Vega Frontier Edition officially launched $999

[Valentyn]

OH MY OH MY! Officially launched at $999. Sadly the water cooled version isn't due until Q3 September.

 

Video shows how switching to Game Mode works from Pro mode.

 

Quote

SUNNYVALE, Calif., June 27, 2017 (GLOBE NEWSWIRE) — AMD (NASDAQ:AMD) today unleashed the first product based on its highly anticipated “Vega” graphics processing unit (GPU) architecture: Radeon™ Vega Frontier Edition. Radeon Vega Frontier Edition is the world’s first graphics card designed to empower the next generation of data scientists, game designers and visualization professionals, with up to 172 percent faster rendering performance than the comparable competitor card1. Through its disruptive High Bandwidth Cache Controller, the cornerstone of the world’s most advanced GPU memory architecture – HBM2 – Radeon Vega Frontier Edition expands the capacity of traditional GPU memory to 256TB, allowing users to tackle massive datasets with ease, and scored up to 33 percent faster than the competition in the DeepBench benchmark that measures the performance of basic operations involved in training deep neural networks2.

“We’re dedicating Radeon Vega Frontier Edition to all the visionaries and trailblazers who embrace new technologies to propel their industries forward to help solve mankind’s greatest problems,” said Ogi Brkic, senior director and general manager, Radeon Pro business, Radeon Technologies Group, AMD. “With this powerful solution, we’ve brought the full weight of our new ‘Vega’ GPU architecture to bear, offering unmatched3 performance in the most demanding design, rendering, and machine intelligence workloads so that the world’s top creators, data scientists and game developers can reach new frontiers in their fields.”

 

Quote

Radeon Vega Frontier Edition graphics cards are available from etailers in select regions today with an SEP of $999 USD for the air-cooled edition. The water-cooled edition is expected to launch in Q3 with an SEP of $1499.

 

 

 

 

 

 

For a prosumer card it actually looks fantastic value. It may not match or beat Radeon Pros, or Quadros; even cheaper ones; but it's ahead of the Titan Xp.

Also very important for Data Scientists, and people doing Machine Learning is that you can get 26 TFLOPs of FP16 compute.
The Instinct MI25 Vega card only has 24 TFLOPs of FP16; where as you need the Quadro GP100 or Tesla P100 to get double rate FP16 from NVIDIA. All which cost a lot more.

Even more interesting is that Frontier Edition has it's own unique drivers. It's not normal Radeon RX ReLive Drivers, of Radeon Pro. It's a true hybrid, and can apparently switch between Pro Mode, and Gaming mode.

It'll certainly be interesting to see what independent reviewers think, although for prosumers that don't game too much, it looks great value over the Titan Xp, and offers a lot of FP16 performance for only $999, compared to the GP100 offerings from NVIDIA.

 

Quote

The Radeon Vega Frontier Edition graphics card simplifies and accelerates game creation by providing a single GPU that is optimized for every stage of a game developer’s workflow. This includes everything from asset production to playtesting and performance optimization. 

With the Radeon Pro Settings user interface, users can seamlessly switch between “Radeon Pro Mode” and “Gaming Mode” to alternate between development on animation applications like Autodesk® Maya and performance optimizations with free, open source tools available through AMD’s GPUOpen initiative.

We'll have to wait and see for gaming performance with these new hybrid drivers though; although the VRMark results are between GTX 1070 and 1080. Hopefully a driver issue.

 

Below is a current bug associated with the driver switching

Quote

When switching between Pro Mode and Gaming Mode, Timeout Detection and Recovery (TDRs) have been observed with multiple display configurations

 

http://support.amd.com/en-us/download/frontier?os=Windows+10+-+64

 

It's also worth noting the drivers used at CES are the same ones currently available for the Frontier Edition. 17.20

 

So I would expect more improvements to come. Especially for the gaming side, as noted before there are issues switching between pro and gaming. This could impact performance of either. 

 

 

Also more pro performance numbers.

