50W Polaris 11 gpu card?

[juno]

At this point nobody who is not an insider can tell you and insiders definately won't. And you are in the wrong subforum, too.

The following is just based on rational assumptions. I would expect cards without an extra PCIe power plug but therefore up to 75 W, not only 50. Raw performance should outmatch Pitcairn (7800/270/370 series) may even come close to Tonga PRO (R9 285/380) or GM206 (GTX 960). When it comes to Linux gaming performance, we'll have to wait and see.

btw: IMHO, a R9 470X below 50 W doesn't make any sense. You shouldn't read those "news" from wccftech. They only repeat rumors over and over and often write "articles" with ridiculous assumptions (just look up past ones and compare to today's products). Go for the original sources instead or choose news sites that don't re-publish every rumor from the internet.

[bridgman]

Yeah, the rumor mill is scary some times, sure fire recipe for disappointment. Everyone seems to be expecting something that runs rings around Fury and sells for $200. I don't think either AMD or NVidia are planning to do that.

[starshipeleven]

(half-decent) GPUs with 50W of TDP?

LOLWHUT?

Not this die shrink, maybe the next die shrink. MAYBE.

assuming it isn't blatant lies, stuff with that TDP is either a GPU for APUs or embedded, not something you can game on. For common meanings of "gaming", anyway.

[juno]

Depends on your expectations. For debianxfce, who is all satisfied with a 750, it's more than enough.
Sure, you won't be using a 144 Hz WQHD panel and highest settings but you can still play.

There are rumors about people that used to pay a few hundred bucks on something like a PS3 and still enjoyed games for years.
playing != "playing"

[starshipeleven]

Depends on your expectations. For debianxfce, who is all satisfied with a 750, it's more than enough.
Sure, you won't be using a 144 Hz WQHD panel and highest settings but you can still play.

There are rumors about people that used to pay a few hundred bucks on something like a PS3 and still enjoyed games for years.
playing != "playing"

Not understood my point.

Dedicated cards are basically getting more and more into the gamer-only-thing realm, because APUs and even Intel integrated stuff basically nuked the whole low-end market segment and sometimes even mid-end cards are not considered in favor of getting a good APU.

I'm """"gaming"""" on a HD4000 (a Ivy Bridge Xeon iGPU actually, but that's the well-known version it is a respin of), and decent APUs can run pretty much anything if you don't need all settings to max (slight exaggeration, but you get the point).

So no, making a new GPU that is total crap for gamers while APUs of the same year or the next one have more or less identical hardware (the same for Intel camp) is total nonsense.

If you want to play without expectations you look for a good APU, or Intel equivalent, not for a total crap pcie card. This is the direction we are going iGPU getting better and better, dedicated cards more and more restricted to niches like gaming and workstations.

And this is why I think that this info is bogus or about a part meant to go in embedded devices, or that it will be part of an APU and not a separate piece of silicon.

[Kano]

Midrage cards usually can run most games fast enough @1080p - with reduced settings. For Intel onboard even 720p can be too much. It often runs but with 2x fps, Ivy Bridge certainly slower than more current chips. It always depends which games you want to play. Some eON ports need OpenGL & GLSL override to play with Intel but it works: Bioshock Infinite and DiRT Showdown. Onboard solutions are usually nothing for hardcore gamers but very common. Windows drivers are often faster with the same games - most are optimized for DX - hard to compete with relatively unoptimized OpenGL especially at the lower end.

[starshipeleven]

http://www.tomshardware.com/reviews/...rchy,4388.html
HD4000 is quite slow gpu.. Prince of Percia 2008 and Tomb Raider 2013 are too heavy games for hd400 at 720p, with wine-staging 1.9.5 csmt enabled. My A8-7600 has r7 250 ddr3 gpu and fps is quite low when you have max action in the games.

1. I and most people I know using low-end GPUs play on windows for a long list of very good reasons. Hell, not even Team Fortress 2 (native linux port from steam) works in linux on the HD4000 (in my Debian Jessie install) while on windows it works fine on the same hardware.
Feel free to buy entry-level gaming GPUs to play like a iGPU on windows due to multiple layers of emulation, command redirection and unoptimized APIs.

2. I am saying that this fictional 50w part is likely to be either total bs (most likely) or with raw gpu power comparable to an iGPU for an APU of the same year. Die shrinking does not work like that on power consumption.

