AMD THREADRIPPER VS INTEL SKYLAKE X

[muon]

So no IPC benefit then. Threadripper will destroy this.

http://www.anandtech.com/show/11550...core-i9-7900x-i7-7820x-and-i7-7800x-tested/16

 

[Kelt]

So no IPC benefit then. Threadripper will destroy this.

http://www.anandtech.com/show/11550...core-i9-7900x-i7-7820x-and-i7-7800x-tested/16

Yeh, it'll rip it a new one.

 

[gavinh87]

So no IPC benefit then. Threadripper will destroy this.

http://www.anandtech.com/show/11550...core-i9-7900x-i7-7820x-and-i7-7800x-tested/16

I like this slide

You could buy x2 ryzen 1700's versus the 7900x. Get the same amount of compiling done and whilst using less energy, cheaper motherboards and on stock coolers.

 

[humbug]

I like this slide

You could buy x2 ryzen 1700's versus the 7900x. Get the same amount of compiling done and whilst using less energy, cheaper motherboards and on stock coolers.

Thats a very bad slide for AMD, the performance there is absolutely atrocious. thats really really really bad, there is almost no difference between the cheapest 6 core and the best 8 core indicating this program just doesn't scale at all with AMD, its only scaling with Mhz, not threads.
Infact i suspect with this even the 4 core Ryzen 3 would look the same as the rest of the Ryzen chips.

If anyone ever wanted to drive home an agenda of how bad AMD's performance is this is the slide they would use as its little more than half as fast as Intel's old 8 core. that is bulldozer 2.0 if anyone was looking for such a thing.

 

[pete910]

Don't tell me, they used intels compiler

Side note, it would be nice if they showed how they derived said benchmark, as in what they used compiler wise, flags ect

 

[humbug]

Don't tell me, they used intels compiler

Side note, it would be nice if they showed how they derived said benchmark, as in what they used compiler wise, flags ect

BINGO!!!!!!!!!!

 

[faceman123]

bring on threadripper. drain the swamp amd. honeslty x299 is extremely disappointing

 

[gavinh87]

Thats a very bad slide for AMD, the performance there is absolutely atrocious. thats really really really bad, there is almost no difference between the cheapest 6 core and the best 8 core indicating this program just doesn't scale at all with AMD, its only scaling with Mhz, not threads.
Infact i suspect with this even the 4 core Ryzen 3 would look the same as the rest of the Ryzen chips.

If anyone ever wanted to drive home an agenda of how bad AMD's performance is this is the slide they would use as its little more than half as fast as Intel's old 8 core. that is bulldozer 2.0 if anyone was looking for such a thing.

Agreed, that program does not favour AMD at all.
However this is a worst case scenario and my point would still stand. I could get the same performance using x2 desktops for the same amount of money.

 

[N19h7m4r3]

Thats a very bad slide for AMD, the performance there is absolutely atrocious. thats really really really bad, there is almost no difference between the cheapest 6 core and the best 8 core indicating this program just doesn't scale at all with AMD, its only scaling with Mhz, not threads.
Infact i suspect with this even the 4 core Ryzen 3 would look the same as the rest of the Ryzen chips.

If anyone ever wanted to drive home an agenda of how bad AMD's performance is this is the slide they would use as its little more than half as fast as Intel's old 8 core. that is bulldozer 2.0 if anyone was looking for such a thing.

Thing is, in Linux and GCC Ryzen doesn't have that issue. This is launch day reviews results and look how well it does there.
I wonder what compiler Anandtech was using.

http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-1700&num=4

 

[humbug]

Thing is, in Linux and GCC Ryzen doesn't have that issue. This is launch day reviews results and look how well it does there.
I wonder what compiler Anandtech was using.

http://www.phoronix.com/scan.php?page=article&item=amd-ryzen-1700&num=4

One carefully chosen.

 

[CAT-THE-FIFTH]

Wow:

https://www.pcper.com/reviews/Proce...X-Processor-Review/Thread-Thread-Latency-and-

Inter-chip latency is pretty high!!

 

[Rroff]

Gonna be an interesting story I think with some of the upcoming CPUs and inter-chip latency. For big data tasks it isn't too much of an issue but for stuff like gaming can be another matter.

EDIT: Oh they do actually touch on that later on in the article:

First, the 7900X is never faster than the 6950X in the gaming tests we ran, although it is able to match performance in Ashes, Deus Ex: Human Revolution, Ghost Recon Wildlands, and GTA V (mostly). However, take a look at Civ 6, Far Cry Primal, or even Hitman and Rise of the Tomb Raider: these show the new 7900X as slower than the 6950X in gaming, even with its substantial clock speed advantages. How? Remember back to this graph I showed you on thread-to-thread communication latency?