 

Frontier Edition

Catia = 135.78

Creo = 83.94

SolidWorks = 114.88

Siemens = 4.08

Cinebench = 183.28

Maya 2017 = 10.38

 

Titan Xp

Catia = 107.29 (~26.55% slower than FE)

Creo = 65.20 (~28.74% slower)

SolidWorks = 67.75 (~69.56% slower)

Siemens = 2.93 (~39.25% slower)

Cinebench = 169.72  (~7.99% slower

Maya 2017 = 3.81 (~172.44% slower)

 

 

VRMark Numbers from May 2017 are not very good at all!

Frontier Edition

VRMark Orange Room = 8157

 

GTX 1080 Reference

VRMark Orange Room = 9408

 

Radeon Duo Pro ( Polaris )

VRMark Orange Room =  6596

 

OcC3D net VRMark results

 

Source: https://videocardz.com/press-release/amd-launches-radeon-vega-frontier-edition-for-999

 

http://www.anandtech.com/show/11583/amds-radeon-vega-frontier-edition-formally-launches-air-cooled-for-999-liquid-cooled-for-1499

[tom_w141]

Well there goes titan's being used for workstation and maybe a price drop for quadro's

[Valentyn]

Just now, Tedny said:

Wait, FuryX faster than 1060? 

 

It's supposed to be competing with the 980Ti, but AMD does horribly in VR Mark it seems.

[Guest]

So blueeee

[Guest]

Just now, Tedny said:

Wait, FuryX faster than 1060? 

Even a non X does little faster

[Valentyn]

Just now, Tedny said:

rx 480, horribly? 

It's okay, at best. 

 

Surprising how bad Vega does for 13.1 TFLOPs and a new arch in VR Mark.

 

Will need to wait for independant reviews.

[WereCat]

8 minutes ago, Valentyn said:

It's okay, at best. 

 

Surprising how bad Vega does for 13.1 TFLOPs and a new arch in VR Mark.

 

Will need to wait for independant reviews.

I suppose the gaming Vega cards will do better in benchmarks like VR Mark. If not, they should be quite a bit cheaper at least.

[DnFx91]

13 minutes ago, deXxterlab97 said:

So blueeee

am i the only one that hates that shroud ? it just looks so, bleh

[Valentyn]

2 minutes ago, WereCat said:

I suppose the gaming Vega cards will do better in benchmarks like VR Mark. If not, they should be quite a bit cheaper at least.

 

Well that all depends on drivers. The Fury X is doing very poorly in there compared to the 980Ti, and RX 480.

So hopefully Radeon RX certainly does better.

[DildorTheDecent]

Team Red is the new Team Blue. 

 

Good job AMD, you've finally lost the plot.

 

 

[PCGuy_5960]

23 minutes ago, Valentyn said:

Cinebench = 169.72  ~7.99% slower

Cinebench Open GL score was bottlenecked by the CPU, it's mostly a single core benchmark

[Valentyn]

1 minute ago, PCGuy_5960 said:

Cinebench Open GL score was bottlenecked by the CPU, it's mostly a single core benchmark

Aye, but how often does an AMD GPU beat an NVIDIA one in OpenGL?

Both used the same systems. 

[Mr.Meerkat]

23 minutes ago, Tedny said:

Wait, FuryX faster than 1060? 

Yes, I mean the fury is faster than a 980 so...  

[PCGuy_5960]

10 minutes ago, Valentyn said:

Aye, but how often does an AMD GPU beat an NVIDIA one in OpenGL?

Idk, but in Cinebench OpenGL, if you pair a 1060 with a 7700K at 5GHz you will get a better score than a Titan Xp with a Ryzen 7 1700, because single core performance

(AFAIK)

[Valentyn]

20 minutes ago, PCGuy_5960 said:

Idk, but in Cinebench OpenGL, if you pair a 1060 with a 7700K at 5GHz you will get a better score than a Titan Xp with a Ryzen 7 1700, because single core performance

True but if you check the footnotes they use a Xeon system for consistency across all the tests.

