AMD THREADRIPPER VS INTEL SKYLAKE X

[Hagar]

I was slightly surprised that Intel have had to respond so forcefully to AMD. Having such a huge market share. They cannot be too worried by the few enthusiast overclocker's, they must have had some serious feedback from OEM's and server suppliers to give them pause.

I cannot really define their strategy either. They do seem to be catering for the enthusiast market at some cost to their bottom line and definitely to the buyers however to do this they migrate onto the professional platform and blur the lines beween gaming and productivity.

 

[Spud21]

Your pricing is off.
R7 1700 goes for £280. 7820K £600. That's more than double.

The 1800x is 450 odd at ocuk, and the 7820x is on pre-order for 539 so it's not over double, and that's the choice I would be making if i were looking, if I was going to drop down to the R7 1700 then id compare that to the 7800x which is only hex core but will still out perform the R7 and comes in at 350, so a fairly modest increase over the AMD chip in terms of price.

 

[DragonQ]

Since Intel was claiming improvements.
Then again, they claimed the same with Kabylake over Skylake, and the only real difference was clocks.

Exactly. I'm pretty sure someone pointed out the small print for their outlandish >30% performance gain claim was comparing two very low powered parts. Top-end parts will not have anywhere near that much. In fact I'd be surprised if the i7-8700K or whatever the 6c/12t part is called will even match the stock clocks of the i7-7700K.

 

[N19h7m4r3]

Exactly. I'm pretty sure someone pointed out the small print for their outlandish >30% performance gain claim was comparing two very low powered parts. Top-end parts will not have anywhere near that much. In fact I'd be surprised if the i7-8700K or whatever the 6c/12t part is called will even match the stock clocks of the i7-7700K.

Even if it only matches, or slightly exceeds their 7800K it's a better buy. The platform cost will be cheaper for similar if not a smidge better performance.

Kabylake X and "low-end" Skylake X just don't look that appealing with Coffeelake also due this year.

 

[zornyan]

Your pricing is off.
R7 1700 goes for £280. 7820K £600. That's more than double.

7820x is £530 not £600 so less than double.

 

[N19h7m4r3]

Alienware claiming they'll be overclocking all the Threadripper cores!

 

[CAT-THE-FIFTH]

The 1800x is 450 odd at ocuk, and the 7820x is on pre-order for 539 so it's not over double, and that's the choice I would be making if i were looking, if I was going to drop down to the R7 1700 then id compare that to the 7800x which is only hex core but will still out perform the R7 and comes in at 350, so a fairly modest increase over the AMD chip in terms of price.

The Ryzen 7 1700 is under £300 and the Ryzen 7 1700X is well under £400. In the end Intel could offer a lower clocked 8C/16T version for those who don't mind overclocking it a bit more,but that is the fault of Intel product segmentation not AMD.

Edit!!

Also if AMD can sell an 8C/16T CPU for well under £300 it makes me wonder how much the entry level 12C/16T and 16C/32T models will cost.

Considering that the Rzyen die is only 60% the size of the one used in the Core i7 7820X I think AMD is not at all in a bad situation.

 

[Chrisc]

Alienware claiming they'll be overclocking all the Threadripper cores!

Thats really interesting, I wonder what kind of clock speeds we will be looking at. . .

 

[subbytna]

The 1800x is 450 odd at ocuk, and the 7820x is on pre-order for 539 so it's not over double, and that's the choice I would be making if i were looking, if I was going to drop down to the R7 1700 then id compare that to the 7800x which is only hex core but will still out perform the R7 and comes in at 350, so a fairly modest increase over the AMD chip in terms of price.

The 1800x is just a slightly overclocked 1700 with marginally higher turbo and the 1700 is £320. The 7820x is £540. Looking at it like that then the cost of the two chips is a huge deciding factor too. You're not really "dropping down" to anything given the 1800x and 1700 are the same silicon.

 

[N19h7m4r3]

One carefully chosen.

I was looking into it more, after some folks at WCCFT were raging on about how Ryzen is useless for programmers and compiling now.

Seems with a simple tick of a box, you can use the Intel Compiler in Visual Studio Community 2015; the version Anandtech used.
https://software.intel.com/en-us/articles/installing-visual-studio-2015-for-use-with-intel-compilers

Considering Ryzen performs worse than a Stock Ivy-Bridge 6 core in their compiling test, it makes me a little suspicious. Since on GCC tests from Phoronix and ServerTheHome Ryzen is doing extremely well against Intel.

