HumanSmokeYes

Not sure how that tracks. Since GCN debuted the FP64 rate has been tied to the market the GPU was primarily aimed for. In 2012, Pitcairn and Cape Verde had 1/16 rate while the more compute minded Tahiti had 1/4. The second GCN iteration was similar - Oland and Bonaire both 1/16 with Hawaii 1/2 (still the GPU basis for AMD's compute cards). The third iteration made 1/16 standard across both Tonga and Fiji. It would be almost a certainty given AMD's "2.5X performance per watt" claim that Polaris will also feature a 1/16 FP64 rate.

Seems highly unlikely but I like your optimism.

I don't think your response relates to what I said. I was taking about the improvements in efficiency with the new Polaris chips The new chips will likely be more efficient in terms of utilizing the computing resources than the previous gen chips based around the same "architecture". I wasn't talking about ration of FP64 to FP32 units.

I don't think I'm exaggerating with my performance predictions. There is a reason why AMD is using a new memory compression technique to improve the effective bandwidth of the new Polaris chips. My guess is that the 480x Polaris card, which is rumored to have a 256-bit memory controller, is hitting the ~250GB bandwidth limit. If this is the case, then the 480X is likely faster the GTX980.

G33k2Fr34kI don't think your response relates to what I said. I was taking about the improvements in efficiency with the new Polaris chips The new chips will likely be more efficient in terms of utilizing the computing resources than the previous gen chips based around the same "architecture". I wasn't talking about ration of FP64 to FP32 units.

Well in that case I don't know what the hell you are talking about. The only other option I can see is basically SIMD power gating, but given the rumoured die size of Polaris, that doesn't seem really doable given the amount of extra gating logic that would need to added to make it effective (i.e ALUs more flexible in workload and allocation). I could see it on a larger die where (say Vega or Navi) where the cost of gating could be mitigated and a larger cache structure could alleviate stalls in workloads of differing sizes, but I fail to see how it translates to a mid size GPU. Care to enlighten me whether you are talking about power gating and variable SIMD unit size or have something else in mind? Your posting is so generalized it is difficult to ascertain what you mean by "improvements in efficiency"

G33k2Fr34kI predict that the 470 will replace the 390X and that the 480 will replace FuryX performance wise.

The <50W R7 470 will replace the 390X?!?

Well, speaking of Vulkan, why would anyone (besides "weareafraidMSStorewillthreatenourmonopoly Valve") embrace it at this point?



arstechnica.com/gaming/2016/02/vulkan-benchmarks-a-boost-for-amd-and-nvidia-but-theres-work-to-be-done/

It's somewhat faster than older OpenGL, but nowhere DX11 perf levels.

As of today, AMD has still not released Vulkan support for Linux. So much for their "big advantage".

BTW, optimized OpenGL will outperform Direct3D 11 (at least on Nvidia hardware), but when a game has three rendering pipelines it's clearly not optimized.

Haha, my Vapor R9 280x with 384 bit memory has close to 300Gbps so 480x with 256bit and 250Gbps will be a downgrade i guess. I can post pics for proof but im not home right now. Ill do that in the evening.

medi01Well, speaking of Vulkan, why would anyone (besides "weareafraidMSStorewillthreatenourmonopoly Valve") embrace it at this point?



arstechnica.com/gaming/2016/02/vulkan-benchmarks-a-boost-for-amd-and-nvidia-but-theres-work-to-be-done/

It's somewhat faster than older OpenGL, but nowhere DX11 perf levels.

Right now, Vulkan in Talos principle is just a wrapper around the OpenGL implementation. Devs themselves have said it will take more work for them to properly harness what Vulkan has to offer, but they have put this out to have a baseline (and check for correctness).

HumanSmokeWell in that case I don't know what the hell you are talking about. The only other option I can see is basically SIMD power gating, but given the rumoured die size of Polaris, that doesn't seem really doable given the amount of extra gating logic that would need to added to make it effective (i.e ALUs more flexible in workload and allocation). I could see it on a larger die where (say Vega or Navi) where the cost of gating could be mitigated and a larger cache structure could alleviate stalls in workloads of differing sizes, but I fail to see how it translates to a mid size GPU. Care to enlighten me whether you are talking about power gating and variable SIMD unit size or have something else in mind? Your posting is so generalized it is difficult to ascertain what you mean by "improvements in efficiency"

I don't know what the hell you're talking about. SIMD power gating improves power usage, I'm not sure how that relates to performance.

