Vega FE benchmarked by random dude

[Valentyn]

4 minutes ago, Its Not Important said:

Fair enough.... i just thought they might be smarter and thinking about 1080 ti too.

Gotta remember AMD is broke as hell, and billions in debt.

 

It's clear most of their R&D went to Zen; and Raja has stated several times the old management thought discrete GPUs were dead by 2015. They thought consoles and APUs were the future.

Took putting Lisa in charge, and head hunting back Raja to start changing things.

 

Vega just needs to be good price to performance, and that's all most would care about. The few % that only want the best would have gotten Titans by now.

[Sacriledgy]

1 minute ago, Valentyn said:

Gotta remember AMD is broke as hell, and billions in debt.

 

It's clear most of their R&D went to Zen; and Raja has stated several times the old management thought discrete GPUs were dead by 2015. They thought consoles and APUs were the future.

Took putting Lisa in charge, and head hunting back Raja to start changing things.

Ahhh if only AMD has more money. From Ryzen its clear that they have the capability to make a great product, only lacking the resource (money). 

 

I really hope Vega is at least around 1080 though, and released on time.

[MageTank]

9 minutes ago, ravenshrike said:

http://www.3dmark.com/hall-of-fame-2/timespy+3dmark+score+performance+preset/version+1.0/1+gpu

 

The $5-7000 dollar workstation card doesn't beat out the full up Pascal gaming cards. That's on mature drivers with a proper PSU for the card. This is a card with new drivers, and possibly underpowered.

We have no clue how old the PSU, whether the 550 rating is nominal or max power draw is or if he's managed to overheat the thing during its lifetime. As such, we have no clue whether it's underpowered or not.

It's a 550w corsair PSU with what looks like flat modular cables (unsleeved). This means it can only be 1 of 5 PSU's. The CXM550W (2015), the TXM550W (2017), the CSM550W (2013), RMx550W (2016), and RM550W (2013). Of these 5 PSU's, only one of them are bronze, with the rest being gold. So at worst, the PSU is either 4 years old and gold rated, or 2 years old and bronze rated. The other options make it a newer, gold rated model.

 

Aging caps aside, if the PSU is following ATX spec, the card should only be pulling at most, 375w (+/- 10w) in it's default configuration. This would be at a default 100% power limit. He is using a 4790k on an ITX motherboard, in an ITX case with an AIO. There is no telling if it's delidded or not, but it's safe to assume that 4790k isn't highly OC'd, as Haswell (with some thanks to FIVR) runs extremely hot when overclocked, even on AIO's. Assuming say, a max of 4.5ghz under those specs, it's safe to assume that CPU isn't pulling more than 100w under load. That's 475w. Still leaves 75w left for the rest of the system. Given the aforementioned ITX board (and the nature of ITX's limited DIMM slots) he's only running 2 DIMMS, a few fans, and an SSD (i see only one sata cable in the case, and I can't tell if that board has an M.2 slot on the underside or not, looks like the ASUS Z97 impact which only had a U.2 port). I understand this is a ton of assumptions to make, but this is me actually giving him the benefit of the doubt, using an overclocked CPU on an ITX board in an ITX case, and actually assuming the card itself is drawing every last drop of power that it can (75w more than it's actual rated TDP). If we assume a 10-20% cap age, then yes, he will be over spec, but these PSU's are also rated to handle peak loads that are beyond their continuous rating as well.

 

It's impossible to say whether or not the PSU is a contributing factor given the many variables, but it's highly unlikely. I would expect a system crash/hang, long before it throttles a random component.

[pas008]

7 minutes ago, ravenshrike said:

Yes, it performs decently. But that's because the card is such an enormous powerhouse in the first place. Imagine two Porche 911s, one optimized for the Nurburgring Nordschliefe vs one optimized for the Brickyard 400. While both cars would be able to easily complete their opposite's tracks in good time, on the track they are tuned for they would be significantly faster than on the one they aren't.

that wasnt the point, plz move on and quit trying to make it something it wasnt

[Sacriledgy]

7 minutes ago, pas008 said:

that wasnt the point, plz move on and quit trying to make it something it wasnt

What do you mean by that? That Vega is a fail GPU way before release? Or you mean you know exactly Vega performance that you are that confident with your statement? 

 

Please stop bandwagoning. Its just a speculation, nobody said its a review. This "Review" is just some random guy with janky setup. Whether you like it or not, this is not the perfect benchmark condition. I'd say relax a little bit and wait for real review.

