Support NeoGAF

[GameAddict411]

1080ti stock is 11.3 tflop FP32, so I'm assuming this implies there are still architectural differences in Nvidia's favor beyond flop count.

It's always been like this. On paper, the RX 580 should go tie to tie with the GTX 1070 while in reality it's on par if not slightly faster than a GTX 1060. The same goes to the TDP. AMD is way behind in power and architectural efficiency.

 

[Timedog]

It's always been like this. On paper, the RX 580 should go tie to tie with the GTX 1070 while in reality it's on par if not slightly faster than a GTX 1060. The same goes to the TDP. AMD is way behind in power and architectural efficiency.

Isn't Vega supposed to be a large architectural jump?

 

[Rushersauce]

Welp, I'll wait for Volta.

 

[SmartWaffles]

don't know why they would do that, the crossfire probably isn't ready for it yet.

Then why on earth are they showing it on a streamed press conference when it was not ready....

 

[NeOak]

Isn't Vega supposed to be a large architectural jump?

supposed is the keyword

 

[Elixist]

the apu was the only exciting thing. that looked pretty sexy. yawn as fuck at all the other shit, tell us the deets on rx vega already.

 

[DemonCleaner]

see you at siggraph...ah yeah thanks amd.


all the gaming hardware youtubers who did go to taipei will be mighty pissed.

 

[Aztechnology]

Nvidia conference: "We just made the world's most powerful laptop GPU in an ultra thin form factor"

AMD:

I wish that we're even the case. You couldn't find a 580 even if you wanted to.

 

[Xyber]

Come on now AMD, this really isn't good enough. =/

Releasing something that will most likely still be behind the 1080Ti not that long before Volta might be a thing. They will pretty much be a whole generation behind at that point for quite some time after that.

 

[VariantX]

Wondering what did they really have to talk about Vega that they really should have saved for another event. Almost 50 minutes of stuff we already knew about with a few extra demos, 5 minutes of telling us our princess is in another castle.

 

[theultimo]

Come on now AMD, this really isn't good enough. =/

Releasing something that will most likely still be behind the 1080Ti not that long before Volta might be a thing. They will pretty much be a whole generation behind at that point for quite some time after that.

before Polaris it was worse.

however knowing Volta is coming sooner rather then later might be a big blow to Vega, unless they price it accordingly.

 

[DieH@rd]

So... what did I miss? Any news?

 

[RandomExpletive]

AMD |OT| PC? I Love It to Threads

AMD |OT| Hows his PC holding up? To Threads you say?

 

[maven]

AMD |OT| Hows his PC holding up? To Threads you say?

AMD |OT| Rip us this day our daily threads

 

[Caayn]

Wait, so all we got was info on Threadripper something about new laptops and basically another delay for Vega?

If AMD keeps this up NVIDIA will have Volta consumer cards out.

 

[Fess]

I missed the stream, is there any changes in the gaming performance race? Nvidia still in the lead?

 

[Aztechnology]

I missed the stream, is there any changes in the gaming performance race? Nvidia still in the lead?

They showed a clip of prey, for like 10 seconds running on Vega without any real infom

 

[Firenze1]

Amd is a joke and Vega is a dud.

 

[KingSnake]

Catching up with this and I guess there​ is a good reason why Nvidia skipped HBM2 so far in their consumer GPUs.

 

[llien]

It's always been like this. On paper, the RX 580 should go tie to tie with the GTX 1070 while in reality it's on par if not slightly faster than a GTX 1060. The same goes to the TDP. AMD is way behind in power and architectural efficiency.

RX 580 is a 230mm2 piece and is hence in no way comparable to 1070, your comparison is broken.

Power consumption figures are also nowhere as bad as perceived, people often jump from OC vs stock vs AIB back and forth. Stock 480 consumed about 35W more than stock 1060. (and a lot of that could be compensated by undervolting and even more than that can be saved with Radeon Chill, nvidia doesn't have that feature at all yet)

 

[llien]

Looking forward to Ryzen + Vega APUs no matter what.

Catching up with this and I guess there​ is a good reason why Nvidia skipped HBM2 so far in their consumer GPUs.

It was a risk nVidia didn't have to take.
AMD, on the other hand, being underfunded and an underdog had to gamble.

