Fiji and HBM.

[xfia]

http://www.forbes.com/sites/jasonev...ds-radeon-fury-video-cards-and-fiji-gpu-core/

 

[lilhasselhoffer]

I'm kinda getting tired of you evangelizing AMD. They're a company making GPUs and CPUs, not a religion.

I'm basing this on the fact that you started yet another article, finding a new source to say the same thing. At this point, you're either not reading through the plethora of existing articles here, or simply don't care because you're deeply entrenched in the fan boy mentality. Let's review:
AMD Fiji implementation detailed - news section - TPU
http://www.techpowerup.com/212740/amd-fiji-hbm-implementation-detailed.html
AMD Fiji block diagram, indicating cool and quiet running - news section - TPU
http://www.techpowerup.com/213695/amd-fiji-block-diagram-revealed-runs-cool-and-quiet.html


You could have reasonably included this in either of those threads, rather than making another. If you continue polluting this forum with copies of the same information I'll be asking the moderators to intervene in the future.



Edit:
Honestly, I'm tired of both sides of this insanity. Neither AMD nor Nvidia is doing anything ground breaking yet. They're stuck on the same node, and treading water. They've gotten communities like TPU into an uproar, with nothing to actually be excited about. Consider me too tired to continue giving a crap about any of this. My only pleasure left is taking fan boys with me on this spiral into bitterness.

 

[MrGenius]

Someone's sounding like a big shiny dickhead. I wonder who that could be?^^^

 

[Zakin]

More like he's just being honest. It's getting old. Just let the 25th come and let's see for ourselves. AMD's spouting right now is as heavy as the Bulldozer announcement..

 

[xfia]

Someone's sounding like a big shiny dickhead. I wonder who that could be?^^^

i would not have posted that if i seen the article.. the forbes article was just rich in content and could spark several conversations but it appears some people just need a hug or something..
good thing i got several threads to help fight off the troll haha

 

[manofthem]

Come on @xfia, you're not helping AMD or those of us that buy their products; you're just making us look bad.

Nobody needs to convince the Internet of how great AMD is, how Nvidia is the devil, or whatever it is you're trying to do.


Not even Morpheus tried this hard with Neo

 

[xfia]

how could it not help to educate.. what if a few people stop being cynical and start understanding trends in technology better.
i mean someone spitting off a bunch stuff that is almost blah blah blah to me trying to be smart finishes by saying api overhead wont really affect performance yet microsoft and amd and demostrate much higher performance with lower latency.
there is just much that is fact that goes largely ignored because of misleading research by reviewers and bad information on forums.
they want people to have better information and not just amd and microsoft are taking initiative but also the oem's with more informative marketing.
i can give the business practice stuff a rest because i think everyone knows how i feel about that haha

 

[Tetsudo77]

Please stop posting new threads every second.

 

[LightningJR]

Yeah man it's becoming spam, all your threads.. You need to stop. You're starting to seem like you've been paid off.

 

[xfia]

i guess i interrupted some riveting conversation happening.. tho it doesnt seem that way.
maybe it was on that so informative and tasteful article that amd doesnt trust its own cpu's?
anyone actually wanna talk about real things going on technologically or just argue about facts?
strange @erocker closed the one that further proves what i have been saying about the mods around here being cynical and not listening to what engineers have been saying for years and didnt close this one that was already a article i didnt see.

 

[Caring1]

I'm going to throw out a pedantic thought about the graphic shown in the intial post.
"Short" is not a unit of measurement, so how can something be two times shorter, surely it is .5 of the length, or half the length!

 

[xfia]

I'm going to throw out a pedantic thought about the graphic shown in the intial post.
"Short" is not a unit of measurement, so how can something be two times shorter, surely it is .5 of the length, or half the length!

thats a good point.. i think its really just for marketing like 2x just looks better to people than .5.
it certainly shows a level of optimization on 28nm thought impossible by some

 

[the54thvoid]

@xfia, despite the fervent tirades we see on forums, please understand that an awful lot of us do read multiple sources about technology.
It's actually not news that HBM is super efficient, its been well known for a very long time. Reposting from Forbes isn't helpful. It's a financial based news regurgitator that is slow with tech innovations.
If you do like science and tech, go to phys.org it'll blow your mind. Also makes the fluff we all argue about on TPU look utterly insignificant.
Real science is way better than AMD or Nvidia.

