Vega FE - Placebo Mode

[HKZeroFive]

Wait for Navi.

[LooneyJuice]

I have to give mad props to them for even going through with this testing and not blurting out a 30 second soundbite along the lines of "Gaming mode: It don't work". What I find almost infuriating is the fact that the comment section, and general responses to the video are inundated with "maybe it's not ready yet, the driver needs optimization, something is probably broken for now" in this insane effort to somehow soften the blow. 

 

1. Why? Why soften the blow? Have you purchased AMD Stock? Is anyone getting any kind of percentage of Vega sales besides AMD?

2. How do you even know it's not optimized yet, or it is indeed pending optimization?

3. Even if it is broken, how is it in any way any kind of justification since a broken product (software or hardware) is going to reach customers in a broken state?

4. Steve Burke specifically said, and I quote "It was marketed as a major feature alongside Vega: FE." in the comments. Ergo, if a major feature isn't working as advertised, it's a massive disservice.

 

Kudos to GN for even going through with this, but damn do I understand if it felt utterly pointless to go through that testing.

[The Sloth]

14 minutes ago, LooneyJuice said:

1. Why? Why soften the blow? Have you purchased AMD Stock? Is anyone getting any kind of percentage of Vega sales besides AMD?

 

it happens to everything these days from CPUs, Games, and GPUs, accept it and move on. 

 

14 minutes ago, LooneyJuice said:

2. Even if it is broken, how is it in any way any kind of justification since a broken product (software or hardware) is going to reach customers in a broken state.

 

rush to compete, since VEGA has been delayed 

 

[LooneyJuice]

2 minutes ago, nerdslayer1 said:

it happens to everything these days from CPUs, Games, and GPUs, accept it and move on. 

 

rush to compete, since VEGA has been delayed 

 

End-users shouldn't be concerned with the latter. It is anyone's prerogative to personally want to consider what's going on under the hood, but that doesn't mean end-users should actually be concerned, and not judge the product as it came out.

 

Regarding the former, just because earthquakes happen doesn't mean you shouldn't build housing that doesn't collapse. Convoluted way of saying that it's an issue that needs to be acknowledged anyway, even if it resembles banging your head against a concrete wall. There's a lot of work going into reviewing products from lots of decent outlets, which doesn't deserve being undermined by arbitrary comments. If on the other hand, there was an official word from AMD (or anyone else) in the comments, or in the video regarding something clearly broken pending revision imminently, that would be a completely different story.

[Taf the Ghost]

2 hours ago, Grinners said:

It is mind boggling that a company as big as AMD is so bad at communicating/marketing/customer-relations. 

 

So much of this bad press could be almost entirely avoided by just being upfront and honest. Something like "the gaming drivers are still being produced so at the moment it is only really UI changes. More to come shortly!"

 

They'd be paying multiple PR/Marketing people hundreds of thousands in combined salaries and they can't even do this???

If RX Vega comes out between the 1080 and 1080 Ti for $399 USD, all of this discussion will be forgotten. The Bad Press is almost wholly in the Tech sphere.

1 hour ago, rattacko123 said:

AMD needed to keep up with the Q2 release date, they had already delayed Vega enough.

Vega appears to have only ever been a Q2'2017 release date for a long while.  2 years after Fury. However, since we never got a "Big Polaris", the space was just empty for a year from AMD. However, it does feel a lot like RX Vega got shoved from Late April to Late July/Early August, so around 3.5-4 months delay. Which doesn't seem too bad given the problems with HBM2.

[Humbug]

LOL, terrible, terrible launch. Poor communication, no cards given to reviewers. And still no indication from AMD about what performance level to expect on the gaming front. Poor launch. Brought on by the fact that they just wanted to stick to their 1st half 2017 Vega launch promise.

 

having said all that you can be pretty sure that Rx Vega and even FE Vega will perform much better than this in gaming.

Why?

because right now clock of clock Vega is as slow as Fury X. Gamers Nexus did a clock for clock comparison. In games the Vega is for now showing zero architectural improvements. Which indicates obviously something isn't working properly in the drivers.

 

The alternative explanation is that Vega is just a terrible architecture and that none of the theoretical architectural improvements worked in real world workloads (like bulldozer CPUs).

[Jito463]

10 hours ago, LAwLz said:

What would not make sense, is if AMD developed two sets of drivers. One which performed well in workstation type tasks, and one that performed well for gaming. I mean... GPUs and drivers doesn't work like that.