It seems it didn't quite work out.
My main concern is not that it doesn't take on 1080Ti, but that they are still not able to actually bring product to the market (faster than 1070/1080 is the biggest part of high end market anyhow)


Also, Raja speaking about that there is only a handful of GPU manufacturers because it is so hard to write drivers smells... like:

Lisa: dafuq Raja, why am I not impressed by how our HBM2 500mm2 chip performs?
Raja: Oh, that's... drivers! It rocks, but we don't have good enough drivers (points hands to dudes responsible for software)

 

[Risk Breaker]

Why not compare AMD to AMD instead of having this endless discussion about AMD flops vs nvidia flops?

Sooo doing some VERY basic math, if this has 64 CUs compared to the 36 in 480/580 meaning 1.77x, and those have a Time Spy score of ~4400, Vega would have a score of 7800, which is like 650 points over a 1080.

Obviously this doesn't account for HBM2 and it implies same clocks, but it doesn't seem bad to me depending on price (hoping for 399).

Without knowing the price and potentially being between a 1080 and 1080ti, with HBM2, I don't see why people would call it a dud already?

 

[llien]

Nvidia conference: "We just made the world's most powerful laptop GPU in an ultra thin form factor"

According to anand's article, all the "Max Q" is about is basically good old M chips with a new name sold at a premium.
They also consume 60-110W so hardly fit ultrathins to be honest.

Without knowing the price and potentially being between a 1080 and 1080ti, with HBM2, I don't see why people would call it a dud already?

It comes too late.

 

[00ich]

Also, Raja speaking about that there is only a handful of GPU manufacturers because it is so hard to write drivers smells... like:

Lisa: dafuq Raja, why am I not impressed by how our HBM2 500mm2 chip performs?
Raja: Oh, that's... drivers! It rocks, but we don't have good enough drivers (points hands to dudes responsible for software)

Once the launch reviews are out a cards reputation is hard to change and a few badly optimized titles make your card look worth less than the competition. That was a real problem with the 480.

 

[DieH@rd]

AMD |OT| Rip and Thread until it is done

 

[Hesemonni]

When is Volta coming out again?

I guess it's team green for me too. Enough waiting is enough.

 

[00ich]

Without knowing the price and potentially being between a 1080 and 1080ti, with HBM2, I don't see why people would call it a dud already?

It's not, but people are losing their patience. Now we get Vega 10 which is supposed to be 1070-80 level. Then it'll be Vega 11 wich will be at 1080ti level.
Also there's Volta looming, probably Q1 '18. That comes already on a smaller manufacturing node and a 2060 might rival the Vega 10 already. 2070 will be an Vega 11 contestant and there will be a 2080. All expected at most half a year after Vega finally launches.
So it looks like AMD is 2/3 of a generation behind Nvidia with Vega, which is disappointing because they were close at beating Nvidia with the Polaris.

 

[Firenze1]

 

[AmyS]

AMD |OT| Rip and Thread until it is done

Bwuhahahaha.

 

[ThoseDeafMutes]

Without knowing the price and potentially being between a 1080 and 1080ti, with HBM2, I don't see why people would call it a dud already?

If AMD's answer to the 1080 / 1070 launched alongside the 1080, people wouldn't. The 1080 will be a full year old by the time this launches and the 1080ti will be 5-6 months old. When you launch 6-12 months after your competition, matching their performance is underwhelming.

Nvidia gets to have their cake and eat it too by enjoying the initial extremely high margin sales for a year, then putting the squeeze on AMD by discounting them just before Vega launches. Volta is expected at the tail end of 2017, meaning outlook not so good for team red.

 

[Delusibeta]

I realise that I was the guy warning against manufacturer benchmarks, but that Prey demo was really bad.

 

[J-Rzez]

Yikes that demo..

I have a bad, bad feeling about Vega at this point. It has to be a good position for Nvidia. They can start with the 11xx teasing. Get into a price war easily. And, if need be at unveiling, just do nothing at all.

