AMD CPU speculation... and expert conjecture

[juanrga]

Just because a Nexus 7 does the same thing as a Ipad doesn't make them the same product. 4+4=8 and 2+6= 8 same outcome however different.

Anyways its still a win for Amd and everyone else. Nvidia had to do nothing really to even get it.

I don't think it is a net win for AMD. They had to move away from the original strategy by 180 degrees.

 

[gamerk316]

On the Mantle/Vulkan thing, let me be clear: Vulkan is going to be slower then Mantle for no other reason then because it's vendor neutral. It's going to be one step higher.

 

[-Fran-]

Just because a Nexus 7 does the same thing as a Ipad doesn't make them the same product. 4+4=8 and 2+6= 8 same outcome however different.

Anyways its still a win for Amd and everyone else. Nvidia had to do nothing really to even get it.

I don't think it is a net win for AMD. They had to move away from the original strategy by 180 degrees.

For AMD to have done a 180°, they would have said something along the lines of "we'll keep MANTLE away from everyone else and no one will ever know what it's about", but they've been very vocal about sharing it with everyone.

As usual, you're being dramatic on your take. They do have to slow down with MANTLE, since DX12 and OGLN (Vulkan) are re-using parts of the API's concepts, so they don't want to compete 1:1 with either for good reason.

Now, I do agree with you on 1 single point: I still don't see where AMD can net a profitable win with this.

Cheers!

 

[juanrga]

On the Mantle/Vulkan thing, let me be clear: Vulkan is going to be slower then Mantle for no other reason then because it's vendor neutral. It's going to be one step higher.

DX12 is vendor neutral, but faster than Mantle even on GCN hardware!

 

[gamerk316]

On the Mantle/Vulkan thing, let me be clear: Vulkan is going to be slower then Mantle for no other reason then because it's vendor neutral. It's going to be one step higher.

DX12 is vendor neutral, but faster than Mantle even on GCN hardware!

OS specific, so you can extract a lot of efficiencies OS side.

 

[juanrga]

Just because a Nexus 7 does the same thing as a Ipad doesn't make them the same product. 4+4=8 and 2+6= 8 same outcome however different.

Anyways its still a win for Amd and everyone else. Nvidia had to do nothing really to even get it.

I don't think it is a net win for AMD. They had to move away from the original strategy by 180 degrees.

For AMD to have done a 180°, they would have said something along the lines of "we'll keep MANTLE away from everyone else and no one will ever know what it's about", but they've been very vocal about sharing it with everyone.

Original: Mantle will be ported to linux and apple
Final: Mantle will be windows exclusive

Original: Mantle will be open to everyone
Final: Mantle will be available only to our partners

Original: New devs would chose Mantle as base API because Mantle-->DX12 is easier than DX11-->DX12
Final: New devs would not target Mantle, but use DX12/Vulkan directly.

Original: Mantle is a complete success. Dozens of new Mantle games will be released before the end of 2014
Final: No word on the promised games.

Original: ...
Final: ...

Maybe it is not a 180 degree spin, maybe it is only a 172.6 degree, but you get the point.

 

[-Fran-]

Original: Mantle will be ported to linux and apple
Final: Mantle will be windows exclusive

Original: Mantle will be open to everyone
Final: Mantle will be available only to our partners

Original: New devs would chose Mantle as base API because Mantle-->DX12 is easier than DX11-->DX12
Final: New devs would not target Mantle, but use DX12/Vulkan directly.

Original: Mantle is a complete success. Dozens of new Mantle games will be released before the end of 2014
Final: No word on the promised games.

Original: ...
Final: ...

Maybe it is not a 180 degree spin, maybe it is only a 172.6 degree, but you get the point.

There's a lot of silver lining on what you're saying. For example:

"Mantle will be open to everyone" -> They're releasing the full API later on this month: http://www.phoronix.com/scan.php?page=news_item&px=AMD-Mantle-Docs-Coming

What hey said, they'll keep supporting those who have chosen to stick with MANTLE before hearing from MS'es DX12 and Vulkan.

"New devs would chose Mantle as base API because Mantle-->DX12 is easier than DX11-->DX12" -> I'd like a direct quote on that one, please. If you think that's what they meant, I have another reading for you: Since DX12 is a behemoth and Vulkan is basically MANTLE, please use either for your own convenience. The "easier to develop" was from the Marketing slide you put some while ago and I found ridiculous.

