Editorial Vega Shows Up Beating a GTX 1080 in CompuBench, But Hold the Hypetrain

[R-T-B]

The Vega based line of AMD GPUs are definitely a big unknown at this point, so any sightings or benchmarks of it are highly sought after by the rumormill. Well, here is another one to add to your pile of rumor-material folks: AMD has posted a card benchmark to Compubench that bests even the GTX 1080.

Why hold the hype?

There are two obvious issues. One, this is a compute only benchmark, and has little relevance to the average gamer. Two, in the same benchmark, a 980TI also beats the 1080. Stranger yet, the 1080 is also beaten by its little brother, the 1070. Take this one with a grain of salt, for the obvious reasons. It won't stop the the hypetrain from using this info to its own end, but maybe you can avoid being smashed by it by using some critical thinking.




PS: If you don't like me telling you how to think, please note this was properly marked as an editorial. You may board the train of your choosing at the next exit. Good luck on reaching your destination.

View at TechPowerUp Main Site

 

[Dippyskoodlez]

aka don't trust amd. If people haven't learned 3 launches later, they will still run this hype train into another brick wall this time..

 

[ensabrenoir]

......cue the count down before someone uses this bench mark to start a Vega smashes 1080ti thread.......starting in 5-4-3.....??? too late...

 

[Prima.Vera]

Wat hype? The card is smashed to pieces by Titan and 1080Ti in this "benchmark":

 

[Darksword]

 

[OneCool]

I mean really lol. It's getting to be funny now.
I was sitting at a bench at the train station for Ryzen and never got off of it. Just watched it go by. Vega? I already know it's going to be mid-high end!

1080ti = beast..Don't see Vega coming close
Intel worried about Ryzen = them blowing 15 billion on a new company

 

[R-T-B]

Wat hype? The card is smashed to pieces by Titan and 1080Ti in this "benchmark":

and 980ti, hence me calling it. I almost didn't label this editorial, because really, you'd have to be a blathering idiot to take this one seriously. Nevertheless, I've seen some people handing it around as fact.

 

[FordGT90Concept]

https://compubench.com/compare.jsp?benchmark=compu20d&did1=45270653&os1=Windows&api1=cl&hwtype1=dGPU&hwname1=AMD+687F:C1&D2=NVIDIA+GeForce+GTX+1080

Vega and GTX 1080 seem to be pretty well matched. Only idea I have why Titan and 1080 Ti are faster is memory related (8 GiB versus 11-12 GiB). Vega presumably has 8 or 16 GiB. If the former, it all makes sense. If the latter...

 

[R-T-B]

https://compubench.com/compare.jsp?benchmark=compu20d&did1=45270653&os1=Windows&api1=cl&hwtype1=dGPU&hwname1=AMD+687F:C1&D2=NVIDIA+GeForce+GTX+1080

Vega and GTX 1080 seem to be pretty well matched. Only idea I have why Titan and 1080 Ti are faster is memory related (8 GiB versus 11-12 GiB). Vega presumably has 8 or 16 GiB. If the former, it all makes sense. If the latter...

What about the 1070 being faster than a 1080? That confuses me to no end.

 

[KainXS]

The device name on this bench GFX900, so is this this cut the cut Vega or full Vega?

http://llvm.org/viewvc/llvm-project/cfe/trunk/lib/Basic/Targets.cpp?revision=297659&view=markup

could be the cut Vega where GFX901 is the full Vega, I still am thinking its only going to be a bit faster than a 1070 though.

 

[drade]

Wat hype? The card is smashed to pieces by Titan and 1080Ti in this "benchmark":

When was the benchmark? Who provided this information, Nvidia? Also what was the complete build setup used with the card? I7, I5? Ram? Many variables to consider.

 

[Captain_Tom]

Yeah but they also post a T Rex graphics and ocean waves GRAPHICS benchmark if you open your god damn eyes.


Don't get me wrong, this leak is far from gospel; but it is a thing. The Vega 10 card underclocked by 30% beat the 1080 and came within 10 - 40% of the 1080 Ti.

 

[TheGuruStud]

Yeah but they also post a T Rex graphics and ocean waves GRAPHICS benchmark if you open your god damn eyes.


Don't get me wrong, this leak is far from gospel; but it is a thing. The Vega 10 card underclocked by 30% beat the 1080 and came within 10 - 40% of the 1080 Ti.

Hey, don't do simple math (especially like Fury X vs Vega Tflop increase excluding uarch improvements).
Nzealots will shut down into full denial mode when they realize Vega is as fast as the new Milk Ti (or faster).

 

[Captain_Tom]

Hey, don't do simple math (especially like Fury X vs Vega Tflop increase excluding uarch improvements).
Nzealots will shut down into full denial mode when they realize Vega is as fast as the new Milk Ti (or faster).

I am not saying it will be, only that it might be; and that anyone who denies this possibility is silly. AMD has done it before numerous times.