[ravenshrike]

7 minutes ago, pas008 said:

that wasnt the point, plz move on and quit trying to make it something it wasnt

Actually, that was the exact argument from everyone except for the strawman that Zmeul built. Nobody actually said that the Quadro couldn't game, just that it was woefully unoptimized for such an application. A 20-30% performance hit vs the same platform when optimized for gaming is a pretty damn significant performance hit.

[pas008]

Just now, ravenshrike said:

Actually, that was the exact argument from everyone except for the strawman that Zmeul built. Nobody actually said that the Quadro couldn't game, just that it was woefully unoptimized for such an application. A 20-30% performance hit vs the same platform when optimized for gaming is a pretty damn significant performance hit.

actually that link was in response to someone saying it couldnt

maybe you should take an understanding that there were many conversations going on here before the thread was cleaned

[Valentyn]

15 minutes ago, Its Not Important said:

Ahhh if only AMD has more money. From Ryzen its clear that they have the capability to make a great product, only lacking the resource (money). 

 

I really hope Vega is at least around 1080 though, and released on time.

Well with Zen across the market they might finally break even and make a profit after 10 years.

 

Then pump money into R&D

[Sacriledgy]

3 minutes ago, Valentyn said:

Well with Zen across the market they might finally break even and make a profit after 10 years.

 

Then pump money into R&D

I really hope Navi will be a good true Nvidia contender. Because honestly, I'd rather pick AMD than Nvidia. They tend to have a better life than Nvidia cards (i wouldn't say anything more about that) Just look at R9 390X vs GTX 970.

[Valentyn]

6 minutes ago, Its Not Important said:

I really hope Navi will be a good true Nvidia contender. Because honestly, I'd rather pick AMD than Nvidia. They tend to have a better life than Nvidia cards (i wouldn't say anything more about that) Just look at R9 390X vs GTX 970.

 

One can only hope they manage it. 

[ARikozuM]

Spoiler

I want to be excited, but after reading the thread (haven't seen so many cleanups in a thread since an AMD v Intel thread went two days without moderation) I'm not too excited. 

 

The variance in the scores seems to be all over the place regardless of what the recorded "core clock" is. Since this is a 375W TDP card with two 8-pins, I'm assuming that they (AMD) put the threshold of power dissipation as high as they could for safety rather than actual need. 

 

Also, why they chose to use their awful reference cooler is beyond my comprehension as we've seen time and time again that AMD's reference is lackluster at best. The reviewer did seem to use the blower in a closed chassis where it should do best vs an open-air test bench. Hopefully AMD allows AIB's to take their designs to this unlike the Fury line-up.  

[Masada02]

1 minute ago, ARikozuM said:

  Reveal hidden contents

Hopefully AMD allows AIB's to take their designs to this unlike the Fury line-up.  

Unlikely. This is a prosumer card, not a gaming card. Nvidia doesn't let AIB's do anything with the Titan cards either. 

 

I bet the gaming versions will have custom coolers and actual gaming drivers, because basically all we are doing now is judging an entire architecture's gaming performance on how the locked down workstation card performs. Seems like the right thing to do...

[Sacriledgy]

1 minute ago, Masada02 said:

Unlikely. This is a prosumer card, not a gaming card. Nvidia doesn't let AIB's do anything with the Titan cards either. 

 

I bet the gaming versions will have custom coolers and actual gaming drivers, because basically all we are doing now is judging an entire architecture's gaming performance on how the locked down workstation card performs. Seems like the right thing to do...

The Gaming card will have custom cooler. But yeah, i agree with you 

[ARikozuM]

1 minute ago, Masada02 said:

Unlikely. This is a prosumer card, not a gaming card. Nvidia doesn't let AIB's do anything with the Titan cards either. 

I'd like to agree with this, but no...

 

Fury X wasn't a prosumer card, but AMD didn't allow the AIB's to change the cooler instead opting for a water-cooler and an air-cooled reference design. Just because company A does something, doesn't mean that company B should follow suit when profits can be gained by straying from the norm.

[The Benjamins]

20 minutes ago, Its Not Important said:

I really hope Navi will be a good true Nvidia contender. Because honestly, I'd rather pick AMD than Nvidia. They tend to have a better life than Nvidia cards (i wouldn't say anything more about that) Just look at R9 390X vs GTX 970.

I think you mean 290X vs 780

[PCGuy_5960]

6 minutes ago, ARikozuM said:

So in other words, Vega at 1600-1682 is barely able to beat a (pretty much) stock 1080 while drawing almost 2x more power.  Not very impressive IMHO. 