"Mantle is a complete success. Dozens of new Mantle games will be released before the end of 2014" -> When there are a lot of Devs that took the API you created and started working with it, yeah, it is a success at the time. If stuff happens down the road with the development process, that's not something AMD can get it's hand dirty on. I'm really surprised you want to put the blame on AMD for that.

And last, but not least:
"Mantle will be ported to linux and apple" -> Apple said no, you want AMD to shove it down Apple's throat or something? And once Khronos wanted to take a look at MANTLE and continue with Vulkan, why even bother on making a Linux port when Khronos will do it for them?

The more you say about AMD, Juan, the more butt-hurt it sounds. You're twisting the words badly. I won't say I'm not doing a positive spin, but what you're doing is like a hate crime, haha.

Cheers!

 

[8350rocks]

See, juanrga, AMD delivered, they just did it in a round about way. Mantle is THE open source API you were looking for on Linux.

Negative, Vulkan is basically mantle, as I said above...AMD just handed it over to somebody who can do more with it...

Vulkan must be the API I was expecting, but it is not Mantle:

we need to be clear that Vulkan is not Mantle, Mantle was used to bootstrap the process and speed its development

This quote from Kronos group president (Nvidia vicepresident) is also relevant:

Any hardware capable of supporting OpenGL ES 3.1 will be capable of supporting Vulkan.

Any modern hardware would be Vulkan compatible, which would help popularization. Here a partial list of compatible hardware

http://en.wikipedia.org/wiki/OpenGL_ES#OpenGL_ES_3.1_2

What Does The Future Hold For AMD's Mantle?

http://www.tomshardware.com/news/amd-mantle-directx-12-vulkan,28676.html

Juan, the functions in Vulkan are all the same functions in Mantle with different extensions down to even having the same names. Check the link above. They literally just modded a few things and boom...Mantle is Vulkan.

Also @Yuka:

Glad to see that come to light. VR has been something they were working on for a while.

 

[gamerk316]

In the meantime:

http://www.techspot.com/review/970-homeworld-remastered-benchmarks/page4.html

i3 4130 > FX-9590
Pentium G3320 > FX-8350

Game scales to two cores (as expected of an AAA game), and loves per-core performance (as expected in non-GPU bottlenecked situations).

 

[gamerk316]

And more Vulkan details:

http://www.pcper.com/reviews/General-Tech/GDC-15-What-Vulkan-glNext-SPIR-V-and-OpenCL-21

 

[Ranth]

In the meantime:

http://www.techspot.com/review/970-homeworld-remastered-benchmarks/page4.html

i3 4130 > FX-9590
Pentium G3320 > FX-8350

Game scales to two cores (as expected of an AAA game), and loves per-core performance (as expected in non-GPU bottlenecked situations).

Though i7-5960x @ 3GHz > i7 4770k @ 3.5GHz, seems kinda odd does it not?

 

[blackkstar]

does the r9 290x perform poorly at 1080p? no, and with newer driver updates, it'll likely get better. it's performance at 4K doesn't look bad either.

Raw performance: At 1080p the 290X is about 20% slower than 980. At 4K the 290X is about 17% slower. The gap is much bigger when looking at performance per watt

https://www.techpowerup.com/reviews/Gigabyte/GeForce_GTX_980_G1_Gaming/27.html

The AMD card performs poorly at resolutions used by most gamers. Releasing a 8GB version of the 290X and comparing it with Nvidia 4GB version on 8K resolutions used by 0.00% of gamers is not unfair, it is ridiculous.

re-read the review posted by you here,
you will see that even 290x have maxed out every game at 1080p (constant 60fps)
so what to do now, just improve efficiency and keep 1080p for forever gamer's choice ?
no, indeed push it to next level that is 4k (8k using vsr/dsr)
so what amd is doing is progressing and now trying to jump to next , and what nvidia is doing is 300fps on 1080p 60hz monitor
and amd is not alone, all display manufacturer are pushing higher resolution even on mobiles where we have reached to the limit of eye and battery is more important than display
even current 1080p owners can do 4k on their 1080p monitor using vsr/dsr
and 290x wasn't meant to compete with 980 but it is still competing

btw i posted that 8k* review link to see what you will say about this because you want us to game on apu
now it looks like that you want us to stick to 1080p so that your apu look good and can kill dgpus because dgpus will sit idle at 1080p resolution

Again its the same argument why buy something like a 290X or 980 for 4K and expect it to work? You need SLI or crossfire and a 290X uses 300 watts vs 190 for a 980.

https://tpucdn.com/reviews/NVIDIA/GeForce_GTX_980/images/perfrel_3840.gif

https://tpucdn.com/reviews/NVIDIA/GeForce_GTX_980/images/perfrel_1920.gif

Also a 970 is just about equal at 4K a res 1% of gamers play on 1080P its 5% or so overall the same.

