Intels pixel sync was used in couple of games.
Not sure what nvidia and amd have done.
AMD Polaris architecture to succeed Graphics Core Next
- Thread starter wachie
- Start date
Not sure what nvidia and amd have done.
Maybe many Gameworks effects are enabled using CUDA or Nvapi, I don't know what AMD have on their side.
Ah yes I remember GRID 2, Total War Rome, GRID autosport and some others using Intel's hardware extension for some neat OIT effects.
I also remember Johan Anderson from DICE wishing for Intel's Pixelsync to be exposed in future APIs.
I guess that has been done.
After how horrible the last cards ended up being (power draw/garbage driver support) can you blame him? I've been rooting for ATI/AMD since their original cards but what difference does that make when they constantly screw their customers?
Locuza
Member
In this regard welcome to Beyond3D and in particular Andrew Lauritzen from Intel and Sebbbi aka Sebastian Aaltonen from Redlynx and one of the creators from the presentation I linked earlier:
https://forum.beyond3d.com/threads/ps4-longevity.56816/page-7#post-1843234
https://forum.beyond3d.com/threads/direct3d-feature-levels-discussion.56575/page-12#post-1848919
https://forum.beyond3d.com/threads/directx-12-api-preview.55653/page-12#post-1855321
They both often times shed some light, about questions "we" have.
To quote simply some jam:
1. Part:
It's a bit problematic, since not every chip supports all the same operations, but we could support some operations, which every IHV nowadays do.
2. Part:
Officially no, but for DX11 they were.
From AMD for example:
http://developer.amd.com/tools-and-sdks/graphics-development/amd-gpu-services-ags-library/
Intel did some to support PixelSync for DX11 or Rasterizer Ordered Views in DX12 speech. (Grid 2/smoke stuff)
I didn't heard a lot about them, I would guess no.
No worries, glad we could talk things out instead of getting into warrior status.
Agreed, it's always fun (for me at least) seeing the PC/PS4/XB1 comparison threads. I like to see how each team handles, prioritizes, and deals with the hardware and limitations/strengths of each platform.
Thank you very much for the posts and info. Very useful.
If Just Cause 3 or F1 2015 are updated (support for DX12, not simply the engines) it could indeed put further pressure on AMD because it would go public, but as far as I understand it neither will receive a DX12 patch. It's just that their respective engines already have DX12_1 support.
F1 2016 might support DX12 along with Avalanche's next project.
It is no surprise that many engines available today have DX12 support, I suppose countless devs have been experimenting with it for quite some time but unless we begin to see a massive wave of games supporting DX12_1 AMD can rest easy.
I bet they will not, and half the 400 series GPUs will be rebranded models.
Locuza
Member
I hope I'm not crazy, when I think it would be crazy when JC3 and F1 2015 wouldn't receive DX12 Patches?
Since you do not announce your session with this headline:
If you would simply talk about engine integration, you would say, look at the new DX12 Features in the Avalanche Engine.
It's also both times presented by Intel, so I'm pretty sure it's the same deal it was before with Grid 2 or Total War.
Intel Marketing, Intel Collaboration.
Since you do not announce your session with this headline:
http://schedule.gdconf.com/session/using-new-directx-features-in-practice-just-cause-3-case-study-presented-by-intel
If you would simply talk about engine integration, you would say, look at the new DX12 Features in the Avalanche Engine.
It's also both times presented by Intel, so I'm pretty sure it's the same deal it was before with Grid 2 or Total War.
Intel Marketing, Intel Collaboration.
Maybe I'm a bit thick but I don't see how that says JC3 and F1 2015 will receive official and public DX12 support. To me it sounds like they are just disclosing what they have been working on, and how their games could benefit from it.
Of course I would absolutely love Just Cause 3 to support DX12, and not only behind the scenes. F1 being an annual IP I don't think they will bother, but JC4 is not planned for 2016 is it ?
Good point, it's also smart to try to recoup your investment on a longer period than the traditional 3 months for typical AAA releases.
Publishers need to think long term.
Maybe you misunderstood. He is stating that the GPU after Polaris this year, say Polaris 2 in 2017, they set a target of 20% improvement. Now it really depends on where Polaris lands this year but if it beats Maxwell by 20% then you sure as hell know that their next year's offerings are also crap against a Pascal that is 40% better than Maxwell.
Locuza
Member
I understood it the exact same way like Thraktor did.