[Zando_]

33 minutes ago, Its Not Important said:

What do you mean by that? That Vega is a fail GPU way before release? Or you mean you know exactly Vega performance that you are that confident with your statement? 

 

Please stop bandwagoning. Its just a speculation, nobody said its a review. This "Review" is just some random guy with janky setup. Whether you like it or not, this is not the perfect benchmark condition. I'd say relax a little bit and wait for real review.

Yes! Technically, proper GPU benches have overkill mobos, RAM, PSUs, CPUs, etc to rule out even the possibility of anything throttling the GPU. 

[ARikozuM]

Just now, PCGuy_5960 said:

So in other words, Vega at 1600-1682 is barely able to beat a (pretty much) stock 1080 while drawing almost 2x more power.  Not very impressive IMHO. 

Again, I don't know the specifics of the methodology. It seems like the case is an ITX cuboid (I don't know if it has a cover or not), but it should have helped the blower achieve better than a fanned design. 

 

Also, using 2 8-pins would provide the GPU with 375W (same setup on the 1080 Ti as far as I know), but the 1080 Ti only draws about 280W according to some of the benchmarks I've read when thinking of purchasing upon release. I doubt this card actually draws the full 375W of power while only maintaining GTX 1080-level performance. I think there may be an issue that wasn't tested for. Whether it was the PSU as some members pointed out, the case (open vs closed), or software, we won't know anytime soon. 

[ravenshrike]

30 minutes ago, pas008 said:

actually that link was in response to someone saying it couldnt

maybe you should take an understanding that there were many conversations going on here before the thread was cleaned

Yet the thread cleaning didn't eliminate the quotes used in the posts that remained. What was said is that the quadro cards don't game well. Which, compared to what the chip in any given quadro card is actually capable of when tuned correctly, they don't. It was never said that they can't game at all. Two fundamentally different statements.

[PCGuy_5960]

Just now, ARikozuM said:

Also, using 2 8-pins would provide the GPU with 375W (same setup on the 1080 Ti as far as I know), but the 1080 Ti only draws about 280W according to some of the benchmarks I've read when thinking of purchasing upon release. I doubt this card actually draws the full 375W of power while only maintaining GTX 1080-level performance. I think there may be an issue that wasn't tested for

Well, the 1080 Ti has a 250W TDP, not a 300+W TDP.... And I doubt that it was thermal throttling, I mean, it does look like the card was running at 1500-1600MHz (which is the base clock of the Vega FE card AFAIK)

[ARikozuM]

1 minute ago, PCGuy_5960 said:

Well, the 1080 Ti has a 250W TDP, not a 300+W TDP.... And I doubt that it was thermal throttling, I mean, it does look like the card was running at 1500-1600MHz (which is the base clock of the Vega FE card AFAIK)

I know, I saw some comments using the "two 8-pins" as an indicator that it draws a LOT of power when it could be that AMD only placed the TDP that high as a precaution. 

[Masada02]

8 minutes ago, ARikozuM said:

I'd like to agree with this, but no...

 

Fury X wasn't a prosumer card, but AMD didn't allow the AIB's to change the cooler instead opting for a water-cooler and an air-cooled reference design. Just because company A does something, doesn't mean that company B should follow suit when profits can be gained by straying from the norm.

I used the Titan as an example case where Nvidia has done the same thing, I never said AMD was following suit just because their competitor does it. And just because a Fury X is NOT a prosumer card does not mean they can't lock the cooler design down. You're relating things that in reality are not.

 

What I'm saying is that the Pro cards and "prosumer" cards have historically always been locked to their reference cooler, and because of this relationship it is highly unlikely that the Frontier Edition of Vega will get a custom cooler from an AIB. I didn't say whether or not this is the correct thing to do, it's just statistics. 

 

The gaming variants are a completely different story and we have no idea how they will be handled. I'm hoping that all of them will have AIB designs.

[PCGuy_5960]

Just now, ARikozuM said:

it could be that AMD only placed the TDP that high as a precaution. 

Nah, why would they do that? It basically makes their card look bad...

[raphidy]

55 minutes ago, Valentyn said:

It's clear most of their R&D went to Zen; and Raja has stated several times the old management thought discrete GPUs were dead by 2015. They thought consoles and APUs were the future.

It's weird, why ryzen exists without apu then? I thought that if that's the direction to go, then ryzen should have come with apu only no?

[PCGuy_5960]

Just now, raphidy said:

why ryzen exists without apu then?

APUs are Coming soon™