 

[N19h7m4r3]

Thats really interesting, I wonder what kind of clock speeds we will be looking at. . .

As do I!

I also want to know how well those new Noctua coolers will perform on it. I'm eyeing up Threadripper now for my next workstation. I want to see how it competes with Skylake X and Broadwell-E.

Heck, if neither are a big jump up, I might just snag a second hand 5960X/6900K or 6950X and plop that into my motherboard; then wait another year or two.

As it stands though, the 7900X is not doing too well in Tom's ( I Love Intel )Hardware review for my friend's and my needs.
I actually expected the 7900X to do far batter in these types of tests, especially the handbrake ones.

http://www.tomshardware.com/reviews/intel-core-i9-7900x-skylake-x,5092.html

 

[Tom B]

There does seem to be something up with Windows multicore performance with Ryzen:

Geekbench 4 1800X @ 4.0ghz Windows http://browser.primatelabs.com/v4/cpu/2027665 - 22722 multicore
Geekbench 4 1700 @ 4.05ghz Linux http://browser.primatelabs.com/v4/cpu/2009407 - 27138 multicore

I know that different schedulers and kernels make a difference (and in fact newer kernels than the one shown there offer some performance increases for ryzen) but windows shouldn't really be 5000 points behind linux in the multi core test. Yes there is a 0.05ghz clock difference, I don't think that accounts for it. Doesn't seem to be RAM either, if you look the Windows test shows a higher RAM bandwidth.

Could explain the difference between the Visual Studio and GCC compilers. Would be interesting to see ryzen vs intel GCC performance on Windows.

 

[B1gbeard]

As do I!

I also want to know how well those new Noctua coolers will perform on it. I'm eyeing up Threadripper now for my next workstation. I want to see how it competes with Skylake X and Broadwell-E.

Heck, if neither are a big jump up, I might just snag a second hand 5960X/6900K or 6950X and plop that into my motherboard; then wait another year or two.

As it stands though, the 7900X is not doing too well in Tom's ( I Love Intel )Hardware review for my friend's and my needs.
I actually expected the 7900X to do far batter in these types of tests, especially the handbrake ones./QUOTE]

What sort of performance were you expecting? It's beating everything including a 6950 fairly well.

Heh my question is in the quote for some reason and won't come out!

 

[humbug]

I was looking into it more, after some folks at WCCFT were raging on about how Ryzen is useless for programmers and compiling now.

Seems with a simple tick of a box, you can use the Intel Compiler in Visual Studio Community 2015; the versio Anandtech used.
https://software.intel.com/en-us/articles/installing-visual-studio-2015-for-use-with-intel-compilers

Considering Ryzen performs worse than a Stock Ivy-Bridge 6 core in their compiling test, it makes me a little suspicious. Since on GCC tests from Phoronix and ServerTheHome Ryzen is doing extremely well against Intel.

Meh, these people are not doing themselves any favours, people who know what they are looking at immediately lose faith in the reviewer, actually the sort of people to whom this matters don't go to places like Anand or Toms Hardware for their tech data anyway, they are not seen as knowledgeable or remotely competent, they exist as a profit magazine aim at the lowest conmen denominator readers for maximum clicks.

Its not necessarily that they are dodgy, its simply that they are generating a lot of hype around their magazine website by subtly dropping these hyperbolic bombshells for the 95% to get exited over.

 

[N19h7m4r3]

What sort of performance were you expecting? It's beating everything including a 6950 fairly well.

It's barely 20% faster at 4Ghz All core boost ( 4.5Ghz dualcore boost) in Handbrake than a stock 1800X ( 3.7Ghz all core boost, 4.1Ghz single core boost), at double the price; and essentially tied in 3DSMax for 100% the CPU cost.
So 20% faster in some, and tied in others, despite 25% more cores.

Intel were claiming a 10% improvement over Broadwell-E, and Anandtech were talking about an 8-13% IPC increase over normal Skylake-S.
Then there's the issue of Skylake X performing worse overall in games than Broadwell-E according to PCPer. I just expected more really. Not just a factory overclocked bigger Skylake, it's like the IPC gaines disappeared thanks to that new Meshed Cache and Inter-core latency.
Especially for the not only the CPU cost, but also the platform cost including motherboard; never mind the thermal issues.

I'll keep my eye on it, and wait for Threadripper before deciding what to move to, if I move to either in the end.