And he made his points very clear.
This is the quote:
I'm sure the head of a division wouldn't casually mention what their performance plans for the next year are.
And he made his points very clear.
This is the quote:
Raja roughly said how it was in the past and what they did now.Raja Koduri (AMD) said:
I'm sure the head of a division wouldn't casually mention what their performance plans for the next year are.
General Lee
Member
One thing that hasn't been mentioned here about the Venturebeat article is that Koduri believes they are ahead of Nvidia with the transition, especially with mobile and mainstream market.
Given we've heard that AMD is doing their lower end chips at Globalfoundries 14 nm, and specifically we know of only two chips coming out, Koduri likely knows that the 14 nm production is ready much sooner than the TSMC 16 nm production. If we assume that they're preparing their enthusiast class GPU at TSMC, yet they only talk about these two mobile and mainstream chips, it would indicate that the chips done at TSMC are coming out considerably later. Of course Nvidia could be faster at TSMC if they do smaller chips first.
Given we've heard that AMD is doing their lower end chips at Globalfoundries 14 nm, and specifically we know of only two chips coming out, Koduri likely knows that the 14 nm production is ready much sooner than the TSMC 16 nm production. If we assume that they're preparing their enthusiast class GPU at TSMC, yet they only talk about these two mobile and mainstream chips, it would indicate that the chips done at TSMC are coming out considerably later. Of course Nvidia could be faster at TSMC if they do smaller chips first.
icecold1983
Member
on the nvidia side i think usage is mostly limited to nvidias gameworks suite. i dont think many devs bother using advanced nvidia only features on their own accord.
Hellgardia
Member
http://wccftech.com/amd-future-gpus-named-stars-galaxies-company-bids-farewell-island-code-names/
Really waiting to see what both AMD and nVidia will come up with.
Really waiting to see what both AMD and nVidia will come up with.
dgrdsv
Member
I don't think that there are any "visual features" that can work only on consoles. The biggest advantage of PS4 over a PC is its fast GDDR5 UMA architecture, the rest is smaller stuff like cross lane operations which are there technically on PC but aren't exposed in modern APIs.
HRAA as I understand it was dropped from PC because NV discontinued CSAA with Maxwell and HRAA is based on coverage reconstruction. It's rather bad in IQ anyway so not a big loss.
He's just promoting their future products. And I would honestly rather hear him flat out saying that Polaris will support FL12_1 instead of going into completely pointless console territory.
SmoothBrain
Member
Samsung Begins Mass Producing Worlds Fastest DRAM Based on Newest High Bandwidth Memory (HBM) Interface
From the article:
A single 8GB package. Crazy.
From the article:
A single 8GB package. Crazy.
It also uses analytical AA (GBAA) which is quite slow without having access to interpolators in hardware and API. (Have to use geometry shaders.)
qa_engineer
Member
Compact, cool, efficient and powerful. Cannot wait to see what Nvidia and AMD put out in late 2016.
Saucycarpdog
Member
AMD is definitely not ahead of Nvidia in mobile GPUs. One look at the gaming laptop market shows that AMD is nowhere to be seen, Nvidia dominates that landscape. Even if we were talking about mobile SOCs Nvidia still is ahead of them in that considering Nvidia has the X1 series that is actually in a tablet while AMDs mullen series is not in one product so far.
Vanillalite
Ask me about the GAF Notebook
Samsung rolling shit out in terms of fab. The upped the amount of ram per chip for their LPDDR4. They announce last week their 14nm finfet 2 process was ready for mass market. Now they announce HBM2.
Couple that with the idea that they are probably gonna be the iPhone OLED supplier, and Samsung is doing well even if their own branded electronics aren't selling at the high clip they use too.
Hellgardia
Member
One of the things i want to see from nVidia are the Tegra Pascal boards. I wonder if they'll pair it with a newer version of their Denver cpu.
If AMD plays their cards well they could very well grab some of the laptop market as well.
If AMD plays their cards well they could very well grab some of the laptop market as well.
dgrdsv
Member
Hellgardia
Member
Yeah but the original roadmap paired the Denver (2?) CPU cores with Maxwell GPU. I guess now it will be Denver + Pascal
dgrdsv
Member
Hm, I don't think that they've specified what CPU cores they'll pair with Maxwell explicitly. Everyone just assumed that since they've used Denver in one configuration of K1 they'll use it in Erista and Parker as well. And they've decided to not use it in Erista/X1 and will use them again in Parker instead.