Here is power consumption as stated before neither card is good enough for 4K you would need 2
https://tpucdn.com/reviews/NVIDIA/GeForce_GTX_980/images/power_peak.gif

here is a comparison in SLI vs crossfire some is faster on the 290X crossfire others are faster with SLI(although the 290X looks like the winner in performance)
http://www.guru3d.com/articles_pages/geforce_gtx_970_sli_review,21.html

Now again keep in mind power consumption on the systems
https://tpucdn.com/reviews/MSI/GTX_970_Gaming/images/power_peak.gif
970= 181 watts
290X= 294
You would want 200 more watts for the 290x crossfire setup. A 750 watt unit would do the 970 sli setup fine but the 290X would need 950 watt unit. Factor that into it. Would you need to buy a new power supply just do even do it? Myself i would never do either and would rather game on 1440P like most here to i think less of a hassle. HECK take a look at this guys
http://store.steampowered.com/hwsurvey

A lot are on 720P still(laptops?)

if 1080p is enough then hd7950 + i5-2300 is enough and its tdp is stated 200w but will consume <150w at max (thus saves initial money and electricity bill too)
or just buy ps4
something to think upon, if you are using any AA then you are not satisfied with your current resolution/pixel density and you want more resolution/density or vsr/dsr

if we consider 4 hours a day gaming then we will be using 292 units extra in 365 day (290x crossfire vs 980 sli max power) means Rs1460 @Rs5/kwh (INR, Rs1460 equals ~24 US $ or around 50$ if i consider 0.17 cents per unit ) per year extra means we will need to game for several years to justify the electricity costs over card's cost

btw point was not if nvidia is more efficient or not, the point was that our top single gpu cards have maxed out 1080p so either we should stop producing powerful card and stick to 1080p or catch up and get ahead of tv and mobile display and get 4k-8k* (as we have most powerful hardware)
point was also to see what juan will say about resolution increase and how his apu will be able to handle this

and every new thing takes time for adaptation, and currently 4k monitor are very expensive so it will take time
even window 8 took time to surpass ages old xp and 8 doesn't costs as much as 4k monitor and we are not loosing any support for our 720,1080p displays due to which we will see quick adaptation rate

LCDs are holding back graphic cards at this point. I was gaming on a 1600x1200 (1.9MP) CRT in the early 00s and most people, over 10 years later, are only at 1920x1080 (2MP). The total number of pixels is not that far off.

The push for VR is in part to the fact that it'll finally get graphic cards pushing something beyond a single 1080p screen. From what AMD showed at GDC, you're basically looking at two 1080p screens for a VR headset. It's still 1080p, but it's twice as much as a single 1080p screen. Considering we've made pretty much zero improvement in pixel density in the last 15 years. I'd even go beyond that, Carmack developed Quake on a 1080p CRT in 1995. 20 years and no major improvements outside of 4k being out of reach for most people.

That's the real problem with graphic cards at the moment, that and we've reached the point where we see massive diminishing returns with the amount of time and money invested into games to increase graphic quality.

http://www.maximumpc.com/amd_announces_liquidvr_sdk_gdc_2015
AMD has some exciting announcements, one of which is massively improving multi-GPU latency for VR (and I'm assuming it'll work just fine for standard Crossfire as well). They're also mentioning using multiple dGPUs to do compute related things, like image warping, like I mentioned they would be doing.

 

[-Fran-]

In the meantime:

http://www.techspot.com/review/970-homeworld-remastered-benchmarks/page4.html

i3 4130 > FX-9590
Pentium G3320 > FX-8350

Game scales to two cores (as expected of an AAA game), and loves per-core performance (as expected in non-GPU bottlenecked situations).

Though i7-5960x @ 3GHz > i7 4770k @ 3.5GHz, seems kinda odd does it not?

FX4100 > Phenom II 980 BE xD!

Yeah, something is really wrong with that game.

Cheers!

 

[logainofhades]

Something really wrong with game, or the person doing the benchmarking.

