DX12 and 6+ CPU cores: savior or irrelevant?

[DavieJG]

---

In contradiction to the above if you look at this video review of Ashes of the singularity, you find the 6700k faster than the 5960x.

Does it have to be physical cores not hyperthreaded cores? If it's the latter then the 6700k has 8 cores at 4ghz. The 5960x has 16 cores but sees very diminishing returns on the extra cores and gets beaten out by a 33% clock speed disadvantage.

If you clock both to the maximum the difference will be 4.5ghz vs 5ghz which is only 11% and at which point is expect the diminishing returns from the extra cores would give the 5960x the win.

I'd agree that if you have an i7 from sandy onwards there's no point to upgrade based on what we've seen. I5 results would be interesting though.

 

[opethdisciple]

Does it have to be physical cores not hyperthreaded cores? If it's the latter then the 6700k has 8 cores at 4ghz. The 5960x has 16 cores but sees very diminishing returns on the extra cores and gets beaten out by a 33% clock speed disadvantage.

If you clock both to the maximum the difference will be 4.5ghz vs 5ghz which is only 11% and at which point is expect the diminishing returns from the extra cores would give the 5960x the win.

I'd agree that if you have an i7 from sandy onwards there's no point to upgrade based on what we've seen. I5 results would be interesting though.

I've seen contradicting evidence on this.

1. I've seen them say that DX12 likes physical cores not hyperthreaded ones.
2. But on the other hand, in that video it shows the 6700k as faster than 5960x.

 

[andybird123]

PCPer did the tests all at stock settings, so in their tests the 6700k will have a massive clock speed advatnage, clocked to the same the 8 core pulls ahead or becomes GPU bound

one of the sites used a 5960X for all of their tests and clocked it to different levels / turned off cores and you could very clearly see that the game scaled up to a certain point and then just stopped as it became GPU bound, at which point you need more GPU's to give the CPU something more to do

it is shaping up that scaling on multi GPU systems should be much easier in the future, but for single GPU users it still looks like pretty much any mid range 4 core will give good results

but yes, DX12 likes physical cores and clock speed, hyperthreading does increase performance by a small amount, but it's only a few percent

 

[joeyjojo]

I posted a thread about DX12 and hardware, but it had no responses.

Here is a link to a presentation and a comparison between a game run on DX11 & DX12. Mentions quite a bit about it, and it can't be far away from launch if they are able to show it like this.

http://www.eurogamer.net/articles/digitalfoundry-2015-why-directx-12-is-a-gamechanger

The images in that gallery are very interesting (10x number of draw calls in DX12 compared to 11). Thanks for sharing.

 

[David Bisset]

I've seen contradicting evidence on this.

1. I've seen them say that DX12 likes physical cores not hyperthreaded ones.
2. But on the other hand, in that video it shows the 6700k as faster than 5960x.

Both have 50% hyperthreaded cores so I don't understand why it would shed any light on here. Different numbers of both, different clockspeeds, different instruction sets all make a pretty muddled picture. We'd need to compare a 5960x with hyperthreading switched off with a 4790k clocked to the same speed with hyperthreading on to be able to compare.

 

[Simon Belmont]

Agreed !!!! Agreed!!!!!!!!!!!! AGREEEEDDDDDD!!!!!

Don't build up his ego higher then it already is.

 

[C64]

By the time some mainstream games are dx12 and need to be, x99 will be old hat the way I see it.

The true was the noctua d15 of it's day not a rubbish h100i, AIO water coolers suck.

Everyone definitely did not have a true copper, hyper 212 evo maybe.

 

[TheF34RChannel]

Cores or no X99 is faster in games at the high end. Lashings of extra L2 help, as do various other things. Including the extra cores in some games.

Can you provide some prove of this? Because all trustworthy reviews I have checked (heck, even the more untrustworthy ones) recommend/say the exact opposite.

Both have 50% hyperthreaded cores so I don't understand why it would shed any light on here. Different numbers of both, different clockspeeds, different instruction sets all make a pretty muddled picture. We'd need to compare a 5960x with hyperthreading switched off with a 4790k clocked to the same speed with hyperthreading on to be able to compare.

Sure but isn't that the point, to compare the CPUs for what they are? Such a comparison would certainly answer my questions.

By the time some mainstream games are dx12 and need to be, x99 will be old hat the way I see it.