 

[Durante]

Also, Raja speaking about that there is only a handful of GPU manufacturers because it is so hard to write drivers smells... like:

Lisa: dafuq Raja, why am I not impressed by how our HBM2 500mm2 chip performs?
Raja: Oh, that's... drivers! It rocks, but we don't have good enough drivers (points hands to dudes responsible for software)

To be fair, GPU drivers are some of the most complex software projects on the planet.
So that's not really an excuse, just a statement of fact.

That's one of the main reasons AMD is/was pushing low-level APIs so hard. Of course, that just pushes the low-level per-GPU optimization to game developers, many of which are even less equipped for that than GPU manufacturers.

 

[Unknown Soldier]

If AMD's answer to the 1080 / 1070 launched alongside the 1080, people wouldn't. The 1080 will be a full year old by the time this launches and the 1080ti will be 5-6 months old. When you launch 6-12 months after your competition, matching their performance is underwhelming.

Nvidia gets to have their cake and eat it too by enjoying the initial extremely high margin sales for a year, then putting the squeeze on AMD by discounting them just before Vega launches. Volta is expected at the tail end of 2017, meaning outlook not so good for team red.

Vega will be launching just as Nvidia will be preparing to launch Volta and leap another generation ahead. At this point AMD is hopelessly behind in GPUs, lagging now by a full generation. When you're launching your competitor to their last generation just as they are about to launch their next generation, you are generally regarded as being in deep trouble in the fast-moving tech industry.

 

[neurosisxeno]

To be fair, GPU drivers are some of the most complex software projects on the planet.
So that's not really an excuse, just a statement of fact.

That's one of the main reasons AMD is/was pushing low-level APIs so hard. Of course, that just pushes the low-level per-GPU optimization to game developers, many of which are even less equipped for that than GPU manufacturers.

It was literally legitimizing a stigma about AMD GPU's that they claim was unfairly attributed to them from back when it was ATi. Now the Nvidia Fanboys can literally just say "Even Raja says their drivers suck".

 

[ThoseDeafMutes]

It was literally legitimizing a stigma about AMD GPU's that they claim was unfairly attributed to them from back when it was ATi. Now the Nvidia Fanboys can literally just say "Even Raja says their drivers suck".

The old AMD drivers suck issue was always a slippery accusation. The original form I heard it in 10 years ago was that ATI drivers were unreliable and crash prone. I was still hearing that in the 2010's and by that stage it was pretty untrue from what I could tell. Later it was that they were less efficient or well suited to games. That at least had some truth to it. The 290 was always a good buy, it's just that they were under-performing compared to their potential at launch. Which of course gave rise to a counter-points from the AMD side, which was that AMD cards aged better. In both cases, not really untrue, but both also a bit too specific to one era of cards to say it was a trend for the companies.

 

[martino]

The old AMD drivers suck issue was always a slippery accusation. The original form I heard it in 10 years ago was that ATI drivers were unreliable and crash prone. I was still hearing that in the 2010's and by that stage it was pretty untrue from what I could tell. Later it was that they were less efficient or well suited to games. That at least had some truth to it. The 290 was always a good buy, it's just that they were under-performing compared to their potential at launch. Which of course gave rise to a counter-points from the AMD side, which was that AMD cards aged better. In both cases, not really untrue, but both also a bit too specific to one era of cards to say it was a trend for the companies.

i left amd because of driver on my 6850 crossfire.
even with one card i was forbidden to use all options.
A combinaison of most aa + high texture filtering and i got systematic visual bugs on screen on most games at the time....
if it wasn"t that it was another combination of options ...
What i know is since i switched for a 980 ti when i scale down or disable something it"s not because it breaks visuals

Remembering that i now understand better this gen consolse problem with texture filtering...
And if my experience with nvidia change i will switch again (even if hdmi 2.1/freesync will probably make me switch etheir ways at some point)

 

[neurosisxeno]

The old AMD drivers suck issue was always a slippery accusation. The original form I heard it in 10 years ago was that ATI drivers were unreliable and crash prone. I was still hearing that in the 2010's and by that stage it was pretty untrue from what I could tell. Later it was that they were less efficient or well suited to games. That at least had some truth to it. The 290 was always a good buy, it's just that they were under-performing compared to their potential at launch. Which of course gave rise to a counter-points from the AMD side, which was that AMD cards aged better. In both cases, not really untrue, but both also a bit too specific to one era of cards to say it was a trend for the companies.