G33k2Fr34kI don't know what the hell you're talking about. SIMD power gating improves power usage, I'm not sure how that relates to performance.

You were talking about "improvements in efficiency" which I was addressing. You clearly stated it:

G33k2Fr34kI don't think your response relates to what I said. I was taking about the improvements in efficiency with the new Polaris chips The new chips will likely be more efficient in terms of utilizing the computing resources than the previous gen chips based around the same "architecture". I wasn't talking about ration of FP64 to FP32 units.

You don't think that power gating to improve performance/watt is related to efficiency? I believe you are just trolling at this point.
As for performance, you clearly did not read the patent link I provided? Obviously not. The technology also allows for SIMD lanes to addititive (increasing their width) on the fly to better optimize for performance depending on workgroups being undertaken. Section 0030 deals with efficiency gains through power gating, while section 0031 deals with workload efficiency by tailoring SIMD width



So. you still haven't answered my question on how AMD are going to achieve efficiency gains if you dismiss the most likely candidate for this to happen, then you turn your attention to performance while dismissing the most likely candidate for that to happen as well. Given your wildly optimistic prediction for performance leaps over the existing products where a new part is supposedly jumping two market segments (something that hasn't happened since the dawn of the unified shader architecture era) and your complete vagueness, I suspect I'm wasting my time.
Enjoy your dreams of a 470 replacing the 390X in the product stack.

RejZoRHD5850 delivered twice the performance of HD4870 iirc. It made one of the biggest performance leaps of all times

Just no.
www.techpowerup.com/reviews/HIS/HD_5850/30.html

I said iirc. It was HD4850. Give or take % so you won't nitpick at that as well.

RejZoRI said iirc. It was HD4850. Give or take % so you won't nitpick at that as well.

Double of a HD 4870 would be the HD 5870, maybe you were talking about that (800 vs 1600 shaders, HD 5850 has only 1440). And also quite true in the benchmarks.

efikkanAs of today, AMD has still not released Vulkan support for Linux. So much for their "big advantage".

BTW, optimized OpenGL will outperform Direct3D 11 (at least on Nvidia hardware), but when a game has three rendering pipelines it's clearly not optimized.

You should talk with Bridgeman at Phornix, AMD already has Vulkan driver for Linux.

efikkanAs of today, AMD has still not released Vulkan support for Linux. So much for their "big advantage".

BTW, optimized OpenGL will outperform Direct3D 11 (at least on Nvidia hardware), but when a game has three rendering pipelines it's clearly not optimized.

Not sure if you guys remember the 3dFX times, when games like Unreal, DeusEx, etc were running so much better on OpenGL or GLide, compared to D3D. It's all in the engine optimisation and implementation.

KananDouble of a HD 4870 would be the HD 5870, maybe you were talking about that (800 vs 1600 shaders, HD 5850 has only 1440). And also quite true in the benchmarks.

Yeah, now I'm slowly remembering. 4850 and 5850 and 4870 and 5870. And yeah, the performance was around 100% increased. Something we rarely ever see in the graphics card industry. It's usually 50% jump or even less between series.

HumanSmokeI suspect I'm wasting my time.

I suspect because "why are you talking about performance and power consumption, I'm talking about efficiency" bit ...

HumanSmokeYou could also make the case that developers are inherently lazy

I suppose when Epic adds Vulcan renderer in Unreal Engine, devs will start considering it.

BiggieShadyI suspect because "why are you talking about performance and power consumption, I'm talking about efficiency" bit ...

Since efficiency is generally expressed as performance-per-watt the connection must still be unfathomable for some people

BiggieShadyI suppose when Epic adds Vulcan renderer in Unreal Engine, devs will start considering it.

UE4 is basically a very popular off the shelf set of tools, so I'd agree that once Vulkan is integrated its visibility should take off in leaps and bounds.

Prima.VeraNot sure if you guys remember the 3dFX times, when games like Unreal, DeusEx, etc were running so much better on OpenGL or GLide, compared to D3D. It's all in the engine optimisation and implementation.

What did we have, back then, DX3-5? (So, end of 1996, v3 was simply terrible, v5 sucked too)

Microsoft kept investing heavily into it, nevertheless.

First "good" DX version was, probably 8. No major fruckups.