 

[LightningJR]

how could it not help to educate.. what if a few people stop being cynical and start understanding trends in technology better.
i mean someone spitting off a bunch stuff that is almost blah blah blah to me trying to be smart finishes by saying api overhead wont really affect performance yet microsoft and amd and demostrate much higher performance with lower latency.
there is just much that is fact that goes largely ignored because of misleading research by reviewers and bad information on forums.
they want people to have better information and not just amd and microsoft are taking initiative but also the oem's with more informative marketing.
i can give the business practice stuff a rest because i think everyone knows how i feel about that haha

You have a lot of trust in companies that make money off this information. I am not a cynic and I do think it's great that they Microsoft has finally fixed overhead in DX12 but how many games have a significant API overhead bottleneck? If developers didn't make their games based on what was possible at the time then they would have a poor performing game and that would kill PR and sales.

My best speculation is that future games will have more potential, current games if updated to DX12 and take advantage of new API improvements will range from decent FPS improvement to minor increases depending on game.

On to the topic.

HBM is revolutionary, no one can deny that if they look at history and understand the technology. The gain in board real estate alone is amazing and that's not taking in to account the power improvements, extra bandwidth, simplicity of block design and others I can't think of. It seems to be the next evolution of VRAM and I like it. I don't know how much performance will be gained by having all that extra memory bandwidth (I am an overclocker and like all other overclockers we know the what the real world benefits of getting more memory bandwidth are and they are slim compared to increasing core clocks) The other concern is the 4GB of VRAM, if we're to assume 4k, it may become an issue in some circumstances and if the trend continues games will continue to use more and more VRAM in the future, obviously plateauing eventually (likely around the times when consoles are at their limit). Compression technologies can only go so far if lossless, look at audio, when was the last breakthrough with lossless audio codecs?

I could say so much more about it all but will stop there.

I never mislead anyone when I post a reply, everything I write is plausible and not pandering to either company or fan-boyism, if I am wrong in anything I will concede based on testable results.

You're not informing people of the facts you are giving us information from slides from the the companies that are selling this product. You need to have a conversation on the topic/technology, further the conversation by bringing constructive arguments and points. Stop taking AMD's and Microsoft's AND NVidia's AND any companies word about their own product that are trying to make a profit from.

 

[xfia]

i could agree with you about the trend of vram usage except the latency improvement with dx12 appears to mean a lot including that vram bandwidth needs will be reduced and no one will need to worry about having 4gb for 4k.
the way that dx12 is implemented is that more of it is a standard especially regarding the command buffer and async shaders. it will make engines for dx11 obsolete but visual enhancement tools will still be popular.
there is more regarding built in bare mental optimizations for specific architecture's so strengths can be well exploited.
i dont think it would be bad information to say it would be wise for buyers to wait on a nv gpu for them to demonstrate significant api overhead reduction in dx12.
not saying that nv couldnt pull a ace but microsoft and amd have been working together for years on the game systems improving amd's api overhead and exploiting the strengths of gcn.

 

[Frick]

Edit:
Honestly, I'm tired of both sides of this insanity. Neither AMD nor Nvidia is doing anything ground breaking yet. They're stuck on the same node, and treading water. They've gotten communities like TPU into an uproar, with nothing to actually be excited about. Consider me too tired to continue giving a crap about any of this. My only pleasure left is taking fan boys with me on this spiral into bitterness.

Aye, this is Molyneux level hype, but all around.

 

[Frag_Maniac]

Just let the 25th come and let's see for ourselves.

Yeah, what else do we have to do but finally look at benches vs speculate. Arkham Knight is a mess and will probably take some time to clean up, so might as well nerd out on the Fury vs 980 Ti debacle.

 

[xfia]

Yeah, what else do we have to do but finally look at benches vs speculate. Arkham Knight is a mess and will probably take some time to clean up, so might as well nerd out on the Fury vs 980 Ti debacle.

they got the backs of gamers wanting to kick ass as batman yesterday
http://support.amd.com/en-us/download/desktop?os=Windows+7+-+64

 

[SonicZap]

I have a special issue against the exclamation mark spam in these topic titles!!! Really, exclamation marks don't belong to topic titles and just make you look immature.

