Ask Me Anything - Official AMD Radeon Representatives

[matta85]

Will amd release anything similar to shadowplay in the next few months?

 

[Thracks]

Will amd release anything similar to shadowplay in the next few months?

Our approach to providing game streaming to users is through the Raptr app: http://raptr.com/amd

 

[Skysnake]

2. Since the HD5k series AMD have GDS, but AMD never sayed something abaout it. The only thing in the programmer guide is on http://developer.amd.com/wordpress/media/2012/12/AMD_Southern_Islands_Instruction_Set_Architecture.pdf ftrom 2012 we can read

This part is gone in the 2013 document: http://developer.amd.com/wordpress/media/2013/07/AMD_Sea_Islands_Instruction_Set_Architecture.pdf

No, it's not. Check chapter 13.5 (page 13-36/37) or table 12.7 (page 12-142/143). It even got one new GWS instruction, DS_GWS_SEMA_RELEASE_ALL.

Hehe hi Gipsel

Sorry, i don't mean the ISA Instructions themself.

I mean the documentation text

2012 Version:

2.3.3.1 Global Data Share (GDS)
The SI-GPU contains a 64 kB global data share memory that can be used by wavefronts of running a kernel on all compute units. This memory enables sharing of data across multiple workgroups. The GDS is configured with 32 banks, each with 512 entries of 4 bytes. It provides full access to any location for use by any wavefront. The GDS also supports 32-bit integer atomic operations to enable fast, unordered atomics. Data can be reloaded from memory prior to kernel launch and written to memory after kernel completion. The GDS block contains support logic for unordered append/consume and domain-launch-ordered append/consume operations through the global wave sync (GWS). These dedicated circuits enable fast compaction of data or the creation of complex data structures in memory.

2013 Version:

2.3.2Global Data Share (GDS)
The AMD Sea Islands series of devices uses a 64 kB global data share (GDS) memory that can be used by wavefronts of a kernel on all compute units. This memory enables 128 bytes of low-latency bandwidth to all the processing elements. The GDS is configured with 32 banks, each with 512 entries of 4 bytes each. It provides full access to any location for any processor. The shared memory contains 32 integer atomic units to enable fast, unordered atomic operations. This memory can be used as a software cache to store important control data for compute kernels, reduction operations, or a small global shared surface. Data can be preloaded from memory prior to kernel launch and written to memory after kernel completion. The GDS block contains support logic for unordered append/consume and domain launch ordered append/consume operations to buffers in memory. These dedicated circuits enable fast compaction of data or the creation of complex data structures in memory.

Without any detailed documentation how to use the ISA-Unstructions for ME they are completely useless. The minimum is a good documentation. And what i/we want is a Api-call

Who really wants to programm GPUs with the ISA-calls :ugly: This is the way to insanity

 

[mblaine]

I just bought a 7970 to replace an old 4850 and get me back into gaming after 4-5 years away. It's fantastic! I have a Radeon Rewards Gold Ticket. Do you happen to know when new games might be added to the tiers? I've been holding off on redeeming in the hopes that some new games would be released, but recently saw that Tomb Raider dropped off. Are new games coming or is the promo "over" and existing supply is being used up, leaving less-desirable games?

 

[Thracks]

I strongly suggest those of you interested in HPC visit our dedicated developer forum: http://devgurus.amd.com/welcome

I'm sure you would be able to find the answers to your questions there. AMD developers frequent that community and can help with ISA, APIs, SDK questions.

 

[Yvese]

Hi, AMD. Great AMA so far!

I'm currently on a 7950 CF setup. Great cards with amazing value. My question is, will Mantle help with CF scaling and GPU usage? In BF4 for example, my 2600k @ 4.5ghz is near maxed out, yet my cards are at 75% usage most of the time, sometimes even dropping to 50%. This is on a 64 player server so this is clearly a CPU bottleneck ( or maybe it's the game being unoptimized? ). Will Mantle help with this?

On the topic of CF and mantle, is there anything else aside from scaling and GPU usage that you feel Mantle will help with?

We're assessing some tweaks that will bring BF4 to 90-95% scaling 1->2 GPUs. That's without Mantle.

Going forward, Mantle could help any game with CPU bottlenecking and multi-GPU scaling. It's deeply parallelized as an API.

Thanks for the response! I can't wait to see those tweaks

Two more questions:

I'm on a 144hz monitor ( Asus VG248QE ). There are issues with this refresh rate on the desktop like cards not downclocking. I have fixed this by dropping down to 120hz. Thing is, I get random artifacts while on the desktop and/or browsing. Is this a driver issue? If so, any plans on resolving it?