 

[logainofhades]

In the meantime:

http://www.techspot.com/review/970-homeworld-remastered-benchmarks/page4.html

i3 4130 > FX-9590
Pentium G3320 > FX-8350

Game scales to two cores (as expected of an AAA game), and loves per-core performance (as expected in non-GPU bottlenecked situations).

Though i7-5960x @ 3GHz > i7 4770k @ 3.5GHz, seems kinda odd does it not?

Only thing I can think of, is the game really loves cache. That or it is a very memory sensitive title, that takes advantage of quad channel DDR4 2400.

 

[gamerk316]

In the meantime:

http://www.techspot.com/review/970-homeworld-remastered-benchmarks/page4.html

i3 4130 > FX-9590
Pentium G3320 > FX-8350

Game scales to two cores (as expected of an AAA game), and loves per-core performance (as expected in non-GPU bottlenecked situations).

Though i7-5960x @ 3GHz > i7 4770k @ 3.5GHz, seems kinda odd does it not?

Cache maybe? Would actually help explain the results if the game is very sensitive to CPU cache.

 

[juanrga]

On the Mantle/Vulkan thing, let me be clear: Vulkan is going to be slower then Mantle for no other reason then because it's vendor neutral. It's going to be one step higher.

DX12 is vendor neutral, but faster than Mantle even on GCN hardware!

OS specific, so you can extract a lot of efficiencies OS side.

Fair remark, but I guess that Valve will optimize SteamOS for Vulkan.

And more Vulkan details:

http://www.pcper.com/reviews/General-Tech/GDC-15-What-Vulkan-glNext-SPIR-V-and-OpenCL-21

It sounds more efficient than Mantle by the use of pre-compiled SPIR-V.

 

[juanrga]

"Mantle will be open to everyone" -> They're releasing the full API later on this month: http://www.phoronix.com/scan.php?page=news_item&px=AMD-...

"AMD no longer has plans to release a public Mantle SDK"

http://www.legitreviews.com/amd-mantle-dead-known-vulcan-api-uses-mantle-technology-opengl_159339

And last, but not least:
"Mantle will be ported to linux and apple" -> Apple said no, you want AMD to shove it down Apple's throat or something? And once Khronos wanted to take a look at MANTLE and continue with Vulkan, why even bother on making a Linux port when Khronos will do it for them?

Kronos has not made a linux port of Mantle. Vulkan is not Mantle. Moreover, the arguments given by AMD to justify Mantle on Windows apply to linux as well.

AMD original plan was DX12/Mantle on windows, Vulkan/Mantle on linux, and Metal/Mantle on Mac. The plan has changed completely.

Juan, the functions in Vulkan are all the same functions in Mantle with different extensions down to even having the same names. Check the link above. They literally just modded a few things and boom...Mantle is Vulkan.

Some code looks like Mantle other code doesn't. Vulkan is a collective development from different members of the Kronos Group. AMD contributed Mantle, other members contributed different stuff. Kronos group has confirmed that Vulkan is not mantle, explained in what parts Vulkan differs from Mantle and even why some part of Mantle were rejected for Vulkan.

AMD has contributed Mantle towards the development of Vulkan, and though we need to be clear that Vulkan is not Mantle

Vulkan is definitely a working group design now

 

[Cazalan]

Considering we've made pretty much zero improvement in pixel density in the last 15 years. I'd even go beyond that, Carmack developed Quake on a 1080p CRT in 1995. 20 years and no major improvements outside of 4k being out of reach for most people.

I remember spending ~$800 for a quality 21" 1280x1024 CRT in the 90s.

Today you can get a UltraHD 24" 3840 x 2160 IPS for about the same. Even better if you adjust for inflation. That's quite a bit more dense. Effectively 6 monitors in 1.

 

[-Fran-]

"AMD no longer has plans to release a public Mantle SDK"

http://www.legitreviews.com/amd-mantle-dead-known-vulcan-api-uses-mantle-technology-opengl_159339

In that same link you can also read this:

WHAT THIS MEANS
OpenGL has long and deservedly commanded respect for being a fast, versatile and wide open API that works on all graphics vendors across multiple operating systems.

Meanwhile, Mantle has seen acclaim for many improvements in gaming and game development: higher framerates, reduced rendering latency, reduced GPU power consumption, better use of multi-core CPUs, and re-pioneering new features like split-frame rendering.

Vulkan combines and extensively iterates on these characteristics as one new and uniquely powerful graphics API. And as the product of an incredible collaboration between many industry hardware and software vendors, Vulkan paves the way for a renaissance in cross-platform and cross-vendor PC games with exceptional performance, image quality and features.