My thoughts exactly! I will not spend a fortune on a platform (X99) based on the hope of forum users from different fora (no offense) that have placed their hope in an idea that yet has to prove itself (see earlier comment below my first quote). I'm sorry but I just don't see it. What I do see is the current win in higher frequencies.

 

[Orangey]

My thoughts exactly! I will not spend a fortune on a platform (X99)

Skylake 550
X99 600

Yeah a real king's ransom.

If money is that tight for you why are you wasting money on a platform with no discernable uplift to begin with? In fact it would make more sense (if any) to go used X99 instead.

 

[TheF34RChannel]

Skylake 550
X99 600

Yeah a real king's ransom.

If money is that tight for you why are you wasting money on a platform with no discernable uplift to begin with?

I'm not from the UK, as can be seen above each of my posts Also, you assume (or I've created the misunderstanding) that I'm upgrading, which I'm not

Additionally, I'd still like to see some of those reviews that back up your recommendations - I'm all about the proof (and again, no offense mate!)

 

[Tephnos]

I'm not from the UK, as can be seen above each of my posts Also, you assume (or I've created the misunderstanding) that I'm upgrading, which I'm not

Additionally, I'd still like to see some of those reviews that back up your recommendations - I'm all about the proof (and again, no offense mate!)

A bit in denial mate, X99 is far better value for money because of those extra cores which help in everything else.

Paying so much for Skylake for a meagre improvement over gaming compared to X99 is hilarious at best. Furthermore, when DX12 is out you're still using 4 cores, when DX12 can use 6.

 

[Gazb]

Personally I think when I went from my i7 920 to the 6 core Xeon it really helped make games feel smoother when I still had my web browser with a lot of tabs open etc in the background.
So while it may not directly increase frame rates I do think it helps when you do multiple things as well as gaming.

 

[David Bisset]

Sure but isn't that the point, to compare the CPUs for what they are? Such a comparison would certainly answer my questions.

Sure you can compare them, doesn't tell you anything about DX12 liking cores or threads though which was what I was responding to.

My thoughts exactly! I will not spend a fortune on a platform (X99) based on the hope of forum users from different fora (no offense) that have placed their hope in an idea that yet has to prove itself (see earlier comment below my first quote). I'm sorry but I just don't see it. What I do see is the current win in higher frequencies.

I've been advising X99 over the Skylake i7 because it's not extra expense so it doesn't matter if there turns out to be almost no gains at all - once skylake gouging drops off then this will stop being true. Skylake does have higher stock frequencies than a 5820k but I'd expect an enthusiast grade CPU to be clocked up anyway and they top out near enough the same frequency on conventional cooling.

 

[MichaelM]

How would Intel's secrecy about skylakes technical capabilities effect this discussion? Rumors of inverse hyperthreading or some other features yet to be unveiled that programmers could start taking advantage of?

 

[Tephnos]

How would Intel's secrecy about skylakes technical capabilities effect this discussion? Rumors of inverse hyperthreading or some other features yet to be unveiled that programmers could start taking advantage of?

Probably not much. Get x99 now, upgrade to Skylake-E if rumours are true.

 

[Telecaster]

I think people are pinning their hopes too much on DX12 and the utilisation of multiple cores, I have a funny feeling it's not going to be as night/day, black/white as that, or whatever other cliché you want to use to describe it.

 

[Telecaster]

I mean, have we had any definitive demos showing DX12 utilisation of extra cores showing massive improvements over standard 4 core CPUs?

 

[Gazb]

I mean, have we had any definitive demos showing DX12 utilisation of extra cores showing massive improvements over standard 4 core CPUs?

Well there is the DX12 benchmark (Ashes of Singularity) that's recently come out.
One review has the 6700K 'slightly' beating the 5960X, but even that isn't black and white.
The 6700K has a 1GHz speed advantage and plus it's 5% or whatever it is IPC improvement over the haswell-E cpu so that will be counter acting the extra cores to an extent.

Plus, people on here I'm the benchmark thread have been demonstrating that the benchmark is using all available cores evenly.
Which has raised questions as to how 6700K is performing the way it is.

More comparitive testing being needed seems to be the consensus.
(Maybe a comparison with both the 6700K and 5960X running all cores at 4GHz would be the best way to do it.)

 

[opethdisciple]

I think people are pinning their hopes too much on DX12 and the utilisation of multiple cores, I have a funny feeling it's not going to be as night/day, black/white as that, or whatever other cliché you want to use to describe it.

Chalk and Cheese?