[WCCFTech]AMD’s FirePro Professional GPUs compared to Quadro counterparts

[WittyRemark]

AMD seems to be doing a very good job with their FirePro series.

Keep in mind that AMD has only
counted the newest generation of
Quadro GPUs so most of these are
rocking the Maxwell GM107 core.
The highest end spectrum, which
consists of the Quadro 5200 and
FirePro W8100, sees AMD leading by
about 50%. The lowest end
spectrum, which consists of the
FirePro W2100 and Quadro K420,
sees AMD leading by around 90%.
The middle end spectrum consists of
the FirePro W5100 and Quadro
K2200, sees Red leading by 33.3%.
Infact the only market segment in
which AMD is not in the lead is the
upper tier of the low end spectrum
(which is usually referred to as entry
level in the mainstream consumer
market). AMD’s FirePro GPUs
(specifically the S9150) have also
recently landed the no.1 spot in the
Green500 list, so basically AMD
powers the world’s most efficient
supercomputer . As far as cost
events of HPC clusters go,It does
look like AMD has both sides of the
coin covered doesn’t it?

http://wccftech.com/amd-compares-firepro-professional-gpus-quadro/

 

[Keysplayr]

Actually, the Quadro 5200 and K4200 and K420 are all Kepler based.
The only Maxwells here are K2200 which is GM107 at 640 CUDA cores and K620, also GM107 but severely neutered at 384 CUDA cores.
We kind of already know that Kepler isn't the computational juggernaut. But Maxwell appears to be.

And this is performance per dollar, not direct performance comparisons. Graph would look MUCH different if it were.

 

[PPB]

Actually Maxwell is another turd in regards of GPGPU performance, because the moment you walk away from canned synthethic benches, you will see lots of devs scratching their heads trying to figure how to get more performance out of the new Maxwell SKUs. OTOY's Octane Render devs are sure having a run for their money trying to figure out the new memory system and trying to satisfy the otherwise rather displeased Maxwell's userbase which, at start, have been having the same render times with a Maxwell based 980 as with a Kepler Based (and totally anemic, in the GPGPU world) 770. After 2.14 release and with a lot of tweaking inbetween, they still cant make a 970 perform to 780 levels, and 980 to 780ti levels.

Like someone at OTOY's summed it up, yay for 75% of GPGPU performance at 50% power consumption! :awe:

 

[Rvenger]

Actually, the Quadro 5200 and K4200 and K420 are all Kepler based.
The only Maxwells here are K2200 which is GM107 at 640 CUDA cores and K620, also GM107 but severely neutered at 384 CUDA cores.
We kind of already know that Kepler isn't the computational juggernaut. But Maxwell appears to be.

And this is performance per dollar, not direct performance comparisons. Graph would look MUCH different if it were.

Last time I checked, performance per dollar is very important to businesses. I sell mostly Quadros at work but I may sway towards the Firepros on our workstations if the price is right. I never get any complaints from the Quadros though.

 

[AnandThenMan]

And this is performance per dollar, not direct performance comparisons. Graph would look MUCH different if it were.

Post these graphs I'd like to see them for comparison.

 

[PPB]

I never get any complaints from the Quadros though.

Implying you would potentially get them if you switch them over very much?

 

[Rvenger]

Well, that's a very good question. I normally don't like to switch product if I don't have any support issues or performance complaints. It seems that AMD has been boosting their FirePro line lately and it may be worth trying out if my customers are willing. As long as the software has FirePro on the compatibility list.

I normally take the approach of if it's not broke, don't fix it.

 

[Erenhardt]

Is VP12 GEOMEAN some kind of seismic calculations?

Because if so, those performance results may not be valid, as suggested by some posters:

So it's easy to assume that the BoD of my company is corrupt than accept that AMD professional support could have bern a failure at the time of the decision? Yeah, *I* should be drinking a lot of kool-aid as of lately. How many professional cards have you tested in seismic applications?

 

[desprado]

Post these graphs I'd like to see them for comparison.

U mean to say PR graph which is always a BS.

And OP has posted a PR graph from AMD is totally irreverent.

 

[Gikaseixas]

U mean to say PR graph which is always a BS.

And OP has posted a PR graph from AMD is totally irreverent.

Prove them wrong then... I bet you can't but decided to come here to troll as usual

 

[desprado]

Prove them wrong then... I bet you can't but decided to come here to troll as usual

bascally is baseless to argue with you because reality is different and fanboyism is different.

 

[3DVagabond]

Well, that's a very good question. I normally don't like to switch product if I don't have any support issues or performance complaints. It seems that AMD has been boosting their FirePro line lately and it may be worth trying out if my customers are willing. As long as the software has FirePro on the compatibility list.

I normally take the approach of if it's not broke, don't fix it.

This is very important. When you are spending someone else's money you can't afford to make the wrong decision.

 

[RussianSensation]

U mean to say PR graph which is always a BS.

And OP has posted a PR graph from AMD is totally irreverent.

http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-k5200-k4200-k2200.html

Go ahead look at the real world benchmarks. Don't forget Quadro is DP crippled which means if you do compute as well you need to spend extra on Tesla. With AMD you get 2 in 1!

As far as Maxwell coming, AMD will have FirePro based on R9 300 series to compete. Professionals are paying attention as AMD went from 12% to 25% in professional market share. Before that they had 5%.

With Apple pushing OpenCL, the proprietary stronghold of legacy CUDA apps is under pressure. The traditional MacPro users might be appalled at the idea of using OpenCL when they have used CUDA since 2008 but if OpenCL keeps improving, newer professionals entering the workforce won't be so close minded. The market is too dynamic to be stuck with old ways of doing things for too long.

 

[mrmt]

Is VP12 GEOMEAN some kind of seismic calculations?