Second question - On the topic of 144hz, when I first got my 2nd card to CF, I would get BSODs when loading games @ 144hz. Games like Hitman where it lets you set your settings before loading the game, and Sleeping Dogs where you can't change refresh rate. There are people with 7990's that also have this issue and only happens in CF.

I've fixed it by adding a 2nd CF bridge ( no idea how or why ), however those with 7990's aren't so lucky. Dropping down to 120hz fixes it. Will this be resolved in a future driver?

 

[nuts32605]

Hey guys, thanks for your answer to my last question. I have another to ask you:
What happens when a 280X and a 7970 are in crossfire and you start a mantle-enabled game. Do you not get mantle? Or does it only run on the 280X?

 

[Thracks]

Hi, AMD. Great AMA so far!

I'm currently on a 7950 CF setup. Great cards with amazing value. My question is, will Mantle help with CF scaling and GPU usage? In BF4 for example, my 2600k @ 4.5ghz is near maxed out, yet my cards are at 75% usage most of the time, sometimes even dropping to 50%. This is on a 64 player server so this is clearly a CPU bottleneck ( or maybe it's the game being unoptimized? ). Will Mantle help with this?

On the topic of CF and mantle, is there anything else aside from scaling and GPU usage that you feel Mantle will help with?

We're assessing some tweaks that will bring BF4 to 90-95% scaling 1->2 GPUs. That's without Mantle.

Going forward, Mantle could help any game with CPU bottlenecking and multi-GPU scaling. It's deeply parallelized as an API.

Thanks for the response! I can't wait to see those tweaks

Two more questions:

I'm on a 144hz monitor ( Asus VG248QE ). There are issues with this refresh rate on the desktop like cards not downclocking. I have fixed this by dropping down to 120hz. Thing is, I get random artifacts while on the desktop and/or browsing. Is this a driver issue? If so, any plans on resolving it?

Second question - On the topic of 144hz, when I first got my 2nd card to CF, I would get BSODs when loading games @ 144hz. Games like Hitman where it lets you set your settings before loading the game, and Sleeping Dogs where you can't change refresh rate. There are people with 7990's that also have this issue and only happens in CF.

I've fixed it by adding a 2nd CF bridge ( no idea how or why ), however those with 7990's aren't so lucky. Dropping down to 120hz fixes it. Will this be resolved in a future driver?

I've taken your post and handed it directly to the project lead for our drivers. A fix will depend on other priorities, but I can confirm personally that they are aware.

 

[Thracks]

Hey guys, thanks for your answer to my last question. I have another to ask you:
What happens when a 280X and a 7970 are in crossfire and you start a mantle-enabled game. Do you not get mantle? Or does it only run on the 280X?

Both GPUs are Mantle-ready. You would have Mantle and CrossFire.

 

[Skysnake]

I strongly suggest those of you interested in HPC visit our dedicated developer forum: http://devgurus.amd.com/welcome

I'm sure you would be able to find the answers to your questions there. AMD developers frequent that community and can help with ISA, APIs, SDK questions.

Not really, i have posted some of the questions there before and after the restart of the develoüer forum. The support ist.. ähh... let us say, there is no official support from techies

I have more or less never get a answer, and a direct contact is more or less impossible to AMD when you are no company, and even than it is hard, because so many people are gone.

My questions 9 and 10 are no HPC/GPGPU specific questions, are you able to answer them? I think many people are really interested in the HPM<->HP question

And FCAT is in the german community a BIG thing.

 

[Thracks]

My questions 9 and 10 are no HPC/GPGPU specific questions, are you able to answer them? I think many people are really interested in the HPM<->HP question

And FCAT is in the german community a BIG thing.

The answer to #9: it's 28nm HP. HPM is for mobile solutions.

The answer to the FCAT question, FCAT is not the end-all-be-all of frame pacing. There are many provable scenarios where the observed performance is stutter-free, whereas FCAT results suggest that it should be stuttering like crazy. If I recall correctly, Tomb Raider is an instance of this. People should use their own eyeballs vs. relying on an automated test, because the automated testing absolutely does not tell the whole story.

What's clear, however. is that we had an issue with the consistency in frame delivery, and we've largely resolved that problem. The remaining scenarios will be resolved this quarter with a driver update that intelligently and algorithmically normalizes frame times.

 

[scopey86]

As a forward looking question, how do you see the discrete GPU market evolving over the next 5 years? What do you think will be the significant challenges for the industry given the continuously improving APUs, the threat of cloud processing (aka gaikai, nvidia's solutions etc), and the rising concerns of shrinking die sizes?

 

[jkhoward]

Hi, AMD. Great AMA so far!

