Support NeoGAF

[Frankfurter]

If they put in a CPU that doesn't share similar performance to say a Skylake i5, then those 6TF are going to waste IMO.

Just like Sonys 4.2 Tflops are a waste?

 

[Nachtmaer]

Hasnt AMD said that this will be somewhere between Haswell and Broadwell in performance per clock? You are right that this isn't going to be THE top of the line gaming CPU. What it could end up being is a Broadwell-E killer. As in, performance on par with the 8 core Broadwell-E but MUCH cheaper.

I've seen plenty of people do the math of where XV + 40% IPC should end up, but I can't really remember what it was. I think it was somewhere inbetween Sandy/Ivy Bridge and Haswell. We don't know if that figure means a worst or best case scenario or an average of a bunch of tests. They're most likely lowballing it a bit because they've been burnt before with giving performance estimations before launch, *cough* Bulldozer.

But yeah, if Zen ends up relatively close to BW-E at decent clockspeeds and at an attractive price, I'm all for it.

 

[napata]

If they put in a CPU that doesn't share similar performance to say a Skylake i5, then those 6TF are going to waste IMO.

Aren't they aiming for 4K native XB1 games? You don't need a better CPU for that. A jaguar will do just fine if that's their goal.

 

[mrklaw]

No necessarily. you need better than Jaguar to play a lot of the most recent games at 60fps instead of 30...

Just look at how many PS4 Pro titles are still stuck at 30fps due to Jaguar...

Xbox One games at 4K60 with upgraded visuals would need the power...

For reference, RotTR on PS4 Pro... you have 4K mode, enhanced visual mode at 1080p, and uncapped framerate mode at 1080p(which still isn't 60fps despite being at standard PS4 settings, this is due to the Jaguar CPU bottleneck...)

In order for Scorpio to roll all three modes at once... and it would need the extra horsepower from both the 6TF and Zen to do it... Jaguar just won't cut it... So, if they are sticking with Jaguary, the extra GPU power is pretty pointless since you con't get better framerates out of it...

So extra GPU power is pointless if you don't get more FPS out of it? Higher resolution, higher details, better shadows etc - all pointless unless we get 60fps?

 

[Nostremitus]

So extra GPU power is pointless if you don't get more FPS out of it? Higher resolution, higher details, better shadows etc - all pointless unless we get 60fps?

According to you any kind advancements are pointless based on your post i was responding to...

If they make it only play Xbox one games in higher resolution then anything more than jaguar is a waste IMO

If they make it only play Xbox one games in higher resolution then those 6TF are going to waste IMO

Again this feels more like people projecting their wishes onto a piece of hardware before it comes out, rather than applying any kind of logic to all the comments about it so far from both MS and AMD and the available technology in that time frame

But yes, if better actual performance isn't on the table, everything else is a bit pointless when your goal is to make better versions of current games that have performance issues...

 

[Space_nut]

So extra GPU power is pointless if you don't get more FPS out of it? Higher resolution, higher details, better shadows etc - all pointless unless we get 60fps?

There's also the fact Scorpio needs to be oculus ready. Last I checked jaguar is severely not able to be compatible with it performance wise

 

[Locuza]

Hasnt AMD said that this will be somewhere between Haswell and Broadwell in performance per clock? You are right that this isn't going to be THE top of the line gaming CPU. What it could end up being is a Broadwell-E killer. As in, performance on par with the 8 core Broadwell-E but MUCH cheaper.

AMD rendered something with Blender and used a 3 Ghz Zen (8 cores) vs. a 3 Ghz Broadwell (8 cores) and finished the Rendering in the same amount of time, with a tiny edge for Zen.

But the thing is, Zen does have only 128-Bit FP-Pipes vs. 256-Bit on Intel, 2 AGUs vs. 3, 128-Bit Datapaths to the Caches vs. 256-Bit (32 Bytes per Load/16 Bytes per Store vs. 64/32 Bytes) and a Dual-Channel vs. a Quad-Channel.

Zen would need to be an absolut genius design to outweight all the disadvantages it has on paper but of course it won't, that's also the reason why AMD said sometimes they will target 80% of the server market, because they can't penetrate the market where AVX/FMA and super high troughput is critical.

 

[mrklaw]

According to you any kind advancements are pointless based on your post i was responding to...



But yes, if better actual performance isn't on the table, everything else is a bit pointless when your goal is to make better versions of current games that have performance issues...