Actually, that's exactly how they work.  It's the reason why - on the professional side - there are optimized and certified drivers for different applications.  AutoCAD?  There's a specific driver for that.  Solidworks?  Same thing.  While the Vega FE drivers lack certifications, they can certainly be optimized for professional tasks.  It's the same reason why newer drivers can improve (or sometimes hamper) gaming performance over time.  It all comes down to optimization.

[Jito463]

What I'd like to see them do, is run professional tasks (like video rendering/editing, or 3D rendering) in both Pro mode and Game mode, to see if that makes any difference at all.  We know gaming vs pro mode does nothing for games at the moment, but does it change anything for professional tasks?

[raphidy]

8 hours ago, iamdarkyoshi said:

If they're as stable as my 1080's are... no thx

I don't know, i owned a 1070 and it's been good since last August. The only issue I had was the one with FAH not folding.

[Taf the Ghost]

54 minutes ago, Jito463 said:

What I'd like to see them do, is run professional tasks (like video rendering/editing, or 3D rendering) in both Pro mode and Game mode, to see if that makes any difference at all.  We know gaming vs pro mode does nothing for games at the moment, but does it change anything for professional tasks?

Considering GN can't dial in an OC in general, I don't think they'd see a difference, as something seems off with their card.  Or the way they have the drivers operating.  Or Game vs Pro mode only works on Ryzen.

[Prysin]

5 hours ago, HKZeroFive said:

Wait for Navi.

pff, AMD is going to be one generation behind Nvidia for the next 2-3 generations, at the very least, unless Nvidia does  major cock-up

[RagnarokDel]

12 hours ago, tom_w141 said:

Has any Titan X sucked at launch?

Titans are never the first GPU of a new architecture.

[Senzelian]

9 hours ago, djdwosk97 said:

Remind me again how the 970 is literally the fourth most common (recent) GPU on Steam (only behind the 1060, 750Ti, and 960). 

 

A few people on tech forums cared about the .5gb of slower VRAM, but 99% of people didn't give a shit since performance was exactly what all reviews showed. 

That was not my point at all, but thanks for useless information.

[HKZeroFive]

53 minutes ago, Prysin said:

pff, AMD is going to be one generation behind Nvidia for the next 2-3 generations, at the very least, unless Nvidia does  major cock-up

It was more tongue-in-cheek, seeing how everyone has been telling me to "wait for Vega" for more than a year now and suddenly it's now "wait for drivers" due to Vega FE's underwhelming performance.

[Jito463]

1 hour ago, huilun02 said:

This is the only version of Vega and drivers that will be released. Horrible perf/price. Vega total failure confirmed. 

I sincerely hope you simply forgot the /s.

52 minutes ago, Prysin said:

pff, AMD is going to be one generation behind Nvidia for the next 2-3 generations, at the very least, unless Nvidia does  major cock-up

Vega was the last architecture launched under the previous management.  Navi will be based on a similar (modular) approach as Ryzen.  If they do it right, they could catch up to Nvidia quite rapidly.  Only time will tell, though.

[79wjd]

1 hour ago, Senzelian said:

That was not my point at all, but thanks for useless information.

Your point was pretty obvious in that it was a really bad thing because it fucked nvidia. But most people didn't give a shit. 

15 hours ago, Senzelian said:

I don't think the answer is that simple and I don't think AMD would be stupid enough trying to trick the consumers this way.
We all know how the 3.5GB GTX 970 fiasco went.

 

[Misanthrope]

5 hours ago, Humbug said:

LOL, terrible, terrible launch. Poor communication, no cards given to reviewers. And still no indication from AMD about what performance level to expect on the gaming front. Poor launch. Brought on by the fact that they just wanted to stick to their 1st half 2017 Vega launch promise.

 

having said all that you can be pretty sure that Rx Vega and even FE Vega will perform much better than this in gaming.

Why?

because right now clock of clock Vega is as slow as Fury X. Gamers Nexus did a clock for clock comparison. In games the Vega is for now showing zero architectural improvements. Which indicates obviously something isn't working properly in the drivers.

 

The alternative explanation is that Vega is just a terrible architecture and that none of the theoretical architectural improvements worked in real world workloads (like bulldozer CPUs).

Again: "Is not possible AMD would fuck up this bad, let me close by stating the exact scenario when they did fuck up this bad in the past: Bulldozer"

[AlwaysFSX]

12 hours ago, raphidy said:

He's a Nvidia shill, he likes to have ready driver on day one.