I'm currently on a 7950 CF setup. Great cards with amazing value. My question is, will Mantle help with CF scaling and GPU usage? In BF4 for example, my 2600k @ 4.5ghz is near maxed out, yet my cards are at 75% usage most of the time, sometimes even dropping to 50%. This is on a 64 player server so this is clearly a CPU bottleneck ( or maybe it's the game being unoptimized? ). Will Mantle help with this?

On the topic of CF and mantle, is there anything else aside from scaling and GPU usage that you feel Mantle will help with?

We're assessing some tweaks that will bring BF4 to 90-95% scaling 1->2 GPUs. That's without Mantle.

Going forward, Mantle could help any game with CPU bottlenecking and multi-GPU scaling. It's deeply parallelized as an API.

Thanks for the response! I can't wait to see those tweaks

Two more questions:

I'm on a 144hz monitor ( Asus VG248QE ). There are issues with this refresh rate on the desktop like cards not downclocking. I have fixed this by dropping down to 120hz. Thing is, I get random artifacts while on the desktop and/or browsing. Is this a driver issue? If so, any plans on resolving it?

Second question - On the topic of 144hz, when I first got my 2nd card to CF, I would get BSODs when loading games @ 144hz. Games like Hitman where it lets you set your settings before loading the game, and Sleeping Dogs where you can't change refresh rate. There are people with 7990's that also have this issue and only happens in CF.

I've fixed it by adding a 2nd CF bridge ( no idea how or why ), however those with 7990's aren't so lucky. Dropping down to 120hz fixes it. Will this be resolved in a future driver?

I've taken your post and handed it directly to the project lead for our drivers. A fix will depend on other priorities, but I can confirm personally that they are aware.

Just to add my two cents on this..

I had the EXACT same issue and it took me forever to figure out the issue. I was able to run CFX with 1 bridge for the longest time and once I got my new monitor I was required to use two CFX bridges.

 

[Guest]

Will amd release anything similar to shadowplay in the next few months?

Our approach to providing game streaming to users is through the Raptr app: http://raptr.com/amd

As an owner of AMD GPU who is very interested in ShadowPlay, this doesn't really answer the question.

ShadowPlay allows users to record 1080p 60FPS video without performance hit thanks to H.264 encoder built directly into the GPU. It essentially records all the time and saves only what you wish to save.

Does AMD have a similar feature planned as a response?

 

[scopey86]

Another question I would love to ask you guys is your opinions on PCI lane bottlenecking. If you were to have 2 290x in Crossfire, do you believe that the performance from an intel x79 chipset's extra pci lanes allowing for dual x16 on the cards will provide a benefit to z87 or amd's boards dual x8 lanes? If so, how much of a difference do you think it would make for?

 

[ltdementhial]

Hi AMD, this is a tough question but some one need to ask.

Can or will AMD make something to make the prices more equitative in other countrys? USA-Europe and some asian countrys seem to have very similar prices, but south america suffers for massive overprices (twice sometimes almost three times the USA price on most ocassions) a great example is the R9 260X wich is almost 400 dolars on major retailers and e-tailers, r9 280x sells at almost 750 dolars and r9 290x above 1200 dollars. I know so much people that buy AMD Cpu's because they're priced many times a couple dollars more expensive than in USA that's not much of a problem, but the GPU market is a very different story. Most people buy HD 7850 at most (above 300 US dollars) 7870? extra 70-100 dollars 7950? 450 dollars and so on.

 

[monstercameron]

no word on wayland/mr drivers? or android support?

 

[joditas]

Why do development and provision of AMD graphics card drivers last for shorter periods of time then Nvidia? They seem to be hosting really, really old drivers, updated for use on modern computers, and I haven't seen that so much with you guys.

Our driver support is identical to NVIDIA. They support GPUs from 2010 (or later) in Windows 8.1, as do we.

Their latest driver (supports up to Win 8.1) supports 8000 series card, which dated back to 2006. Radeon 4000 series came up later, yet you guys already ceased support them in Win 8.1

AMD Catalyst 13.9 supports HD 2000, 3000 and 4000 in Windows 8.1. The driver must be installed through the device manager, however, as it has not yet passed Microsoft driver certification, rendering it ineligible for automatic installation via .exe.

Ok will it ever pass the certification? I think information like this needs to be published. Even a beta driver for older cards for Win 8.1 would be better than nothing.


Regarding TrueAudio, since only R9 290X, R9 290 and R7 260X support it, and 4 other don't. Don't you think it will create confusion, and anger perhaps as people thought they bought the latest and greatest "series" from AMD only to know later that their particular model doesn't support it. Even for someone like me who pay close attention to the latest have to double check the models, it's going to be tougher for average consumers.