 

[B1gbeard]

It's barely 20% faster at 4Ghz All core boost ( 4.5Ghz dualcore boost) in Handbrake than a stock 1800X ( 3.7Ghz all core boost, 4.1Ghz single core boost), at double the price; and essentially tied in 3DSMax for 100% the CPU cost.
So 20% faster, and tied, despite 25% more cores.

Intel were claiming a 10% improvement over Broadwell-E, and Anandtech were talking about an 8-13% IPC increase over normal Skylake-S.
Then there's the issue of Skylake X performing worse overall in games than Broadwell-E according to PCPer. I just expected more really. Not just a factory overclocked bigger Skylake, it's like the IPC gaines disappeared thanks to that new Meshed Cache and Inter-core latency.
Especially for the not only the CPU cost, but also the platform cost including motherboard; never mind the thermal issues.

I'll keep my eye on it, and wait for Threadripper before deciding what to move to, if I move to either in the end.

Maybe to a certain extent it's us. I'm looking at the numbers you posted and at the moment the 7900X is the fastest CPU available for those tasks and 20% seems a fair bit tbh. Even the gaming benchmarks I have looked at has the 7900X at the top mostly. It boosts higher than a 6950 and still costs more than Ryzen but when you look at the numbers I'm not seeing the disaster some are making it out to be.

I'll look to Threadripper too before deciding.

 

[humbug]

@N19h7m4r3 actually you know what, tell the flocks at WCCFT to get a grip, tell them no one with anything other than an Intel CPU would use Intel's compiler, that would just be daft, and yes if they ask; Anand are daft.

 

[N19h7m4r3]

Maybe to a certain extent it's us. I'm looking at the numbers you posted and at the moment the 7900X is the fastest CPU available for those tasks and 20% seems a fair bit tbh. Even the gaming benchmarks I have looked at has the 7900X at the top mostly. It boosts higher than a 6950 and still costs more than Ryzen but when you look at the numbers I'm not seeing the disaster some are making it out to be.

I'll look to Threadripper too before deciding.

Aye, it's fastest for many things, but the platform cost of that 20% far outweighs it for me.

I recently build my old man a little solidworks, and computational workstation using a Ryzen 1600, and the entire thing cost under £700.
For myself, I'm a price to performance guy usually, so 20% more performance for 25% more cores than 1800X in my use cases is a bit iffy for me, especially since thermals( HSF/AIO costs), and motherboard costs also come into play.

For me looking at it clock for clock the 7900X just looks like an overclocked 6950X with bad thermal interface on more expensive motherboard at the moment.
A shame OC3D didn't include the 6950X in the rest of their tests, although once they get a 7900X I expect they'll compared both stock and OC.

I have to admit though; I haven't had so much to read, and enjoy regarding CPUs in nearly a decade. It's an exciting time for sure; and I'm glad there's some decent competition again.
To think the original slides for Skylake X and Basin-Falls capped out at 10 cores, and now we have Intel actually planning on 18 Cores to stay ahead.

Wonderful time to be a tech enthusiast.

@N19h7m4r3 actually you know what, tell the flocks at WCCFT to get a grip, tell them no one with anything other than an Intel CPU would use Intel's compiler, that would just be daft, and yes if they ask; Anand are daft.

Surprisingly after linking some charts, and a few comments I stop getting replies. I almost never engage with people there, it's a truly messed up place full of trolls. I don't even do compiling or programming but the circlejerk about that really didn't sit well with me.

 

[humbug]

The internet needs somewhere to put them all, WCCF is the perfect place for them.

 

[DragonQ]

I have to admit though; I haven't had so much to read, and enjoy regarding CPUs in nearly a decade. It's an exciting time for sure; and I'm glad there's some decent competition again.
To think the original slides for Skylake X and Basin-Falls capped out at 10 cores, and now we have Intel actually planning on 18 Cores to stay ahead.

Wonderful time to be a tech enthusiast.

It's strange because it seems like tech reviewers have collectively forgotten what it's like to have a competitive industry. They are used to rolling out their test Windows installs, updating chipset drivers, and running the same bunch of programs as last year. They show the 5% bigger bar in the graph and praise be to Intel, job done.

Now there's actual competition they're slowly realising there's actually a lot more involved here. How does Ryzen's infinity fabric work? How does RAM speed and latency work with it? What happens on Linux? What programs are badly optimised for non-Intel CPUs? What compilers should we use for each system? What GPUs work best with different CPUs? What's more important, average or minimum frames? etc.

In all honesty it's impossible to come to a matter-of-fact solution with so many variables. All you can do is choose the few that are most important to you and test (or find other people's tests for) those variables to make a decision.