AMD Vega (FE and RX) Benchmarks [Updated Aug 10 - RX Vega 64 Unboxing]

[Veradun]

Don't get it how an architecture a bit bigger than the 1080ti, more transistors, 1 year more development time and it can barely even compete with the GTX 1080 and consumes more power!

The point most people miss here is Vega is supposed to trade blows with GP100, GP102 and GP104. One single jack-of-all chip that has to battle three specialized chips from the competition. And if you take into account raw flops it also competes with V100.

In fact Vega FE competes with GP102 in PRO workloads. Competes with GP104 in gaming. Destroys GP100 and is supposed to be on par with V100 in 16b workloads.

Waiting for the RX it seems atm they focused somewhere else the thin budget they had.

If priced competitively I'm still buying an RX for gaming, since I have an ASUS MG279Q and I really want to retire my HD6870 (that I just switched in removing the HD6950). I'd like an AiO version, or a Nano (and I'm pretty sure I can make a Nano out of any AiB partner card with some downclocking and a lot of downvolting).

 

[nathanddrews]

I've added leaked RX FireStrike scores, specs, and images to the first page. I'll try to get any and all reviews condensed there as they roll in...

 

[cytg111]

if it trades blows with 1080 for sub 300 watts at 400$, i am getting one too

 

[_UP_]

Don't get it how an architecture a bit bigger than the 1080ti, more transistors, 1 year more development time and it can barely even compete with the GTX 1080 and consumes more power!

The answer, as many things in life, is money.
Nvidia spent $351M a quarter on R&D since 2014, according to this data. *
AMD spent $253M a quarter on R&D since 2014, according to this data. *
* Note: Last 13 quarters; dates are a bit different for both companies.

Now, even if we assume that AMD spends 100% of their budget on R&D only for GPUs, the difference is still $1.274B. Over the last 13 quarters. That's a lot of money. For a lot of engineers and other expenses.

But let's not forget that AMD actually splits its R&D costs on 2 markets, and in the other one, they are competing with Intel, a company that according to this spent an average of $3.056B A QUARTER for the same last 13 quarters.
I'm not even going to calculate the difference here, as it's quite staggering as it is.

AMD is in 2 markets, with significantly less R&D budget than either of its competitors, and has to split its R&D budget to research both of these fields.
And somehow manages to stay competitive, maybe not each generation and not in every market or market segment, but overall competitive.
We should also make an honourable mention to poor management before Dr. Su.

PS
How nice when companies are publicly traded and one can just Google these things

 

[leoneazzurro]

That, and the fact that Vega has a superior feature set compared to Pascal. Granted, it is probable that even without these enhancements RX Vega had ended consuming more than GTX1080 but anyway features are not cost free and they are definitely there.

 

[Rifter]

Vega FE already launched.
Vega RX will "launch" on July 31st, but, 99.999% chance of it being paper launch that day.

Thanks, as long as they send some out to reviewers i dont care if its a paper launch or not it always takes 3-6 months for any new GPU to be available in canada at decent prices anyways, likely wont buy till Christmas, just want to know if vega is in the running or if im going 1080/1080ti.

 

[Phynaz]

That, and the fact that Vega has a superior feature set compared to Pascal. Granted, it is probable that even without these enhancements RX Vega had ended consuming more than GTX1080 but anyway features are not cost free and they are definitely there.

Features just for features sake don't provide value.

 

[cytg111]

Features just for features sake don't provide value.

Posting just for postings sake dont provide value... I love features I might get the card on feature merits alone. Threads have been pretty objective and balanced so far lets not change that ok?

 

[n0x1ous]

Features just for features sake don't provide value.

you have a point. look at the much hyped tru-audio at the Hawaii launch. used in 1 or 2 games and basically useless now because it has no developer support

 

[96Firebird]

Posting just for postings sake dont provide value... I love features I might get the card on feature merits alone. Threads have been pretty objective and balanced so far lets not change that ok?

Which of these features do you really love, and why?

 

[Despoiler]

you have a point. look at the much hyped tru-audio at the Hawaii launch. used in 1 or 2 games and basically useless now because it has no developer support

It's used in consoles....

 

[nathanddrews]

It's used in consoles....

While true, it serves no purpose in the PC space, which is all that matters. #PCMR

 

[Elixer]

The answer, as many things in life, is money.
Nvidia spent $351M a quarter on R&D since 2014, according to this data. *
AMD spent $253M a quarter on R&D since 2014, according to this data. *
* Note: Last 13 quarters; dates are a bit different for both companies.
...