No, it is not. VP12 is short for SPEC ViewPerf V12. And I didn't imply that the results from any AMD cards weren't valid, what I implied is that while you have 0 experience with the professional GPU environment, and more important, 0 experience with the kind of high complexity environment the company I work for has, that didn't stop you from calling the entire technical team and BoD of corrupt because they adopted a policy of no AMD professional cards on the company.

 

[mrmt]

With Apple pushing OpenCL, the proprietary stronghold of legacy CUDA apps is under pressure. The traditional MacPro users might be appalled at the idea of using OpenCL when they have used CUDA since 2008 but if OpenCL keeps improving, newer professionals entering the workforce won't be so close minded. The market is too dynamic to be stuck with old ways of doing things for too long.

I don't think Apple is pushing OpenCL, it is having to deal with OpenCL the same way they had to deal with CUDA. Apple's long term vision seems to be using the Metal API for both graphics and compute.

 

[3DVagabond]

I don't think Apple is pushing OpenCL, it is having to deal with OpenCL the same way they had to deal with CUDA. Apple's long term vision seems to be using the Metal API for both graphics and compute.

Apple were the designers of OpenCL. Maybe they were tired of having to deal with CUDA and wanted to have the freedom of an open standard?

 

[mrmt]

Apple were the designers of OpenCL. Maybe they were tired of having to deal with CUDA and wanted to have the freedom of an open standard?

They can have even more freedom with their own proprietary API (Metal).

I don't think CUDA is their real target, I think their real target is to close the garden by evicting OpenGL of their mobile platform, and if they are going to develop a Graphics API, they are building a compute API as a secondary venture.

 

[ShintaiDK]

A company sldie that shows company is better than competition. Surprise!

 

[3DVagabond]

They can have even more freedom with their own proprietary API (Metal).

I don't think CUDA is their real target, I think their real target is to close the garden by evicting OpenGL of their mobile platform, and if they are going to develop a Graphics API, they are building a compute API as a secondary venture.

I don't think you understood me. Apple isn't having to deal with OpenCL like they did with CUDA, as you said. Apple wrote/designed OpenCL. Metal and OpenGL are rendering API's. OpenCL is for computational tasks.

 

[mrmt]

I don't think you understood me. Apple isn't having to deal with OpenCL like they did with CUDA, as you said. Apple wrote/designed OpenCL. Metal and OpenGL are rendering API's. OpenCL is for computational tasks.

Nope. Metal is also GPGPU:

https://developer.apple.com/library...ogrammingGuide/Introduction/Introduction.html

About Metal and this Guide

The Metal framework supports GPU-accelerated advanced 3D graphics rendering and data-parallel computation workloads. Metal provides a modern and streamlined API for fine-grain, low-level control of the organization, processing, and submission of graphics and computation commands and the management of the associated data and resources for these commands. A primary goal of Metal is to minimize the CPU overhead necessary for executing these GPU workloads.

 

[dacostafilipe]

I don't think Apple is pushing OpenCL

OSX uses more and more OpenCL under the hood. Some apps (ex: Preview.app) do not work or crash without it. This is certainly a sign that Apple is interested in OpenCL.

Even if Metal supports GPGPU in iOS, I don't see why Apple would try to use it on OSX. OpenCL does the job already, it did not in iOS because of the lack of drivers.

 

[Keysplayr]

http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-k5200-k4200-k2200.html

Go ahead look at the real world benchmarks. Don't forget Quadro is DP crippled which means if you do compute as well you need to spend extra on Tesla. With AMD you get 2 in 1!

As far as Maxwell coming, AMD will have FirePro based on R9 300 series to compete. Professionals are paying attention as AMD went from 12% to 25% in professional market share. Before that they had 5%.

With Apple pushing OpenCL, the proprietary stronghold of legacy CUDA apps is under pressure. The traditional MacPro users might be appalled at the idea of using OpenCL when they have used CUDA since 2008 but if OpenCL keeps improving, newer professionals entering the workforce won't be so close minded. The market is too dynamic to be stuck with old ways of doing things for too long.

http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-k5200-k4200-k2200_11.html

"The new cards are also superior to AMD’s products of the same price categories, offering the most attractive price/performance ratio to professional users."

This seems to go directly against the graph in the OP here.
From the conclusion of the article you linked to. There are more gems in there as well.

 

[desprado]

http://www.xbitlabs.com/articles/graphics/display/nvidia-quadro-k5200-k4200-k2200_11.html

"The new cards are also superior to AMD’s products of the same price categories, offering the most attractive price/performance ratio to professional users."

This seems to go directly against the graph in the OP here.
From the conclusion of the article you linked to. There are more gems in there as well.

u are wasting time.They Dont judge performance and e.t.c on facts because they have different method with no source and links.

 

[mrmt]

OSX uses more and more OpenCL under the hood. Some apps (ex: Preview.app) do not work or crash without it. This is certainly a sign that Apple is interested in OpenCL.

It would be a very interesting move if Apple developed their own compute API but in the long run kept supporting OpenCL.

 

[Flapdrol1337]

As far as Maxwell coming, AMD will have FirePro based on R9 300 series to compete. Professionals are paying attention as AMD went from 12% to 25% in professional market share. Before that they had 5%.

With Apple pushing OpenCL, the proprietary stronghold of legacy CUDA apps is under pressure. The traditional MacPro users might be appalled at the idea of using OpenCL when they have used CUDA since 2008 but if OpenCL keeps improving, newer professionals entering the workforce won't be so close minded. The market is too dynamic to be stuck with old ways of doing things for too long.

Those mac pro's are good for market share yes, 2 "firepro" gpu's in there, only cheaper, and without the firepro perks.