AMD Next Gen Compute Unit?

>>58430766

Its only accurate if you cant get enough buzz words and wishful thinking.

>>58430766
Not really. AMD always disappoints on product launches.

>bulldozer will be good is 8 core
>polaris is good, crossfire will be better than 1080 but cheaper
>fury x will beat a 980 ti in several cherrypicked games using settings specified by AMD (not ultra)
>7970 is better than a gtx 680

On and on and on

>>58430836
True. I mean it's nice their cards age well a year later but by then nvidia releases a new card that shits on them.

>>58430766
>How accurate is this pasta?
The guy who first wrote this is asserting some speculation as if its factual, when it hasn't been confirmed. Its basic information mixed with misinformation.

>Vega is the biggest fundamental architecture change since they switched from Terascale to GCN 1.0
Fact.

>The new NCU means a massive boost to efficiency in the shader core and a much easier ability to fully saturate the GPU with work.
This is what one would call a non statement. Might as well say fast things are fast.

>The biggest problem with Fiji(furyX) is that its 4096 shaders didn’t mean very much because of large bubbles in the pipeline and large amounts of shaders idling and doing no work each cycle.
This isn't totally accurate. Fiji is really ROP limited, and has a slight issue with its memory controller providing lower throughput than it should.

>Even with a low Level API like Vulcan which is best case scenario the FuryX is still not running with all cylinders firing because Vulkan only solved the Driver overhead problem.
In correct. Vulkan and DX12 similarly let the GPU do more by letting more than 1 CPU thread be utilized for draw queues among other things. Drivers themselves are a separate and much more complex issue.

>There is still bubbles in the workload because of the fundamental design of the GCN architecture itself.
Another non statement.

>>58430870

That needs to be upgraded as AMD keeps unleashing new tech thats compatible with their original building blocks for the last 5 years.

>>58430870
I just want a launch that isn't disappointing. I have an A8 7650k and love it, but it's only my pc at my parents home. I use a 4770k at home and a GTX 980.

I'll buy from whichever brand as long as its offering better value for the performance I want, but it's kinda disappointing that AMD doesn't meet that so often.

>this time it's really going to be good, i promise

>>58430885
>AMD is switching from standard 4x16-bit blocks of a normal GCN CU(compute unit) to This new design of a variable 8+4+2+2 block and either four of these or a mix of these and standard 16-bit blocks in this new NCU(next compute unit)
This is 100% speculation based on an AMD patent that was circulating a while ago. There is no confirmation that this patent is Vega, and it is actually very unlikely. The patent described variable size CU for the purpose of saving power. With Vega AMD touted the ability of an NCU to process 128 FP32 ops per clock.
This has to mean that they magically found a way for a 64ALU CU to process two single precision OPs in a single clock, or the CU is now 128ALUs.
There is no indication anywhere that the NCU is of varying size, and AMD's Vega slides hint against it.

>They could not do this in the past because they did not have the technology to build a Hardware Scheduler and Memory Scheduler Complexed enough to run this kind of architecture.
This is just an asspulled guess.

>We have test samples of Vega with NCU's disabled and with very low clock speeds and its beating a 1900Mhz overclocked GTX 1080 in DOOM.
There is no indication whatsoever that the Vega 10 engineering samples AMD showed off had any NCU binned off. There is no indication that they were running at low clocks either.
This simply is a lie.

>A lot of people dismiss the test because it’s Vulcan but they don’t know what they are looking at. The Vega sample is running FIJI 1.2 drivers
The Vega engineering samples were in fact using drivers not tailored for the new arch, this much is correct.

>This New architecture is more advanced than Maxwel/Pascal.
Naturally if follows that newer things tend to be more advanced than older things.

The fanboy who typed all this up only has a very loose grasp of basic facts.

>>58430940

Damage control

>>58430766
Just shill for amd and laugh at retards who wasted their money in support threads. That's what I do. ;^)

>>58430940
>This is 100% speculation based on an AMD patent that was circulating a while ago.

It's not speculation anymore, it was confirmed in the architecture reveal a week ago. How much do you get paid for this damage control?

>>58431472
No, retard, no it wasn't.
The NDA on the Vega arch was lifted just a few days ago, and it included absolutely nothing from the patent in question. No variable size CU, no increased sALUs. Nothing. This is a total asspull on behalf of the retard who wrote that copypasta.

http://www.anandtech.com/show/11002/the-amd-vega-gpu-architecture-teaser

Don't try to correct people who know more than you.

>>58431520
Those slides literally show the new compute units you fucking retard.

>>58431563
Is your IQ 50 or are you intentionally being a total fucking retard.
Exactly as I stated, plain as day:
>The NDA on the Vega arch was lifted just a few days ago, and it included absolutely nothing from the patent in question. No variable size CU, no increased sALUs. Nothing. This is a total asspull on behalf of the retard who wrote that copypasta.

The Vega arch does not utilize variable size CU as described in that shit tier copypasta. You not wrapping your head around this simple fact makes me think you're the dipshit who wrote it.

>>58430766
late as fuck If this shit not 3x faster then fury X then it is a failure. might as well make other midrange to save face