It is something people probably will want to keep a close eye on especially if they have multi die CPUs in mind for gaming.

In a slightly different and more complex way it is why you don't tend to see GPUs forever rehashing the same architecture with ever increasing SP counts.

 

[N19h7m4r3]

Wow:

https://www.pcper.com/reviews/Proce...X-Processor-Review/Thread-Thread-Latency-and-

Inter-chip latency is pretty high!!

Gonna be an interesting story I think with some of the upcoming CPUs and inter-chip latency. For big data tasks it isn't too much of an issue but for stuff like gaming can be another matter.

EDIT: Oh they do actually touch on that later on in the article:



It is something people probably will want to keep a close eye on especially if they have multi die CPUs in mind for gaming.

In a slightly different and more complex way it is why you don't tend to see GPUs forever rehashing the same architecture with ever increasing SP counts.

It's the cause of RyZen's lower gaming performance as well, yet even with that competes well against Broadwell, despite lower clocks.

Here Skylake X is losing to Broadwell-E while being clocked much higher. It's like an IPC decrease in gaming nearly.

To think people and even Anandtech thought this new cache structure could provide nearly up to a 13% IPC increase. Yeah... So far it seems Skylake X only beats Broadwell-E in compute because of its drastically increased clock rates, and then losing in games despite higher clocks. So one can say a decrease in IPC nearly.

http://www.anandtech.com/show/11464...ng-18core-hcc-silicon-to-consumers-for-1999/3

The good element of this design is that a larger L2 will increase the hit-rate and decrease the miss-rate. Depending on the level of associativity (which has not been disclosed yet, at least not in the basic slide decks), a general rule I have heard is that a double of cache size decreases the miss rate by the sqrt(2), and is liable for a 3-5% IPC uplift in a regular workflow. Thus here’s a conundrum for you: if the L2 has a factor 2 better hit rate, leading to an 8-13% IPC increase, it’s not the same performance as Skylake-S. It may be the same microarchitecture outside the caches, but we get a situation where performance will differ.

 

[=XDC=FluphyBunny]

So for 6 core wait for Coffee Lake and for multi-threaded wait and see what Threadripper brings.

Does that sound about right?

 

[jaffveyda]

So for 6 core wait for Coffee Lake and for multi-threaded wait and see what Threadripper brings.

Does that sound about right?

Nope, smash your head against a wall get concussion and order bundle twice expensive than AMD with little to no benefit at all. sarcasm off

 

[N19h7m4r3]

So for 6 core wait for Coffee Lake and for multi-threaded wait and see what Threadripper brings.

Does that sound about right?

If CoffeeLake brings some nice IPC gains and 6 cores, 12 threads, I wouldn't touch the lower end Skylake X chips at all.

Not with them performing worse in games than Broadwell and Kabylake

 

[DragonQ]

It's the cause of RyZen's lower gaming performance as well, yet even with that competes well against Broadwell, despite lower clocks.

Here Skylake X is losing to Broadwell-E while being clocked much higher. It's like an IPC decrease in gaming nearly.

To think people and even Anandtech thought this new cache structure could provide nearly up to a 13% IPC increase. Yeah... So far it seems Skylake X only beats Broadwell-E in compute because of its drastically increased clock rates, and then losing in games despite higher clocks. So one can say a decrease in IPC nearly.

http://www.anandtech.com/show/11464...ng-18core-hcc-silicon-to-consumers-for-1999/3

Remember that code optimisations could improve matters, just as with Ryzen.

If CoffeeLake brings some nice IPC gains and 6 cores, 12 threads, I wouldn't touch the lower end Skylake X chips at all.

Not with them performing worse in games than Broadwell and Kabylake

Why do people still expect IPC gains? It's a hex core Core i7-7700K.

 

[Panos]

Must admit the i7 7820 looks like a bit of a monster and only 80 odd quid more than the Ryzen 7 8 core, as I currently have a 5930k i'm not looking to upgrade for a good while yet, but I'm looking forward to what threadripper has to offer.

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

 

[N19h7m4r3]

Why do people still expect IPC gains? It's a hex core Core i7-7700K.

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

 

[Tom B]

Unfortunately it looks like Intel is hitting the limits of both frequency and IPC. I can't imagine we'll start seeing an improvement in single core performance until Intel move to 10nm