[juno]

Wow, are we now in the official phoronix rumor mill?

Dedicated cards are basically getting more and more into the gamer-only-thing realm, because APUs and even Intel integrated stuff basically nuked the whole low-end market segment and sometimes even mid-end cards are not considered in favor of getting a good APU.

Now I get your point and I agree up to a certain degree.
It is true for the low-end budget cards with DDR3 and the most crippled chips. It is not true for the "normal" low-end like the GTX 750 Ti or the R7 360.
And obviously, these products are not gamers-only but also data centers, HPC supercomputers, professional workstations, etc. etc.

So no, making a new GPU that is total crap for gamers while APUs of the same year or the next one have more or less identical hardware (the same for Intel camp) is total nonsense.

How dare you call it "total crap for gamers" when it is not released or not even official information available?
Today an Intel iGPU, even in the highest-end configuration consuming up to 65 Watts on the desktop with eDRAM is nowhere near those first not-crappy dedicated graphics cards: http://www.tomshardware.com/reviews/...ll,4169-6.html
And that is for 720p, the min FPS already dropping down to 23. For higher resolutions, the memory bandwidth limitations hit critical.
AFAIR, Intel is doubling their GPU part on Skylake and even put the next-gen graphics part into Kaby Lake. It will be interesting how Kaby Lake catches up.
On the AMD side, Intel's fastest solutions are already faster, due to eDRAM. iGPUs are highly bandwidth-limited (and also power-limited on the AMD side, as they have their current generation only in Notebooks, not on AM4 yet). As long as AMD doesn't put a decent HBM stack next to their APU or some GDDR5 or some eDRAM (which I don't expect, though), too, there is no way they are catching up on dGPUs.

If you want to play without expectations you look for a good APU, or Intel equivalent, not for a total crap pcie card. This is the direction we are going iGPU getting better and better, dedicated cards more and more restricted to niches like gaming and workstations.

Again, we are not talking about "total crap". P11 was demoed on the RTG summit and outperformed the GTX 950 easily in terms of efficiency. It ran just around ~800 MHz (!) and did the same framerate the 950 did (v-synched, though, so capped to 60).
If there is nothing completely wrong with the architecture and the manufacturing process at GlobalFoundries, we are talking about the performance sector of Pitcairn-Tonga or GM206 (GTX 950/960) here. APUs are nowhere near that.

And this is why I think that this info is bogus or about a part meant to go in embedded devices, or that it will be part of an APU and not a separate piece of silicon.

No, It won't. The roadmaps are clear*. P11 was demoed to the press back in December, also spotted on the GDC. There are no new APUs coming anytime soon. Bristol Ridge, sure, which is Carrizo with improved power management. It's not new silicon.
However, BR will obviously gain some performance due to DDR4. It still has 8 CUs from VI generation, P11 should at least double that and run much higher clocks.

*
mid-year:
Polaris 10 & 11

end-year:
AM4/Summit Ridge (Zen CPU)
Bristol Ridge (Excavator APU)
Vega (Fiji-successor)

2017:
_Maybe_ the first Zen really High-End APUs. Nothing you would buy for your, compare this more to Knight's Landing or Tesla P100.
Zen-based APUs (Raven Ridge), hopefully paired with HBM *dreaming*
Even bigger Vega

Sure, you can speculate that BR is in 14 LPP too, utilising a new Polaris GPU, but I wouldn't expect that. Why should they do the effort and bring BR to 14 LPP when they could bring Zen-based APUs just a few months later?

Polaris 11 uses amdgpu driver, so this forum is just right place to discuss it.

So what? Are you expecting bridgman to jump in and say "yes, AMD is planning a card like this. It gives you 123% FPS improvment over the 750 and come for just 45 bucks"? That's obviously not gonna happen.

[juno]

750ti 2GB ram do cost 130usd, not 45 usd.

You DO realise, I was kidding, using made-up numbers (12345) right?

https://youtu.be/5g3eQejGJ_A?t=125

P11 > GM206 > GM107
However, that is DX11, of course.

Now make your own guess about performance (and price!)

[aht0]

750ti 2GB ram do cost 130usd, not 45 usd. And if the polaris 11 pci-e powered version is faster and do cost 130 usd, of course that is a good alternative. Good open source driver is also plus, you can fix problems by yourself.

Probably depends if you are buying one from retail store or buy it already used from previous owner.