Because expecting a product to work when it's available to the public makes you a shill, lols.

[cj09beira]

8 hours ago, LooneyJuice said:

I have to give mad props to them for even going through with this testing and not blurting out a 30 second soundbite along the lines of "Gaming mode: It don't work". What I find almost infuriating is the fact that the comment section, and general responses to the video are inundated with "maybe it's not ready yet, the driver needs optimization, something is probably broken for now" in this insane effort to somehow soften the blow. 

 

1. Why? Why soften the blow? Have you purchased AMD Stock? Is anyone getting any kind of percentage of Vega sales besides AMD?

2. How do you even know it's not optimized yet, or it is indeed pending optimization?

3. Even if it is broken, how is it in any way any kind of justification since a broken product (software or hardware) is going to reach customers in a broken state?

4. Steve Burke specifically said, and I quote "It was marketed as a major feature alongside Vega: FE." in the comments. Ergo, if a major feature isn't working as advertised, it's a massive disservice.

 

Kudos to GN for even going through with this, but damn do I understand if it felt utterly pointless to go through that testing.

1-we aren't the normal people, we are tech fans and love to know as much as possible about new interesting things Vega is the newest interesting thing, the fact that some important new features in Vega aren't enabled is something people should take into account before preaching "Vega is dead"

2-because the tests done on games mostly show that most gaming features aren't enabled, example the extra geometry output per clock, the new rasterizer isn't enabled also

3-they have stated multiple times that this isn't a gaming card, and almost exclusively marketed it as a card for devs and creators 

look at the pro benchs, there vega does great because that part of the driver has been optimized for.

10 minutes ago, Misanthrope said:

Again: "Is not possible AMD would fuck up this bad, let me close by stating the exact scenario when they did fuck up this bad in the past: Bulldozer"

right now we have huge ipc gains in pro apps, and exactly no ipc gains on the  gaming side, that clearly shows that it isnt a architecture problem 

[Senzelian]

27 minutes ago, djdwosk97 said:

Your point was pretty obvious in that it was a really bad thing because it fucked nvidia. But most people didn't give a shit. 

Nope
I even explained it in another post and still you're trying so hard to prove me wrong for something I never said. 

[Misanthrope]

26 minutes ago, cj09beira said:

right now we have huge ipc gains in pro apps, and exactly no ipc gains on the  gaming side, that clearly shows that it isnt a architecture problem 

Well yes and no: AMD has had the raw power lead in workstation apps for a while now, several years iirc. Their cards when it comes to horsepower are just more powerful than Nvidia offerings. They still almost always get tied or loose in gaming because Nvidia is that much better at optimizing and addressing gaming developers directly (to the point of basically "cheating" with some titles) This is further confirmed by the DX12 optimized title (because it boils down to really just one: AOTS) where they take a comfortable lead too cause the software is build around taking advantage of their resources more efficiently and more resources they do have.

 

So yes it's probably not an architecture problem, but you can't dismiss driver and game optimization so easily when it weights on performance so heavily. Heavy enough that it keeps Nvidia competitive with less powerful hardware at times and keeps AMD as a lukewarm option because of their lack of optimization. Hardware architecture is always just half of the equation, software is just as important.

 

[cj09beira]

Just now, Misanthrope said:

Well yes and no: AMD has had the raw power lead in workstation apps for a while now, several years iirc. Their cards when it comes to horsepower are just more powerful than Nvidia offerings. They still almost always get tied or loose in gaming because Nvidia is that much better at optimizing and addressing developers directly (to the point of basically "cheating" with some titles) This is further confirmed by the DX12 optimized title (because it boils down to really just one: AOTS) where they take a comfortable lead too cause the software is build around taking advantage of their resources more efficiently and more resources they do have.

 

So yes it's probably not an architecture problem, but you can't dismiss driver and game optimization so easily when it weights on performance so heavily. Heavy enough that it keeps Nvidia competitive with less powerful hardware at times and keeps AMD as a lukewarm option because of their lack of optimization. Hardware architecture is always just half of the equation, software is just as important.