And that's taken from the AMD blog site according to your link. In any case, LIKE I SAID, AMD sees no need to rush MANTLE out in the open since Vulkan and DX12 will take most of it and do something big enough to satisfy most things MANTLE promised. Now, for the "openness" itself, they said they'll announce something at GDC. Wait until then to keep on bashing AMD on this one.

In any case, AMD has to think what to do with MANTLE. I'm thinking they might use it to showcase GCN strengths and force some standardization. That's a double edged sword, but the first try went fine, I'd say. I wonder what's in for AMD on this; I still can't come up with something that explains all of this "trouble" AMD has gone through with MANTLE.

Kronos has not made a linux port of Mantle. Vulkan is not Mantle. Moreover, the arguments given by AMD to justify Mantle on Windows apply to linux as well.

AMD original plan was DX12/Mantle on windows, Vulkan/Mantle on linux, and Metal/Mantle on Mac. The plan has changed completely.

For short, and how most people understand it: Vulkan > MANTLE. Clear enough on what I mean when I say "Khronos will do the job for AMD"? Independent on what the original plan was, they haven't done a 180° on that. Things just didn't work with Apple and on Linux Khronos gave them an offer they could not resist; and it didn't involve dead horses, haha.

Some code looks like Mantle other code doesn't. Vulkan is a collective development from different members of the Kronos Group. AMD contributed Mantle, other members contributed different stuff. Kronos group has confirmed that Vulkan is not mantle, explained in what parts Vulkan differs from Mantle and even why some part of Mantle were rejected for Vulkan.

AMD has contributed Mantle towards the development of Vulkan, and though we need to be clear that Vulkan is not Mantle

Vulkan is definitely a working group design now

"Vulkan is MANTLE just like OpenGL is Glide". Let that sink for a minute, will you?

Cheers!

 

[juanrga]

Without Mantle it's likely that Vulkan would be a year or two behind DX12.

From my former post on Vulkan:

http://www.tomshardware.co.uk/forum/352312-28-steamroller-speculation-expert-conjecture/page-370#15418239

Mantle was used to bootstrap the process and speed its development

 

[cdrkf]

Oh for **** sake... will you lot *stop* bashing on Mantle and think about all this for a seccond?

Frankly it doesn't matter 1 jot if Vulkan is or isn't based on Mantle (though I think you'd have to be pretty biased to argue it had no influence over it's design).

It also doesn't matter if Mantle was or wasn't an integral part / reason for DX12.

1 thing is clear, graphics api's are finally developing after a very very long hiatus. I personally think Mantle kicked the various parties into action (undoubtedly there would have been some ground work in place before Mantle, however I also think that without it the new api's could have been quite a way off).

As for 'is this a win or loss for AMD'? From a purley financial view point, it doesn't make much difference. However, it will allow more modern games to make better use of multi threaded architectures which is definitely a benefit for AMD's cpu side of things. On the graphics side, as much as some like to claim how 'AMD is doomed', frankly I see no evidence of that. They've been keeping pace with nVidia for the last for generations (when starting from behind when they took over ati). nVidia currently have a lead with Maxwell, but lets see what amd's latest looks like before declaring them dead in the water, as it's always like this when one vendor is a generation ahead (look at reviews of HD 7970 vs GTS 580 ).

 

[de5_Roy]

AMD: We will actively promote ‘Zen’ processors for server market
http://www.kitguru.net/components/cpu/anton-shilov/amd-we-will-actively-promote-zen-processors-for-server-market/

Nvidia's PhysX joins the open-source party
http://techreport.com/news/27910/nvidia-physx-joins-the-open-source-party
"cpu-based"

 

[juanrga]

AMD: We will actively promote ‘Zen’ processors for server market
http://www.kitguru.net/components/cpu/anton-shilov/amd-we-will-actively-promote-zen-processors-for-server-market/

The impact of Zen on servers will be small. I guess AMD will sell Zen to former x86 Opteron customers (2--3% of servers). AMD only hope to gain significant market share is K12:

It's a longer-term play. We think it gets there and maybe if you ask me right now in the 2019 timeframe, I say 15% of the overall server space is in the ARM server business. When you get transitional space, AMD's market share today is like 2% or 3%, nothing to really talk about.

http://www.fudzilla.com/news/processors/37184-amd-hopes-15-of-servers-will-be-arm-based

 

[logainofhades]

2015 is going to be a rough year for AMD, CPU wise.