Except my original post that you quoted was clearly referencing the post I quoted saying that 6TF was a waste without an i5 level CPU - in other words disagreeing with that statement

 

[mrklaw]

There's also the fact Scorpio needs to be oculus ready. Last I checked jaguar is severely not able to be compatible with it performance wise

This is a wildcard, agreed. We don't have anything solid about oculus support though.

Personally if MS are serious about VR then they are more likely to go with a discrete CPU and GPU than they are to get a zen APU.

 

[Pop-O-Matic]

So extra GPU power is pointless if you don't get more FPS out of it? Higher resolution, higher details, better shadows etc - all pointless unless we get 60fps?

 

[Nostremitus]

Except my original post that you quoted was clearly referencing the post I quoted saying that 6TF was a waste without an i5 level CPU - in other words disagreeing with that statement

Then I don't know what to tell you... based on your posts you don't seem to know what you want or think is a worthwhile use of the processing power...


Edit: but then I did just have back surgery and am lying in a hospital bed on meds... I may not be following your reasoning very well.

But yes, extra GPU power does become pointless if your CPU can't feed it and becomes a bottle neck... If it's already choking on draw calls with the current Xbox One GPU, what do you think will to happen with a 6TF GPU?

 

[mrklaw]

Then I don't know what to tell you... based on your posts you don't seem to know what you want or think is a worthwhile use of the processing power...


Edit: but then I did just have back surgery and am lying in a hospital bed on meds... I may not be following your reasoning very well.

But yes, extra GPU power does become pointless if your CPU can't feed it and becomes a bottle neck... If it's already choking on draw calls with the current Xbox One GPU, what do you think will to happen with a 6TF GPU?

Sorry - I think extra GPU power is very welcome for better image quality, higher resolutions and higher detail. While I would welcome a big CPU increase I don't think it is likely, given the technology landscape at the moment. And considering MS' comments about it playing XB1 games I think they are looking at a PS Pro type machine.

Given all that, a jaguar based approach would be realistic and deliver what is wanted - XB1 games in 4K.

But because the machine doesn't yet exist, we are seeing lots of posters hoping it'll have this or that, and hoping for 60fps. We saw the same with PS Pro. Nice to want things, but at the same time I think expectations should be tempered.



(And nothing will happen with draw calls on a 6TF GPU if they aren't drawing more polygons, simply the same polygons at a higher res. I'd also expect them to do a frequency boost like pro did)

 

[dragonelite]

Sorry - I think extra GPU power is very welcome for better image quality, higher resolutions and higher detail. While I would welcome a big CPU increase I don't think it is likely, given the technology landscape at the moment. And considering MS' comments about it playing XB1 games I think they are looking at a PS Pro type machine.

Given all that, a jaguar based approach would be realistic and deliver what is wanted - XB1 games in 4K.

But because the machine doesn't yet exist, we are seeing lots of posters hoping it'll have this or that, and hoping for 60fps. We saw the same with PS Pro. Nice to want things, but at the same time I think expectations should be tempered.



(And nothing will happen with draw calls on a 6TF GPU if they aren't drawing more polygons, simply the same polygons at a higher res. I'd also expect them to do a frequency boost like pro did)

I rather have a better performing CPU if Scorpio needs to be supported when Xbox 2 comes out.

 

[Nostremitus]

Sorry - I think extra GPU power is very welcome for better image quality, higher resolutions and higher detail. While I would welcome a big CPU increase I don't think it is likely, given the technology landscape at the moment. And considering MS' comments about it playing XB1 games I think they are looking at a PS Pro type machine.

Given all that, a jaguar based approach would be realistic and deliver what is wanted - XB1 games in 4K.

But because the machine doesn't yet exist, we are seeing lots of posters hoping it'll have this or that, and hoping for 60fps. We saw the same with PS Pro. Nice to want things, but at the same time I think expectations should be tempered.



(And nothing will happen with draw calls on a 6TF GPU if they aren't drawing more polygons, simply the same polygons at a higher res. I'd also expect them to do a frequency boost like pro did)

Oh i thought you were talking about more enhancements than just higher resolution.

I personally don't think it will be Zen or that it will do 4K60, but with Microsoft's statements about how the Scorpio will be a noticeably bigger difference than PS4 Pro, they'd need more than just a GPU boost to keep from looking like blowhards.

