AMD Confirms Press Conference for Computex 2017 - Vega is (Almost) Here

[Steevo]

AMD is seeing the ligght at the end of the tunnel.

1) For many years AMD/ATI leapfrogged Nvidia with smaller process nodes to gain performance advantage with smaller chip sizes. It worked well for the most part.
1a) This ended when process node development stalled. 28nm anyone?
1b) AMD was and still remains tightly tied to a single foundry for financial reasons, even when it hurts them.

2) ATI before AMD bought them was free to do what worked best for graphics, as that WAS their business.
2a) AMD made a point of pushing the ATI team to play ball with integrating and making more modular systems that played better with APU/CPU technology
2b) AMD tied graphics to a single supplier, see 1b.

3) AMD spun off many of the parts of ATI that were becoming more relevant.
3a) Mobile anyone?
3b) My TV has an AMD modulator demodulator, now there are many companies that do this with ARM products that ironically feature ATI tech.
3c) AMD's piss poor direction at first caused a huge problem for them up till recently when they finally took notice they weren't the cock of the walk and the IP they have is as if not more valuable than the chips they make.

All of us know Ryzen isn't a magic bullet, the same that VEGA isn't. Its going to be at least a couple years of strategic actions and planning before AMD is as great as they once were, and unfortunately many new people will continue to play games on the GT240 and 2500 and bash AMD for not having a competitor to Nvidia or Intel that "blows it out of the water", not realizing that the money Nvidia and Intel make isn't so much on high end consumer stuff, but on server and other hardware.

 

[ZoneDymo]

Jesus f-ing Christ, not this shit with AMD being broke and being borderling bankrupt for the past xyz years again.
Seriously, stop it. Or go to Wfctech.

its a bit like that "pc gaming is dying" bs that seems to stick with people

 

[xkm1948]

This is getting fun. Pitching crazy fan boys against each other.

Don't forget kittens, we all belong to the PC MASTER RACE, so play nice!

 

[TheoneandonlyMrK]

Never Forgotten How can you forget anything AMD these days, it's bread crumbs are in every headline

No argument from me that AMD has intended this release to be competitive for them.
For the new arch in 6 years , used for 1 - 2 Hopefully you're wrong there. If it takes 6 years to design a new architecture, you better use it for 6 years
AMD cards have always been "powerful" cards. But whether be it drivers or just implementation, they rarely seem to live up to their potential. Though I do agree with the AMD "fine wine" approach. Even with that, you can only squeeze so much before the juice runs dry.

I have my fingers crossed it does kick ass, that would so cool. However I really wonder that if they have something that competes with Volta, why not wipe the floor with Pascal today with it, and ride the wave into Volta's release. Why hold your hand......... this smells a little off to me.

I will only comment on the die-shrunk Maxwell comment. Technically speaking, Maxwell was a larger die pre-implementaiton of Pascal. Not that other way around....

Anyways, like I said. I REALLY REALLY would love to see AMD release a card that just wipes the floor in the current market. One that is so powerful that they survive two life-cycles from Nvidia, which is what I think we were all expecting now. Not sure how to end my statement without trying to lean one way or the other, other than saying yes, fingers crossed, but not holding my breath anymore. I have mentally moved on........

Its not a mystery why they haven't always shown it in games , it's the way its meant to play.

Plus nvidia ditched double compute on consumer cards for the most part ergo efficiency gained.

Plus the breadcrumbs likely are bull and despite getting blamed they don't usually originate from Amd ,why would they.

When nvidia are releasing something ,then Amd release a tidbit of info not every other day a fairly average bench , how do some of you connect these dots?

 

[Cybrnook2002]

Don't forget kittens, we all belong to the PC MASTER RACE, so play nice!

And that's really the double edged sword isn't it We "should" all be happy for the advancement of things that advance us, instead we bicker and argue back and forth

But, don't be fooled, that's a business model and a marketing strategy that Manf's have happily tapped into. We get things like this thread, that turn into heated and heated discussions about things we have no idea about, to the point where it becomes polarizing. And at that point one of the two sides now have a dedicated customer I think we all argue back and forth as much as those companies go back and forth. OR, maybe Lisa is over an Jen's house right now smoking a stogie with her feet up and a glass of cognac, pitching the next 10 years.

And I think, in my mind, the REAL advancement opportunities for these companies are in Big Data crunching, and of course the "deep learning" autonomous coming in. THIS is where the real directions are. These card's architectures just so happen to also run games, so a double dip.

I know we have Nvidia CUDA farms for data crunching here at work.

 

[xkm1948]

And that's really the double edged sword isn't it We "should" all be happy for the advancement of things that advance us, instead we bicker and argue back and forth

But, don't be fooled, that's a business model and a marketing strategy that Manf's have happily tapped into. We get things like this thread, that turn into heated and heated discussions about things we have no idea about, to the point where it becomes polarizing. And at that point one of the two sides now have a dedicated customer I think we all argue back and forth as much as those companies go back and forth. OR, maybe Lisa is over an Jen's house right now smoking a stogie with her feet up and a glass of cognac, pitching the next 10 years.

