Won't that be swell?! Haha. Might break the piggy bank though . I'll be content with a 7820X

 

[Atari2600]

But it should be about stability, foremost.

Enthusiasts aren't slaves to stability unlike say professional markets.

But I suspect you'll dance around different priorities till it fits your agenda.

I believe most are aware of AMD's upcoming processor(s).

So they have detailed performance comparisons of AMD's upcoming processors against Intel's upcoming processors?

Any chance these people could share this information?

 

[Atari2600]

There are always early adopters, people who want to be on the cutting edge - and are willing to endure a few rough edges if necessary. It's just the way of the world. Just because you regard it as 'stupid' doesn't mean it's wrong.

The cutting edge one month at $X looks pretty stupid if the month later you can buy 1.5x the performance for $ 0.8X. (Hypothetically.)

 

[DrMrLordX]

I thought this was funny.

I made up my mind last week not to buy an i9 and the X299 on the release date. I'll be going 16 core Threadripper if the 10 and 12 core i9 performance falls behind. If performance of Threadripper trails, it's the 18 core i9 and an Asus Rampage VI Extreme for me in October.

It's good to have AMD back in the game.

Why not wait until October and just get the 18c chip then, assuming Intel is able to launch it on time? Not to besmirch AMD or anything, but let's face it, the 18c Intel chip will be the fastest thing out there in HEDT land.

Threadripper will be cheaper, and it'll launch sooner . . . but I don't see the 16c part being faster than Intel's 18c part in anything.

 

[frozentundra123456]

No, ONE E5-2683 14 core@2.5 turbo vs ONE 1700x 8 core@3.7 turbo. So its hard to tell the IPC difference. Too many variables. Where I disagree is that Intel has higher overclocks and higher IPC. At least based on current information and benchmarks.

Edit: And thats STOCK turbo, not overclocked on both CPU's/

Seems just the opposite to me regarding overclocking. Ryzen seems to have a hard limit of about 4.0 ghz. Even BW-E usually can reach low to mid 4ghz. I would think SKL-X on 14 nm+ would do better than that.

 

[SpoCk0nd0pe]

Do you have a source for this reliability data?

My main source for that is an acquantince of mine who has repeatedly bought large batches of mobos (100+) to supply his customers (small enterprises) with pcs. Since falues are on his bill, he also talked to his business vendor a lot. He said falue rates as a whole have fluctuations because cap quality varies over time but he had repeated bad experiences with Asus and good experiences with Gigabyte (as a side note: I don't like Gigabyte cards btw because they give too much voltage imho).

Hardware fr usually has fairly high volumes and releases rma statistics regularily.

 

[Rngwn]

Why not wait until October and just get the 18c chip then, assuming Intel is able to launch it on time? Not to besmirch AMD or anything, but let's face it, the 18c Intel chip will be the fastest thing out there in HEDT land.

Threadripper will be cheaper, and it'll launch sooner . . . but I don't see the 16c part being faster than Intel's 18c part in anything.

That's all depends on the clock speed wouldn't it? It's worth noting that 18c is "only" 12.5% more than 16c, multiply that with the IPC advantage of 10% for the total of about 24%. This 24% is the (hypothetical) maximum all-core clock speed advantage TR allowed to have over the SKL-X's 18c parts for the latter not to lose out.

In short, the 18c SKL-X should have at least 80% the clock speed of the 16c TR to beat it. The current top-tier TR sample is running at 3.4 GHz base (based videocardz), that translates to the minimum of 2.7 GHz or faster all-core clock speed for the 18c part.


https://en.wikipedia.org/wiki/List_...kylake-SP.22_.2814_nm.29_Scalable_Performance

Now, seeing that Intel have already acheived 3.0 GHz all-core boost on the 18c Xeon (Xeon Gold 6140), I would say Intel have no problem raising the clock for HEDT counterpart and surpass the TR by a wider margin.

 

[Sweepr]

All-core Turbo for Xeon Gold 6150 is 3.4 GHz with 18C and 165W TDP. Expect equal or better for Core i9-7980XE via binning.

 

[Rngwn]

All-core Turbo for Xeon Gold 6150 is 3.4 GHz with 18C and 165W TDP. Expect equal or better for Core i9-7980XE via binning.

Hmmm, that could mean up to 25-30% advantage in the Multi-threaded performance for the 18c i9. More than what I once estimated.

Oh boi...


Btw, I am curious to see if those Xeons are soldered or TIM'd. That would also somewhat influcence how much clock speed those i9s would have without overheating.

 

[Zucker2k]

That's all depends on the clock speed wouldn't it? It's worth noting that 18c is "only" 12.5% more than 16c, multiply that with the IPC advantage of 10% for the total of about 24%. This 24% is the (hypothetical) maximum all-core clock speed advantage TR allowed to have over the SKL-X's 18c parts for the latter not to lose out.

In short, the 18c SKL-X should have at least 80% the clock speed of the 16c TR to beat it. The current top-tier TR sample is running at 3.4 GHz base (based videocardz), that translates to the minimum of 2.7 GHz or faster all-core clock speed for the 18c part.


https://en.wikipedia.org/wiki/List_...kylake-SP.22_.2814_nm.29_Scalable_Performance

Now, seeing that Intel have already acheived 3.0 GHz all-core boost on the 18c Xeon (Xeon Gold 6140), I would say Intel have no problem raising the clock for HEDT counterpart and surpass the TR by a wider margin.

Why not just pitch TR against the i9 7960X? It's also 16 cores and would most likely have the better clocks, especially turbo.

 

[mikk]

I am not sure about the IPC lead. I have many Xeons E5 and Ryzens. They are equal, My E5 has more cores, my Ryzen has more clock speed, thus maybe more IPC. I will wait for benchmarks.

14 cores and 2.5 ghz vs 8 cores and 3.7 ghz. My gut says Ryzen has more IPC against this Xeon

Ryzen has Haswell IPC at best.

 

[tamz_msc]

Ryzen has Haswell IPC at best.

You need to define a workload to talk about IPC.