Seems like someone with a pretense of knowledge trying to cover up his disappointment. I suggest hanging out at B3D to try and learn what's really happening.
Seems like someone with a pretense of knowledge trying to cover up his disappointment. I suggest hanging out at B3D to try and learn what's really happening.Regarding the overall shortcomings of GCN, this post probably gives the best overview.
Highlights:
AMD's design spends so much space and energy tilting at windmills that it's at a massive size and energy disadvantage compared to NVIDIA's pared-down/graphics-oriented architecture. This chip uses twice the die size and twice the power of a GP104 for the same performance.
By "tilting at windmills" I mean that AMD devotes an enormous amount of hardware to solving general scheduling problems and work-stealing and spinning off async tasks to try and fill its pipeline bubbles. All of that die space and power go to something that is not drawing graphics on your monitor. AMD tries to optimize shitty instruction streams in hardware, rather than optimizing them on the CPU where it's fast like NVIDIA does and spitting out a single optimized command queue that minimizes pipeline bubbles in the first place. GPUs are not a general architecture and it's perfectly sensible to feed them in the way they like to be fed. It's not bad, it's not "cheating", etc etc. It's called bare-metal optimization.
What's really happening then?Seems like someone with a pretense of knowledge trying to cover up his disappointment. I suggest hanging out at B3D to try and learn what's really happening.
What is your source for this? Has it been posted by someone who would have a clue? Or, are you educated about HBM2 controllers?
They have launched Vega. People have it. RX Vega is the same size as Vega FE.
The reason is probably simpler: embarrassment.
Gen 2 is bigger.
https://www.hotchips.org/wp-content...VIN_TRAN-SKHYNIX-VERSION_FINAL-dcrp-t1-4_.pdf
I think AMD chose Rambus's PHY, and doing some digging, seems this is coming up in the pipeline from Samsung...
I wonder what the exact cost savings are?
To those saying fermi to kepler was complete architectural change. You are wrong. Kepler was fermi distilled. Kepler was still very much fermi.
http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/2
I'd say Pascal is a smaller change from Maxwell than Kepler was from Fermi, but the point is the same.Kepler : Fermi = Pascal : Maxwell
So move from hot clock 512SP architecture to normal clock 2880SP + move from hardware scheduling to software scheduling + delta color compression + other things is not big change ok...Its bigger change than what AMD did last 6 years with GCNTo those saying fermi to kepler was complete architectural change. You are wrong. Kepler was fermi distilled. Kepler was still very much fermi.
http://www.anandtech.com/show/5699/nvidia-geforce-gtx-680-review/2
So move from hot clock 512SP architecture to normal clock 2880SP + move from hardware sheduling to software sheduling + delta color compression + other things is not big change ok...Its bigger change than what AMD did last 6 years with GCN
You can argue that nothing is a new architecture if you follow the logic of some here. Every new architecture share similarities with the previous one. Even several architectures back.
Can you provide the slide deck used for launch, please?
AMD can claim that IPC have increased all day long for Vega as long as it beats Fiji in professional workloads. The same can be said about efficiency too.I wonder... How much faster than Fuji can we honestly expect VEGA to be? It has the same amount of ROPs, probably TMUs and shaders... Sure, it has some changes that are supposed to help with geometry especially but it's a big question mark how much those will help. By getting the clock speeds up the IPC might have somewhat got lower and the new things just help to make up lost performance clock per clock. They got the clock speed up a lot (hopefully) so the performance is up too. It's totally possible that VEGA FE shows more or less what to expect from VEGA RX (mature drivers likely add 10-15% more performance).
I guess AMD has similar problems with GCN now that what NVIDIA had with Fermi. Maybe they should really think again how to handle that scheduler. It might very well be the reason why VEGA is so large and seems to use lots of power.
Titan Xp costs $1199AMD can claim that IPC have increased all day long for Vega as long as it beats Fiji in professional workloads. The same can be said about efficiency too.
Easy for them to boast Vega as an amazing architecture as long as they dont compare against the competition (Nvidia). They have already tried one spin to try and sell the card, and that was with comparing Vega FE against Titan X, which is not a professional card, and boasting Vega as superior.
AMDs whole argument was that both cost the same ($999).
As for Vega RX and how they will instruct reviewers to spin the truth once reviews get out and people get the truth, hard to say really.
I think it will be about how much faster the card is vs Fury X in percentage, but doesnt want coverage how poor gaming IPC is vs Fury X and dont want no power draw comparison vs the competition.
As for RX Vega...
I think it will be about how much faster the card is vs Fury X in percentage, but doesnt want coverage how poor gaming IPC is vs Fury X and dont want no power draw comparison vs the competition.
Are you seriously saying Vega has not been launched? You're still in the first stage?
Not only is Vega out, they're releasing a water cooled version too.Are you seriously saying Vega got a full launch with no seed, no slide decks, no press meeting? I'm at the "i don't care" stage atm.
I highly doubt there will be any OC headroom to speak of.Price is the only thing that can really save RX Vega at this point. Even though benchmarks show it currently slower (on average) than a 1080, I think RX Vega will somehow end up 5% faster than 1080 on average, but will probably see it's small lead entirely evaporate when both are OC'd.
Yeah, pretty much, and nvidia's counter punch may just KO the entire launch price(s), assuming nvidia wants to do that.I personally just don't see RX Vega coming in at less than $550 - a price point which will unfortunately not attracted anyone besides die hard AMD fans. But, as with all AMD 2nd tier cards, if there is a cutdown version I'm guessing it will be 10-15% faster than the 1070 for $399-449, making it much more attractive in perf/$ than RX Vega.
They say, over three times faster than previous iMac GPUs.The Radeon Pro Vega is over three times faster than any previous iMac GPU, packing the power of a double-wide graphics card into a single chip. All of which translates to higher frame rates for VR, real-time 3D rendering, more lifelike special effects, and gameplay at max settings. It’s one huge reason iMac Pro is power incarnate.