And I think, in my mind, the REAL advancement opportunities for these companies are in Big Data crunching, and of course the "deep learning" autonomous coming in. THIS is where the real directions are. These card's architectures just so happen to also run games, so a double dip.

I know we have Nvidia CUDA farms for data crunching here at work.

CUDA/OpenCL has been contributing to the explosion of Bioinformatics mining for quite some time now. It is truly a wonderful freaking diamond mine that requires some serious computation power. Nvidia caught on this just in time.

 

[_larry]

This is all a load of hot air at this point. No evidence VEGA is going to trouble the 1080Ti, yet it won't be much cheaper I am quite sure. 1080s can be had super cheap now, so VEGA needs to be SERIOUSLY competitive if it has the slightest chance succeeding at this stage in the game.

1080? Cheap???
I hope I get to a point in life where $525~ is chump change.
I was happy to find my R9 290 for $200 a year ago!

 

[Duality92]

Dang guys, chill. Vega will be out soon, we'll get numbers from the reviewers and actual leaks soon enough.

On that note, even if it would beat a 1080 ti, the price needs to reflect that, if AMD thinks beating a 10% is worth 20% more $$, we'll still be crying about the price.

Big Vega should, by the timeline we're currently on, match a 1080 Ti (±5%) for 5-10% less $$. Big Pascal has been out long enough for this to be needed. If they're too late, 1180 will be out, with 1080 Ti performance with 1080 price a tag and it won't be competitive.

 

[TheLostSwede]

1b) AMD was and still remains tightly tied to a single foundry for financial reasons, even when it hurts them.

Huh? TSMC makes AMD's graphics parts and GloFo makes their processors. Please read up on things before you write dumb comments https://www.extremetech.com/computi...samsung-could-tap-foundry-for-future-products

Considering how long you've been on TPU, you ought to know better.

 

[Ferrum Master]

Isn't fullmoon? Here and in Vega thread all went cuckoo.

Okay, I forgot to brush my teeth this morning, I guess same happens with prescription pills.

 

[moproblems99]

Its not a mystery why they haven't always shown it in games , it's the way its meant to play.

Plus nvidia ditched double compute on consumer cards for the most part ergo efficiency gained.

Perhaps AMD should have done the same? What good does double compute do many of us right now? Most of the complaints about AMD are power and heat. Axing those paid off for nv in that regard so it could have helped AMD as well.

I look forward to Vega as I most likely will purchase one to upgrade my 980 from.

 

[R-T-B]

AMD has had a HUGE perf/mm advantage:

Since Maxwell that hasn't really been true.

 

[cdawall]

Well I just bought a 1080Ti Strix. I am tired of waiting. Sorry AMD, I am done playing the one day vega will be here game.

 

[R4E3960FURYX]

Mathematics today for RX Vega Performance preview.

RX VEGA 1200MHz 700MHz HBM2 3DMark Time Spy 1.0.
Graphic score was 5721 point.
RX VEGA 1600MHz 900MHz HBM2 may = 5721 x (4/3) = 7628 point.

RX VEGA 1200MHz 700MHz HBM2 3DMark FireStrike.
Graphic score was 17801 point.
RX VEGA 1600MHz 900MHz HBM2 may = 17801 x (4/3) = 23735 point.

Could it faster than your card? check.

 

[Prima.Vera]

...1080s can be had super cheap now,

Define "super cheap".

 

[cdawall]

Mathematics today for RX Vega Performance preview.

RX VEGA 1200MHz 700MHz HBM2 3DMark Time Spy 1.0.
Graphic score was 5721 point.
RX VEGA 1600MHz 900MHz HBM2 may = 5721 x (4/3) = 7628 point.

RX VEGA 1200MHz 700MHz HBM2 3DMark FireStrike.
Graphic score was 17801 point.
RX VEGA 1600MHz 900MHz HBM2 may = 17801 x (4/3) = 23735 point.

Could it faster than your card? check.

That's slower than a regular 1080 and isn't besting a 1070 by much. That is comfortably slower than a 1080Ti.

This is a rig I personally built and ran 1800x+1070.

http://www.3dmark.com/spy/1447068

 

[TheinsanegamerN]

It will be interesting to see how the VEGA falls on the price/performance spectrum, and it's final price. Nvidia has a lot of power here, and AMD cant afford a price war, and have more expensive HBM hamstringing any price cuts.

Hopefully it is 1080 level, to provide an AMD choice for what is a fast growing lucrative mid range market. I expect a mix of 1070/1080 performance on average, however.

Its not a mystery why they haven't always shown it in games , it's the way its meant to play.

Plus nvidia ditched double compute on consumer cards for the most part ergo efficiency gained.

Plus the breadcrumbs likely are bull and despite getting blamed they don't usually originate from Amd ,why would they.

When nvidia are releasing something ,then Amd release a tidbit of info not every other day a fairly average bench , how do some of you connect these dots?