General Lee
Member
Of course they're not, and that's not what was said. They'll be first to 14 nm products, and they'll have them for back-to-school period which is super important. Nvidia will apparently only do a refresh with 970MX and 980MX since Pascal won't be ready for mobile in time. This is very good chance for AMD to take back market share in mobile where they've been nearly non-existant.
https://translate.google.com/transl...oard.php?bo_table=news&wr_id=15839&edit-text=
Slightly OT, but I figured we were talking about AMD already so why not.
According to the AMD investors' report
I wonder why there's a decline in the royalties for console APUs? Are the royalty structures front-loaded, or possibly related to # of consoles sold (since there were more sold this FY than last, right?)?
This also means that Sony/MS are making more per console sold now, which increases profit and also gives them another chance for a price drop while still being profitable.
According to the AMD investors' report
http://www.wsj.com/articles/amd-fourth-quarter-revenue-tumbles-23-but-loss-narrows-as-costs-fall-1453242928
I wonder why there's a decline in the royalties for console APUs? Are the royalty structures front-loaded, or possibly related to # of consoles sold (since there were more sold this FY than last, right?)?
This also means that Sony/MS are making more per console sold now, which increases profit and also gives them another chance for a price drop while still being profitable.
Haha, thank you. I just realized I've been reading that wrong this whole time. For some reason I was thinking in my head that it was talking about the royalty fee per APU, not the overall royalties.
Silly me.
Thanks.
jeff_rigby
Banned
Both of you are correct. Royalties do drop with time as the price drops and sales are seasonal. At least that's my understanding from AMD transcripts (Stock disclosures).
We should get a bump in sales for both the XB1 and PS4 when they start being used as DVRs, Cable TV STBs and Antenna TV ATSC 2 STBs supporting XTV as well as Vidipath servers and clients. Then UHD IPTV and Cord cutter IPTV (Playstation Vue) using HEVC and likely as 4K blu-ray players with digital bridge.
PS5 and XB2 will likely drop after 2018 so both generations will be shipping making money for AMD.
Polaris spotted on Zauba for import price around $600 (rumor)
In the same article, another product was spotted around $450, presumably the pared down part.
In the same article, another product was spotted around $450, presumably the pared down part.
Not related to Polaris, but Falcon Northwest showed off a new desktop with a Radeon R9 Fury X2 dual-GPU card running a HTC Vive demo at the Virtual Reality Los Angeles 2016 Winter Expo.
https://twitter.com/FalconNW?ref_src=twsrc^google|twcamp^serp|twgr^author
http://wccftech.com/amds-dual-fiji-based-radeon-r9-fury-x2-graphics-card-spotted-vrla/
http://www.fudzilla.com/news/graphics/39761-amd-shows-off-radeon-r9-fury-x2-dual-gpu-at-vrla-2016
https://twitter.com/FalconNW?ref_src=twsrc^google|twcamp^serp|twgr^author
http://wccftech.com/amds-dual-fiji-based-radeon-r9-fury-x2-graphics-card-spotted-vrla/
http://www.fudzilla.com/news/graphics/39761-amd-shows-off-radeon-r9-fury-x2-dual-gpu-at-vrla-2016
jeff_rigby
Banned
Zen + Polaris + HBM (likely some other Star name GPU) as a APU. The XB1 can emulate a Xbox 360 and likely the next generation APU with Zen can emulate the PS3 and any older generation chipset designed for consumers.
The Stock market is speculating on how long AMD can last. AMD was seriously impacted by the end of Moore's law (leakage increases with smaller nodes) and the costs in implementing HBM. CPUs require a larger cache as they get faster which Intel designed in and that raised the cost of Intel CPUs. GPUs as they get larger are memory starved; the answer for AMD was HBM and then HSA with a new memory interconnect fabric. To efficiently use HBM the cache and registers in CPU and GPUs need a redesign, this is also necessary for new power efficiency schemes. If they can do this and meet the price point the consumer is willing to pay they will win back market share. I think that is their plan and by 2018 their stock price should reflect how well they succeded.
jeff_rigby
Banned
Which is why the XB1 and PS4 are using AMD not Intel Nvidia?