 

[bistrocrat]

I never care about absolute values (like who beats who by how much) - only thing I care is price/performance... and if this Vega in absolute is smashed by Ti, is little worse than 1080, and costs below 1070 - then I will hype the shyte out of my self.

so far - AMD has not "crushed my dreams", "lied to the kids, AGAIN", "the hypetrain has railed off" or whatever metaphors internet tries to use these days - AMD did deliver a CPU in price of 329$, that is on par with 1000$ intel - in CPU relevant tasks (except if you care about 1080 @144+ gaming - then the 1000$ CPU is not the smartest choice anyway)

 

[FordGT90Concept]

What about the 1070 being faster than a 1080? That confuses me to no end.

Look at the numbers. It is very, very close. Margin of error close.

1070 and 1080 both have 8 GiB which kind of confirms the theory about memory being the key factor.

The 980 Ti is the anomaly at only 6 GiB of RAM.

 

[ShurikN]

Look at the numbers. It is very, very close. Margin of error close.

1070 and 1080 both have 8 GiB which kind of confirms the theory about memory being the key factor.

The 980 Ti is the anomaly at only 6 GiB of RAM.

That entire benchmark is an anomaly. 1080 destroys a 980ti in gaming yet falls behind it in this particular "test", or whatever it is. Don't see how cuda and opencl are relevant for gaming.

 

[R-T-B]

so far - AMD has not "crushed my dreams", "lied to the kids, AGAIN", "the hypetrain has railed off" or whatever metaphors internet tries to use these days - AMD did deliver a CPU in price of 329$, that is on par with 1000$ intel - in CPU relevant tasks (except if you care about 1080 @144+ gaming - then the 1000$ CPU is not the smartest choice anyway)

I don't think I ever claimed it did anything but deliver a good product. They have. But the hype generated was unrealistic. For those of us who kept to realistic expectations, it is remarkable progress for them.

This benchmark is completely off the wall though, and when someone passed it to me to try to show superiority of Vega, I almost had to make an article to counter it. It's pretty amazingly bad to use this as "proof" of anything in the gaming realm.

 

[RejZoR]

Unrealistic? A 600€ CPU beating a 1100€ one is an "unrealistic" hype? I must be dreaming then...

 

[FordGT90Concept]

That entire benchmark is an anomaly. 1080 destroys a 980ti in gaming yet falls behind it in this particular "test", or whatever it is. Don't see how cuda and opencl are relevant for gaming.

They aren't. This test could be using double precision floats for all we know and 980 Ti may have a higher double precision rate than the rest.

Edit: Both cards are 1/32 where Vega is 1/16.
http://www.anandtech.com/show/9306/the-nvidia-geforce-gtx-980-ti-review
http://techreport.com/news/30174/rumor-a-gp102-geforce-titan-and-gtx-1080-ti-are-in-the-works

I've tried and I have no explaination for 980 Ti being where it is.

 

[medi01]

Vega chip is vastly bigger than 1080, nearly 500mm^2 vs 314 in 1080. (Titan is 472mm^2)
It can NOT not wipe the floor with it.

aka don't trust amd. If people haven't learned 3 launches later, they will still run this hype train into another brick wall this time..

Hello hater.

Ryzen was HOPED to reach Haswell IPC, it actually reached Broadwell.

 

[R-T-B]

Unrealistic? A 600€ CPU beating a 1100€ one is an "unrealistic" hype? I must be dreaming then...

You know how hype trains work. You can't pretend there weren't people expecting more returning their CPUs on day 1. It IS a great CPU. It was NOT everything the hype promised, nor could it ever have been.

 

[Yorgos]

According to your 480 review, you'd rather steal and get some more money to buy a novidia card.
So, why would you suggest something opposing the site's long tradition of novidia loving?
I told you before, on some pr thing you usually have here as an article, that the last year you had almost ZERO inside info on anything. You have been reduced to posting pr about keyboards and fans.

I personally don't have the sources I used to have, due to being in a different workplace, but a little bird told me to expect big changes with vega.
My former colleagues already had details about zen 1 year ago, since we used to have intel amd dell and very often altera representatives there.
It's nice to see you guys going full damage contol with almost every amd product. Keep up the laughter.

 

[uuuaaaaaa]

There are various vega revisions/versions. The CompuBench 687F:C1 is the same that has been shown running Doom 2016 4K 60fps and it appears to perform a little bit faster than a gtx 1080. 687F:C1 appears to be running at 1200MHz and that would fall short of the 12.5TFlops reported for the Radeon MI25.

Recently 687F:C3 appeared on SiSoftware Sandra 2015 and it beats the the gtx1080 by 35% in this particular benchmark.

http://ranker.sisoftware.net/show_r...efdbe9dcebdaebcdbf82b294f194a999bfccf1c9&l=en

 

[ZoneDymo]

Im not sure whats more baffling, the results of this benchmark, or the reactions to this article...