Overclocking ability, cheaper cost to manufacture, lower rate of faulty parts, and bigger chips are all a side affect of that "ergo gain".

That "ergo gain" allowed nvidia dominant control of the gaming laptop market, which continues to grow, as well as being able to put together monster titans and x80tis that dominate the enthusiast desktop space, produce arches like maxwell and pascal that OC incredibly well and are cheap to build, while AMD has to use hugely expensive HBM and water coolers in a vain attempt to keep up.

The less complicated arch is also easier to tweak, allowing nvidia to continue improving capability even when stuck on a transistor size for too long. AMD, OTOH, was hamstrung, with no real answer to maxwell, and falling a full generation behind on efficiency 18 months later.

And of course, that ditching of compute parts that consumers do not use allowed nvidia to make a ton of cash and capture a huge marketshare.

 

[TheoneandonlyMrK]

It will be interesting to see how the VEGA falls on the price/performance spectrum, and it's final price. Nvidia has a lot of power here, and AMD cant afford a price war, and have more expensive HBM hamstringing any price cuts.

Hopefully it is 1080 level, to provide an AMD choice for what is a fast growing lucrative mid range market. I expect a mix of 1070/1080 performance on average, however.
Overclocking ability, cheaper cost to manufacture, lower rate of faulty parts, and bigger chips are all a side affect of that "ergo gain".

That "ergo gain" allowed nvidia dominant control of the gaming laptop market, which continues to grow, as well as being able to put together monster titans and x80tis that dominate the enthusiast desktop space, produce arches like maxwell and pascal that OC incredibly well and are cheap to build, while AMD has to use hugely expensive HBM and water coolers in a vain attempt to keep up.

The less complicated arch is also easier to tweak, allowing nvidia to continue improving capability even when stuck on a transistor size for too long. AMD, OTOH, was hamstrung, with no real answer to maxwell, and falling a full generation behind on efficiency 18 months later.

And of course, that ditching of compute parts that consumers do not use allowed nvidia to make a ton of cash and capture a huge marketshare.

so does the fanboi element it allowed them to up prices for mid range build,priced cards, to top notch ,win. design more concurrent asics , in fairness opening new revenue streams and gouge everyones eyes out akin to apple on crack, thats my opinion and its true and unbiased as was my last.

 

[moproblems99]

so does the fanboi element it allowed them to up prices for mid range build,priced cards, to top notch ,win. design more concurrent asics , in fairness opening new revenue streams and gouge everyones eyes out akin to apple on crack, thats my opinion and its true and unbiased as was my last.

I used to say everyone who bought Titans was stupid, nv is price gouging, and nv is selling mid-range gpus at high-end prices. However, for points 1 and 2, if the market will bear it - keep selling. For point 3, if an nv mid-range can sometimes outpace an AMD high-end then something is wrong. I hate to say it, but if I had the best crack in town I would keep raising the price until people stop buying it or someone starts producing crack that is just as good and cheaper.

When AMD starts catching up, things will change. Don't get me wrong, I have an a Ryzen coming up in my future and most likely a Vega. I would have bought my Ryzen on launch day but I got tired of waiting and bought a Sandy Bridge Xeon to use as my vm server.

AMD's biggest problem is that they can't get stuff out on time. You can clearly see that Vega was slated for early 2017 and end of May and early June is not early. Most people have already bought a 10 series and those that have a 9 series or 300 series are probably now going to wait until Volta to see which is better and by how much. I have been contemplating buying a 1080Ti because I am tired of waiting.

 

[R4E3960FURYX]

Next week 16th May we gonna see what AMD Vega GPU spec or details.

Many people are waiting to analysis them.

 

[Steevo]

Huh? TSMC makes AMD's graphics parts and GloFo makes their processors. Please read up on things before you write dumb comments https://www.extremetech.com/computi...samsung-could-tap-foundry-for-future-products

Considering how long you've been on TPU, you ought to know better.

RX480 and how many APUs?
Again....28nm?

Shall I ask the age old question of is a shot sarnie better with bread or toast?

 

[jigar2speed]

Dang guys, chill. Vega will be out soon, .

No Shut up, we want to bash /s

 

[Assimilator]

People really need to educate themselves on how important die size is, and how often AMD has had a HUGE perf/mm advantage:
View attachment 87816

Vega is going to be 500 - 533mm^2. By all means it should be stronger due to its size.

The only thing that imo can really hold them back is if GF is still garbage when it comes to clockspeeds. If AMD can't get above 1400 MHz, then they will only compete with the 1080 Ti, if they get to 1600 MHz they will compete with the Titan XP, and if they hit 1700 MHz they will crush everything in their path.

Using die size as a predictor of performance.... yeah.
Using clock speeds as a predictor of performance... yeah.

 

[cdawall]

For comparison using a pair of 480's as a basis for performance on vega. (Vega is equal to two in specs, but in theory weaker per clock due to longer pipelines)

 

[R4E3960FURYX]

Vega cannot beat 1080Ti overall.