It isn't as much the dollar amount difference, it is how you spend the $$$ that counts.

There are lots of small startups that can compete with the big guys because they are focused on the task at hand.

AMD just needs the right people in the right area.

 

[n0x1ous]

It's used in consoles....

didnt know that....you would think it would get more use on the PC in ports then if its already being used on console

 

[_UP_]

It isn't as much the dollar amount difference, it is how you spend the $$$ that counts.

There are lots of small startups that can compete with the big guys because they are focused on the task at hand.

AMD just needs the right people in the right area.

That's true, to an extent.
Hardware is slower/more expensive, especially the more complex your design gets.
There are quite a few reasons that most players in the CPU/GPU are very much the big guys (Intel, Samsung, Qualcom, etc.), and many are very specialized (Intel, Nvidia, AMD, ARM, etc.).
It takes many resources - both to design and develop and fab these chips, the IP costs money (everyone has a deal with everyone since they all hold some of the patents) and to be able to see a project through.
There's a reason not too many new startups compete in these markets. It's very difficult.
Even the new players for dedicated chips for neural nets and machine learning are big guys like Google, IBM, etc.

Not to mention that having more money allows you to just brute force your way into a better design by sheer amount of engineer hours put at a task.

 

[Phynaz]

Posting just for postings sake dont provide value... I love features I might get the card on feature merits alone. Threads have been pretty objective and balanced so far lets not change that ok?

Your post is bording on trolling.

So you are happy that AMD bothered with the time and expense of Tru-audio 1, or Eyefinity 3D, or TruForm, etc. instead of spending that time, money and effort on increasing graphics throughput?

I think my graphics card should be a graphics card. If I want positional audio I can get a sound card for that.

 

[Phynaz]

It's used in consoles....

Do the consoles use Tru-Audio 1 or 2?

Edit: Looks like the PS4 has v1. I wonder how they are going to carry this forward as v2 is incompatible.

 

[Magic Hate Ball]

 

[cytg111]

Your post is bording on trolling.

So you are happy that AMD bothered with the time and expense of Tru-audio 1, or Eyefinity 3D, or TruForm, etc. instead of spending that time, money and effort on increasing graphics throughput?

I think my graphics card should be a graphics card. If I want positional audio I can get a sound card for that.

I think you got that the wrong way around but feel free to report me.

 

[cytg111]

From HardOCP :

"Totally unscientific and subjective testing is scorned by many, so if that gets your panties in a bunch, I suggest you stop reading."
- Ok, ill save my breath then .

 

[Paratus]

For those who didn't watch

Spoiler

• 1 Prefered the 1080Ti - Gsync system
• 6 Rated them equal
• 3 Preferred the RX Vega - Freesync system

Test was done at ultra wide 1440P 100hz monitors.

 

[guachi]

Don't get it how an architecture a bit bigger than the 1080ti, more transistors, 1 year more development time and it can barely even compete with the GTX 1080 and consumes more power!

nVidia probably spends more on R&D than RTG generates in revenue. That's your answer. How awful would nVidia products be if they had to shut down R&D in March and take 10 months off?

 

[Despoiler]

Do the consoles use Tru-Audio 1 or 2?

Edit: Looks like the PS4 has v1. I wonder how they are going to carry this forward as v2 is incompatible.

By v2 do you mean True Audio Next? Keep in mind that console chips are semi custom designs. They could decide to keep the DSPs, keep the DSPs for backwards compatibility and push for TAN to be used for new titles, or just go to TAN and port the audio via patches to their back catalog.

 

[Despoiler]

nVidia probably spends more on R&D than RTG generates in revenue.

Almost. JHH said they spent 3 billion developing Volta.

 

[Mopetar]

But let's not forget that AMD actually splits its R&D costs on 2 markets, and in the other one, they are competing with Intel, a company that according to this spent an average of $3.056B A QUARTER for the same last 13 quarters.
I'm not even going to calculate the difference here, as it's quite staggering as it is.

It's not like NVidia is just in the graphics market, which itself isn't quite as simple as just gaming cards, but never mind that. They've also spent a lot of money building SoCs over the years and probably dumped a lot into building their own custom ARM core (project Denver) instead of just licensing the stock ARM designs like they used to.