That's true, but having the head of Radeon Group try and hand wave a disappointing presentation by saying it's because of bad drivers is not good optics for a company like AMD which has that (not entirely accurate) reputation. If anything you probably should blame anything but the product you're presenting...

I can only speak personally, but I had issues back in the day with my HD4870 and HD6850 pretty regularly--the X800 I had before that ran great though.. After that I switched to Nvidia for a GTX670 (and held onto that until last year when I got a GTX1060) and had almost no problems since then. I can't speak to how smooth things were from the 7xxx series onwards but I heard they resolved a lot of issues.

 

[Tacklemacain]

I dont think the drivers are the problem, it is probably HBM2 availability. And his statement about drivers beeing the reason for not more competition. The drivers are a big part, why intel is still lagging behind with their integrated graphics.

 

[neurosisxeno]

I dont think the drivers are the problem, it is probably HBM2 availability. And his statement about drivers beeing the reason for not more competition. The drivers are a big part, why intel is still lagging behind with their integrated graphics.

That and the design of their iGPU's is much different from traditional GPU's--or so I've heard. On top of that they didn't bother seriously working on iGPU's until Sandy Bridge, so within the last what 5 years? Iris Pro is no joke considering AMD had basically a 25 year head start on them...

 

[Calm Mind]

Ryzen mobile was the only thing that caught my eye.

 

[Rodin]

Ya, doesn't bode super well for the card if it can't run Prey at 4K w/ a decent frame rate. That game isn't even very demanding.

That was one of the least impressive "reveals" i have ever seen.

I didn't watch the event (it was around 4-5am here), are we sure it wasn't the stream quality? Wouldn't be the first time.

 

[Steel]

Ryzen mobile was the only thing that caught my eye.

Threadripper boards having >60 PCIE lanes was interesting, too. Wonder what the price will be.

 

[V_Arnold]

Okay, I am saving up for an 1070.

 

[Calm Mind]

Threadripper boards having >60 PCIE lanes was interesting, too. Wonder what the price will be.

Too rich for my blood MSRP.

 

[Colbert]

AMD is really really bad at PR and marketing. A simple press release had done a better job than a stream with almost no info in it.

 

[kami_sama]

So, in the end, the conference was for nothing, right?
Why the hell do they do this? It only hurts them.

 

[llien]

Threadripper boards having >60 PCIE lanes was interesting, too. Wonder what the price will be.

Max i9 has is 44 lanes and that only on $999 7900x.
Models below use 28, gg to those who want to run dual GPU with full x16 PCIe.

And that while whoknowshowmuchbutsurelycheaprthaniintels ThreadRipper comes with yeehaa 64 lanes, so you could go quad GPU at full x16 PCIe. And it is soldered too, vs TIM on Intel's lineup.

It's actually so interesting that Intel will inevitably lose some of the HEDT market,

To be fair, GPU drivers are some of the most complex software projects on the planet.
So that's not really an excuse, just a statement of fact.

That's one of the main reasons AMD is/was pushing low-level APIs so hard. Of course, that just pushes the low-level per-GPU optimization to game developers, many of which are even less equipped for that than GPU manufacturers.

Yeah, complex, so what. Is it complex to a point of "Intel's GPUs are underwhelming, because Intel doesn't know how to write drivers"? (Raja's point) I doubt it, to be honest.

So, in the end, the conference was for nothing, right?

Bottom line:

1) ThreadRipper looks shiny
2) Vega isn't there yet
3) Vega's is highly unlikely to beat 1080Ti
4) OEMs are all onboard with Ryzen, and even Ryzen + Polaris notebooks are available (Dell has that lovely 4k all in one with 580)

 

[ethomaz]

SIGGRAPH is 30th of July.

Holy shit. 5 month head start for the 1080 Ti. This thing better be sex on a PCB.

Well probably 5 months before Volta...

 

[dgrdsv]

Well probably 5 months before Volta...

Probably a lot less at this point. Launching early August, real availability in September and first Volta cards (GV104 based) may easily launch around Sep-Nov. So we're probably looking at ~2-3 months between Vega and Volta launches which means that whatever Vega was supposed to compete with will probably be obsolete very fast.