 

those features not being enabled is a driver issue, optimization will also help a lot,

i didn't focus on the driver side because its easier to prove that a feature isnt enabled than saying it could do better with more optimized drivers

[LooneyJuice]

22 minutes ago, cj09beira said:

1-we aren't the normal people, we are tech fans and love to know as much as possible about new interesting things Vega is the newest interesting thing, the fact that some important new features in Vega aren't enabled is something people should take into account before preaching "Vega is dead"

2-because the tests done on games mostly show that most gaming features aren't enabled, example the extra geometry output per clock, the new rasterizer isn't enabled also

3-they have stated multiple times that this isn't a gaming card, and almost exclusively marketed it as a card for devs and creators 

look at the pro benchs, there vega does great because that part of the driver has been optimized for.

right now we have huge ipc gains in pro apps, and exactly no ipc gains on the  gaming side, that clearly shows that it isnt a architecture problem 

Well, I feel I have to say I never mentioned anything about Vega being dead. I for one would kill for some competition. I want to see it succeed just by virtue of the fact that the market expands as a result. My problem is a lot of the hearsay when it comes to what's supposedly missing. All that again alludes to issues end-users shouldn't be expected to be concerned about when it comes to the performance of the delivered product.

 

For example, your assertions about geometry output per clock, rasterizer etc. Do you have any legitimate information regarding something that's internally being tested, has been broken or is currently being worked on? This isn't snark, If you do, I'm totally cool with that, and I would love the information. Besides that, I cannot take anyone's word simply because they said so. I think it's only fair to ask for a legitimate source or inside scoop.

 

Additionally, I really don't know where the "not a gaming card" argument comes in again, seeing as I never inferred that it sucked for gaming because it was marketed as a gaming card. Rather, my issue solely lies with the fact that the gaming mode offered none of the advertised improvements according to GN. Again, the quote was "It was marketed as a major feature alongside Vega: FE."  This is of course a prosumer product, and that's totally fine. The problem is the false advertisement, at least in this stage.

[XenosTech]

4 minutes ago, Misanthrope said:

Well yes and no: AMD has had the raw power lead in workstation apps for a while now, several years iirc. Their cards when it comes to horsepower are just more powerful than Nvidia offerings. They still almost always get tied or loose in gaming because Nvidia is that much better at optimizing and addressing developers directly (to the point of basically "cheating" with some titles) This is further confirmed by the DX12 optimized title (because it boils down to really just one: AOTS) where they take a comfortable lead too cause the software is build around taking advantage of their resources more efficiently and more resources they do have.

 

So yes it's probably not an architecture problem, but you can't dismiss driver and game optimization so easily when it weights on performance so heavily. Heavy enough that it keeps Nvidia competitive with less powerful hardware at times and keeps AMD as a lukewarm option because of their lack of optimization. Hardware architecture is always just half of the equation, software is just as important.

 

From what I've noticed I'd say hardware vs software optimizations is a 30-70 relationship. Like it's way easier and cheaper to optimize software vs rebuilding hardware from the ground up and launching a new product.

[cj09beira]

7 minutes ago, LooneyJuice said:

Well, I feel I have to say I never mentioned anything about Vega being dead. I for one would kill for some competition. I want to see it succeed just by virtue of the fact that the market expands as a result. My problem is a lot of the hearsay when it comes to what's supposedly missing. All that again alludes to issues end-users shouldn't be expected to be concerned about when it comes to the performance of the delivered product.

 

For example, your assertions about geometry output per clock, rasterizer etc. Do you have any legitimate information regarding something that's internally being tested, has been broken or is currently being worked on? This isn't snark, If you do, I'm totally cool with that, and I would love the information. Besides that, I cannot take anyone's word simply because they said so. I think it's only fair to ask for a legitimate source or inside scoop.

 

Additionally, I really don't know where the "not a gaming card" argument comes in again, seeing as I never inferred that it sucked for gaming because it was marketed as a gaming card. Rather, my issue solely lies with the fact that the gaming mode offered none of the advertised improvements according to GN. Again, the quote was "It was marketed as a major feature alongside Vega: FE."  This is of course a prosumer product, and that's totally fine. The problem is the false advertisement, at least in this stage.

i didn't meant to say you did it, my point was that people shouldn't.

amd has released the most important Vega features a long time ago, in them there is 2.75 times the geometry output per clock (probably a mix of double output and using other parts o f the gpu to help(0.75 part)) and the tiled based rasterizer, the rasterizer is confirmed not enabled by pcper, and the geometry output would be noticeable on the first firestrike graphics test, as that test is mostly geometry incentive.

amd didn't push the gaming side of Vega at all, they mostly focused on the pro side, probably because they don't have the man power to do both drivers at the same time, now they should have been upfront about it and said they would improve it latter, but this is amd we are talking about marketing isn't their forte.