 

[BriareosGAF]

CPU is almost entirely irrelevant to shifting a title's resolution. I can only think of two areas where it would have an effect--putting pressure on a CPU-based decompression routine as the higher resolution brings in textures with higher mips (but you shouldn't be using CPU compression on a console). Another minor involvement is if you have some sort of screen pixel based LOD metric, and shifting the LODs higher due to unadjusted LOD algorithm results in meshes with more discrete material sections being submitted to the renderer, which results in higher effective draw call throughput (this is something you have to be careful with when doing dynamic resolution even sub 4K).

These threads are almost entirely devoid of rational discussion (and I try to stay out of them but occasionally my will fails). Even Seb on B3D, who is very, very smart, vastly underestimates the amount of branchy general purpose CPU code we ran on e.g. SPUs last generation.

 

[Nostremitus]

CPU is almost entirely irrelevant to shifting a title's resolution. I can only think of two areas where it would have an effect--putting pressure on a CPU-based decompression routine as the higher resolution brings in textures with higher mips (but you shouldn't be using CPU compression on a console). Another minor involvement is if you have some sort of screen pixel based LOD metric, and shifting the LODs higher due to unadjusted LOD algorithm results in meshes with more discrete material sections being submitted to the renderer, which results in higher effective draw call throughput (this is something you have to be careful with when doing dynamic resolution even sub 4K).

These threads are almost entirely devoid of rational discussion (and I try to stay out of them but occasionally my will fails). Even Seb on B3D, who is very, very smart, vastly underestimates the amount of branchy general purpose CPU code we ran on e.g. SPUs last generation.

I was talking about games that currently struggle to maintain 30fps regardless of resolution getting better performance and higher framerate from a better CPU. You can't argue that the Jaguar isn't a bottleneck.

 

[mrklaw]

I was talking about games that currently struggle to maintain 30fps regardless of resolution getting better performance and higher framerate from a better CPU. You can't argue that the Jaguar isn't a bottleneck.

I think/hope those will be ok - I'd expect a clock boost at least

 

[Nostremitus]

I think/hope those will be ok - I'd expect a clock boost at least

Even with the PS4 Pro's clock boost, we've seen how difficult it can be to maintain 30fps .

 

[Papacheeks]

What gives you that idea? Even taking AMD's own claims of 40% IPC improvements over their previous generation completely at face value, they still wouldn't match current Intel IPC. Do you believe they will make up for that by 5 GHz clocks?

Now performance/price? They could well be competitive there. But I jut don't see a way for that to occur in absolute performance.

No more cores will be how they make up the gap. Which seeing as some leaked tests showed them beating Broadwell with fewer cores.

But they touted there will be options with more cores than what we know about.

 

[Irobot82]

What gives you that idea? Even taking AMD's own claims of 40% IPC improvements over their previous generation completely at face value, they still wouldn't match current Intel IPC. Do you believe they will make up for that by 5 GHz clocks?

Now performance/price? They could well be competitive there. But I jut don't see a way for that to occur in absolute performance.

That's what I am hoping for. If AMD's 4C/8T is say 20% slower than Intels, fine, what if their 6C/12T or 8C/16T costs the same as intels's 4C/8T. It would help a lot in games like BF1 that will soak up every core you throw at it.

 

[dogen]

CPU is almost entirely irrelevant to shifting a title's resolution. I can only think of two areas where it would have an effect--putting pressure on a CPU-based decompression routine as the higher resolution brings in textures with higher mips (but you shouldn't be using CPU compression on a console). Another minor involvement is if you have some sort of screen pixel based LOD metric, and shifting the LODs higher due to unadjusted LOD algorithm results in meshes with more discrete material sections being submitted to the renderer, which results in higher effective draw call throughput (this is something you have to be careful with when doing dynamic resolution even sub 4K).

Good points, that's exactly my take on this. Resolution is usually cpu independent, and PS4 Pro is just gonna be running PS4 games with graphics improvements for the most part. The faster cpu is just a bonus thrown in. I don't see the issue as long as some care is taken.

These threads are almost entirely devoid of rational discussion (and I try to stay out of them but occasionally my will fails). Even Seb on B3D, who is very, very smart, vastly underestimates the amount of branchy general purpose CPU code we ran on e.g. SPUs last generation.

Yeah, I don't think he ever did a ps3 game. That woulda been cool to see.