 

[RCoon]

good thing i got several threads to help fight off the troll haha

Meanwhile, in the real world, somebody who makes several threads consisting of largely the same thing gets banned for forum spam. I don't think the thread should be closed, and perhaps you shouldn't even receive a warning, after all, you've not really done anything wrong.

I would however recommend you take a good hard look at yourself and how you've behaved in each of these threads, and what the general self-fulfilling purpose of them is, and consider how the community is going to view you from now on.

 

[erocker]

i guess i interrupted some riveting conversation happening.. tho it doesnt seem that way.
maybe it was on that so informative and tasteful article that amd doesnt trust its own cpu's?
anyone actually wanna talk about real things going on technologically or just argue about facts?
strange @erocker closed the one that further proves what i have been saying about the mods around here being cynical and not listening to what engineers have been saying for years and didnt close this one that was already a article i didnt see.

I could not care less about the topic. The thread was closed due to you spamming news articles, most of which are on the same topics. Nobody likes a blabbermouth.

 

[Aquinus]

anyone actually wanna talk about real things going on technologically or just argue about facts?

What?! Since when are "real things" and "facts" two different things?

I could not care less about the topic. The thread was closed due to you spamming news articles, most of which are on the same topics. Nobody likes a blabbermouth.

Yes, but no one like an arrogant ass either. Even more so when they can't even back up their own claims with anything other than AMD PR propaganda. I'm also getting tired of the constant rhetoric that I'm seeing in just about every post Xfia is making.

 

[Batou1986]

This article is hilarious, comparing die size to die size plus memory footprint woo ~60% smaller, my sides are in orbit
let's compare actual die size

Just shows you how much forbes doesn't know
I hope AMD put as much money into R&D as they did marketing.

 

[xfia]

i wanted to insert everything the threads i already had but they felt largely crapped up to.. yes partially due to me but im not the best at handling a hardcore troll that has nothing to do with anything but insulting. i could be nicer all around and yeah its been a curse and blessing for me how quickly i can learn compared to some.
not to go on.. yeah sorry for being a ass @mods

 

[LightningJR]

i could agree with you about the trend of vram usage except the latency improvement with dx12 appears to mean a lot including that vram bandwidth needs will be reduced and no one will need to worry about having 4gb for 4k.
the way that dx12 is implemented is that more of it is a standard especially regarding the command buffer and async shaders. it will make engines for dx11 obsolete but visual enhancement tools will still be popular.
there is more regarding built in bare mental optimizations for specific architecture's so strengths can be well exploited.
i dont think it would be bad information to say it would be wise for buyers to wait on a nv gpu for them to demonstrate significant api overhead reduction in dx12.
not saying that nv couldnt pull a ace but microsoft and amd have been working together for years on the game systems improving amd's api overhead and exploiting the strengths of gcn.

I am not sure how you see lower latency lowering VRAM usage, unless I am missing something that makes no sense.

Most of the optimizations being implemented in DX12 are already in Mantle, did you know this? All of the big numbers you see thrown around about "**% performance increase over DX11" by Microsoft and AMD are just % improvements on that specific area of the API and are not a direct % increase on FPS. If you look at Mantle vs DX11 numbers in actual games the performance improvement is on average only +%5 if you have a good Intel CPU. The only area Mantle really shows it's legs is when there's a CPU bottleneck, I expect AMD created Mantle to offset their lower than Intel CPU performance and they succeeded to an extent. I don't understand what what you mean by this statement "the way that dx12 is implemented is that more of it is a standard" you'll have to clarify that for me.

If Mantle is any indication of the performance we'll get from DX12 then it's equivalent to an overclock of you GPU. This is only achievable if the games support DX12 which is quite a ways off too.

Wait for an NVidia GPU to demonstrate API overhead reduction? You know that 3DMark has a draw call benchmark for DX11/12 & Mantle right? The API overhead/Latency/Draw Call Benchmark are all basically the same. The API overhead is measured in ms which is latency, with a reduction of this the GPU can receive more draw calls from the CPU. This test has been done with current NVidia GPUs and they get great numbers.