 

So if it's accurate it should be faster than the Xp if use a faster cpu as well. 

[WereCat]

3 minutes ago, Valentyn said:

True but if you check the footnotes they use a Xeon system for consistency across all the tests.

 

So if it's accurate it should be faster than the Xp if use a faster cpu as well. 

I saw some benchmarks (quite some time ago) NVIDIA vs AMD on different CPUs and NVIDIA seems to consistently perform better than AMD on CPUs with lower IPC or clock speed, once the IPC or clock speed is higher the AMD GPU performance is starting to catch up.

If the CPUs were Xeons, they were probably low clock speed and it may have somehow affected the performance on AMD cards more.

 

I would like to see up to date benchmarks like these again but so few reviewers actually do it as it takes a lot of time and boring work.

[Valentyn]

15 minutes ago, WereCat said:

I saw some benchmarks (quite some time ago) NVIDIA vs AMD on different CPUs and NVIDIA seems to consistently perform better than AMD on CPUs with lower IPC or clock speed, once the IPC or clock speed is higher the AMD GPU performance is starting to catch up.

If the CPUs were Xeons, they were probably low clock speed and it may have somehow affected the performance on AMD cards more.

 

I would like to see up to date benchmarks like these again but so few reviewers actually do it as it takes a lot of time and boring work.

It's also worth noting the drivers used at CES are the same ones currently available for the Frontier Edition. 17.20

 

So I would expect more improvements to come. Especially for the gaming side, as noted before there are issues switching between pro and gaming. This could impact performance of either. 

[Agosto]

I have a bad feeling about this.

[i_build_nanosuits]

1 hour ago, Valentyn said:

Aye, but how often does an AMD GPU beat an NVIDIA one in OpenGL?

Both used the same systems. 

i think you are confusing OpenCL and OpenGL...traditionaly AMD GPU's are more efficient at OpenCL computing than Nvidia but when it comes to OpenGL rendering Nvidia is usually better.

[Valentyn]

6 minutes ago, i_build_nanosuits said:

i think you are confusing OpenCL and OpenGL...traditionaly AMD GPU's are more efficient at OpenCL computing than Nvidia but when it comes to OpenGL rendering Nvidia is usually better.

Nothing confused at all. Please reread my comment.

"but how often does an AMD GPU beat an NVIDIA one in OpenGL?"

 

The answer is rarely, because AMD is traditionally worse at OpenGL than NVIDIA.

Where in this case the Titan Xp was nearly 8% slower than the Vega card in OpenGL.

[i_build_nanosuits]

2 minutes ago, Valentyn said:

Nothing confused at all. Please reread my comment.

"but how often does an AMD GPU beat an NVIDIA one in OpenGL?"

 

The answer is rarely, because AMD is traditionally worse at OpenGL than NVIDIA.

Where in this case the Titan Xp was nearly 8% slower than the Vega card in OpenGL.

i read you wrong sorry about that

[VagabondWraith]

Wait for Volta™

[i_build_nanosuits]

Just now, VagabondWraith said:

Wait for Volta™

yeah let's face it, it took AMD a full 2 years to come up with something slightly worse than Nvidia's current architecture which came out a year and a half ago...they are a full generation behind now.

[PCGuy_5960]

1 minute ago, i_build_nanosuits said:

yeah let's face it, it took AMD a full 2 years to come up with something slightly worse than Nvidia's current architecture which came out a year and a half ago...they are a full generation behind now.

Well, if you want a good workstation card, Vega is probably a better option, but for gaming I bet that Vega will be on par with a 1080 +/-5%

[The Benjamins]

4 minutes ago, PCGuy_5960 said:

Well, if you want a good workstation card, Vega is probably a better option, but for gaming I bet that Vega will be on par with a 1080 +/-5%

I think it will be around 1080ti area or better. well these cards hit the selves on 7/1 and some reviews have pre ordered the card. PCPer is one that I know of that will be getting the card on release.