The current iMac is using a small Intel CPU/APU and AMD dGPU. Zen is competitive with Intel CPUs and AMD GPUs are better than Intel. For the first time a APU could replace the combo APU + dGPU the iMac is using and it would be faster or at least cheaper than Intel and adequate at the worst. Then their is AMD using open standards, many of which still haven't been implemented because too few platforms support them.
michaelius
Banned
Right OEM contracts are definitely all about performance and not about maximizing margins at the cost of supplier
unapersson
Member
Which is why we're all on Itanium on our high end desktops now rather than X86-64.
jeff_rigby
Banned
I take it that was a sarcastic comment but I don't see how it applies: "this deal would allow Apple to secure a high performance x86 SOC design at a significantly lower cost than competing Intel solutions." A next generation APU/SoC should be cheaper than a CPU/APU + dGPU combination. AMD has recommended small APU + dGPU until next generation which has to incorporate multiple power saving schemes and HBM so that the memory can support all of them. The XB1 and PS4 as game consoles get a break in power efficiency requirements that still applies to computers.
Currently the US government can not purchase a Computer unless it complies with Energy Star/EU power requirements.
Yes CPU/APU + dGPU at the present time favors Intel as it has the more efficient more powerful CPU/APU. The same is now seen in the iMac. When we move to larger APUs with the newer open source standards this should change for PCs and iMac. If most PCs come with a APU that is equal to a XB1 or PS4 and you can add a dGPU to this I expect most PCs or iMacs will ship with this chipset. The new memory fabric confirms this: 4 TF APU! I suspect in the short term the mainstream/cheapest APU will be PS4 level performance.unapersson said:
dgrdsv
Member
A CPU+dGPU combo will always be faster than just the CPU with iGPU, nothing will change with the new APUs from AMD. Intel's Skylake iGPU is more advanced than GCN and Iris graphics performance is nothing to be ashamed of - its usage of EDRAM is more or less the same thing which AMD plans to achieve with HBM and it's been out for a couple of years already. There is nothing new in what AMD plans to offer, the only reason Apple may use AMD's chips is because they'll be cheap while providing an acceptable level of performance.
jeff_rigby
Banned
EDRAM is expensive and can not be provided in the amounts that GPUs need. In the short term EDRAM can improve the efficiency of iGPUs that have to rely on DDR3 or DDR4 (XB1 and Intel). HBM changes everything which is why Intel has plans to use HBM just as AMD does. Will AMD have a cheaper larger APU than Intel? The rumor thinks so and I think AMD has to to survive.
A CPU + dGPU because of TDP limitation can be larger and faster. Not saying this is not true, what I am saying is that next generation APUs becuse of HBM will be larger and the basic building block for a PC or iMac not a small APU + dGPU. So it changes to something close to a PS4 level APU (2 TF or smaller with larger faster CPU) as the minimum in a PC with dGPU optional. Sockets still supported with DDR4 on the motherboard so upper end PCs can have a 4TF APU and larger dGPU.
"AMD Has Three Semi-Custom Designs In The Pipeline In Addition To The Two Current Console Designs" Of the three new semi-custom designs one is gaming, one is beyond gaming and one is ARM with the beyond gaming either X-86 or ARM. AMD just announced a ARM server design so a guess is Nintendo NX, Apple iMac and the ARM server.
Edit: In the short term we could see a XB1 or smaller APU with EDRAM as the PC or iMac APU but for a larger APU it needs HBM and register - memory fabric changes.
Thanks!
Hands of Stone
Member
Can we expect these to be on par, if not outright better than Pascal?
Aztechnology
Member
There's really no way to know outside of projected bandwidths etc. The high ends seem to be similar but we have no idea if the low, medium or high end will be released in what order for each respective company and how they will compare.
dgrdsv
Member
Did you see CPU with Iris graphics? The EDRAM chip on it is implemented in the exact same fashion as HBM stacks are implemented on Fiji. The difference is in UMA/NUMA obviously but for a typical PC usage scenario this difference is negligible as you will never be able to omit NUMA support on PC platform.
Last time I checked Intel did not plan to use HBM as they opt for HMC instead. In any case the memory type is not important, what you get with HBM is more bandwidth and this is just extensive evolution which does not change anything.
Nothing changes. dGPUs will progress at the exact same rate as iGPUs. It doesn't matter how PC APUs compare to console parts, reaching PS4 GPU performance in a PC APU won't change anything on the PC market as it's already way beyond of what is possible on current gen consoles.
Back to school is what I've read last time.
Similar threads
- Poll