And in parallel, NV was doing typical NV things to OpenGL (which it, kinda, actually loved) - proprietary "let me mess up the competitor" extensions.
ATI discovers it needs to push yet another OpenGL extensions if they want to support shaders on it. (they kinda were in good relations with M$)
So, dear developers, wanna develop for OpenGL? And with shaders? Ah, do 2 incompatible implementations, pretty please!
Yay. The excitement.
So there goes OpenGL.

Now the question: WHO is going to invest a fortune into Vulkan? (to bring it on par to DX, forget about being better)
Lazy arse monopolist Valve? Well, maybe. Let's hope so. DX is the only thing that keeps me on Windoze.

@medi01 Dare I get off topic and comment? I will though, since you mentioned it twice.

Valve is not a monopoly. As is well-known on here, I am not a huge Steam-lover. I grudgingly use them only for games I cannot play through other means.

Valve is not a monopoly. Very influential, yes. Monopoly? No.

www.dictionary.com/browse/monopoly

monopoly
[muh-nop-uh-lee] /məˈnɒp ə li/

• Examples
• Word Origin

See more synonyms on Thesaurus.com
noun, plural monopolies.
1.
exclusive control of a commodity or service in a particular market, or a control that makes possible the manipulation of prices.
Compare duopoly, oligopoly.
2.
an exclusive privilege to carry on a business, traffic, or service, granted by a government.
3.
the exclusive possession or control of something.
4.
something that is the subject of such control, as a commodity or service.
5.
a company or group that has such control.
6.
the market condition that exists when there is only one seller.
7.
(initial capital letter) a board game in which a player attempts to gain a monopoly of real estate by advancing around the board and purchasing property, acquiring capital by collecting rent from other players whose pieces land on that property.

medi01What did we have, back then, DX3-5? (So, end of 1996, v3 was simply terrible, v5 sucked too)

Microsoft kept investing heavily into it, nevertheless.

First "good" DX version was, probably 8. No major fruckups.

They did it with DX, Office, Xbox and now they're trying to do the same with Universal Apps/MS Store and Windows Phone. The trouble is, it seems they're running out of cash to keep pushing these things.

rtwjunkieValve is not a monopoly.

With your definition, even google is not a monopoly, after all, we got bing, yahoo, eh?

Dictionary definition has nothing to do with it. It's about laws. In most countries it is defined as "having dominant position" in certain markets. It start with market shares as low as 50% and even 30% in some countries.

Valve, with whopping >70% share is a monopoly.

bugThe trouble is, it seems they're running out of cash to keep pushing these things.

Last quarter, 22 billion revenue, 6 billion income. Running out of what cash?
www.microsoft.com/en-us/Investor/earnings/FY-2015-Q4/press-release-webcast

Luckily for Valve, M$ decided to also leverage it for its mobile OS, which would badly impact its adoption.

Since Valve is such a big thing at the moment, it takes another big thing to shatter its monopoly.
There is nothing unique / challenging in what they are doing, it's just about established market share.
gog.com, sadly, has no chances.

medi01With your definition, even google is not a monopoly, after all, we got bing, yahoo, eh?

Dictionary definition has nothing to do with it. It's about laws. In most countries it is defined as "having dominant position" in certain markets. It start with market shares as low as 50% and even 30% in some countries.

Valve, with whopping >70% share is a monopoly.

Whatever, bro. You've obviously got some irrational feelings and hate you need to sort out. I'll leave you to it.

rtwjunkieirrational

Nothing in that post is irrational, bro. Maybe you have problems admitting you were wrong.

medi01Nothing in that post is irrational, bro. Maybe you have problems admitting you were wrong.

Valve is not a monopoly in such as defined. There are plenty of other online stores to source games from. Blizzard seems to be doing just fine without any agreements with Valve. Same with a majority of EA titles. Not to mention brick and mortar locations.

medi01Valve, with whopping >70% share is a monopoly.

I don't know anything about Valve, but monopoly is where the big money is. Happens all the time in software and the internet where it's easy to do if you are first to the game and make monopoly your goal. Even a shitty company can easily parlay an early dominance into eternal dominance and big $$$. Microsoft is a great example. Ebay. Facebook.

Remember when there were a bunch of different online auction sites and how cheap it was? Ebay was the worst one but they were first and always the biggest. Briefly there was a site where you could go to browse listings from all the auction sites at once. Great for consumers and competition. Ebay sicced their lawyers on it and got it shut down. Competition disappeared and fees went up by ~4x.

Anyway, this is way OT for this thread and has nothing to do with video cards....