If you can add something to the model name as a transition, for example TA behind the name that support it, it would be better than nothing.

 

[jpishgar]

Hey there friends!

It's getting on in the day, and so we're going to break for the evening to provide our guests with an opportunity to eat, rest, and relax for a bit from all the great questions. Please feel free to continue posting your questions here to this thread, and the reps from AMD will back back tomorrow morning to follow-up with responses from the overnight. Thanks all - and keep the great questions rolling!

-JP

 

[wolfkraut]

I apologize if this has been asked all ready, but does AMD have any plans to release an equivalent to Shadowplay? This is one of the few features Nvidia has offered recently that has me interested and it would be great if something like this could come to AMD cards.

Also, this may not be the appropriate place to ask this question, but a version of a mod for DX9 PC games called "ENB" specifically for Fallout games has recently become incompatible for AMD 7xxx series card owners, causing the graphics driver to crash, and instead only works on older models and Nvidia cards. This is unfortunate as these mods are very popular among PC gamers, significantly improving the graphical quality of these games. The developer of the mod claims this is an issue with AMD drivers, what is AMDs stance on this? Do you plan to support developers who create programs such as this in the future or perhaps make it easier for the them to communicate with and report bugs to AMDs driver team at a level higher than that of say, an average user who doesn't understanding the coding aspect of drivers and such?

Thanks!

 

[blackpanther26]

I just recently build a second computer with the A6 5400k APU. I also bought a Radeon 7750. Is there any plans in the future to release a driver to allow dual GPU plus x-fire with another 7750. Similar to 2 radeon 7750's and the APU in dual graphics working together in a small form factor case as a low profile case. I think that would be a awesome idea. I do like where you guys are going with the APU's. And is there plains to shrink the SOC to 22nm from 32nm?

Sry if my question is confusing answer what you can and I'm glad you can be here today.

 

[coupe]

Hi, AMD. Great AMA so far!

I'm currently on a 7950 CF setup. Great cards with amazing value. My question is, will Mantle help with CF scaling and GPU usage? In BF4 for example, my 2600k @ 4.5ghz is near maxed out, yet my cards are at 75% usage most of the time, sometimes even dropping to 50%. This is on a 64 player server so this is clearly a CPU bottleneck ( or maybe it's the game being unoptimized? ). Will Mantle help with this?

On the topic of CF and mantle, is there anything else aside from scaling and GPU usage that you feel Mantle will help with?

We're assessing some tweaks that will bring BF4 to 90-95% scaling 1->2 GPUs. That's without Mantle.

Going forward, Mantle could help any game with CPU bottlenecking and multi-GPU scaling. It's deeply parallelized as an API.

Thanks for the response! I can't wait to see those tweaks

Two more questions:

I'm on a 144hz monitor ( Asus VG248QE ). There are issues with this refresh rate on the desktop like cards not downclocking. I have fixed this by dropping down to 120hz. Thing is, I get random artifacts while on the desktop and/or browsing. Is this a driver issue? If so, any plans on resolving it?

Second question - On the topic of 144hz, when I first got my 2nd card to CF, I would get BSODs when loading games @ 144hz. Games like Hitman where it lets you set your settings before loading the game, and Sleeping Dogs where you can't change refresh rate. There are people with 7990's that also have this issue and only happens in CF.

I've fixed it by adding a 2nd CF bridge ( no idea how or why ), however those with 7990's aren't so lucky. Dropping down to 120hz fixes it. Will this be resolved in a future driver?

I've taken your post and handed it directly to the project lead for our drivers. A fix will depend on other priorities, but I can confirm personally that they are aware.

Just a FYI. I heard that a second crossfire bridge helps because the bandwidth of one bridge isn't suitable for the amount of data 144 hz is transmitting.

So with 7990 it depends on the bandwidth of the bridge on the device.

You have to figure any bridge has a bandwidth limit (I think its 9 GB/s with a crossfire bridge). So when you increase resolution and/or refresh rate more data is going across the bridge and when it hits its cap the signal breaks down.

Much like Display Port is popular on high end devices because it offers the most bandwidth for high resolution, high refresh rate, and multi monitor solutions.

 

[Calculatron]

What is the future for AMD Dual Graphics? Will this option be be perfected and built upon?

 

[smeezekitty]

Hi AMD, this is a tough question but some one need to ask.

Can or will AMD make something to make the prices more equitative in other countrys? USA-Europe and some asian countrys seem to have very similar prices, but south america suffers for massive overprices (twice sometimes almost three times the USA price on most ocassions) a great example is the R9 260X wich is almost 400 dolars on major retailers and e-tailers, r9 280x sells at almost 750 dolars and r9 290x above 1200 dollars. I know so much people that buy AMD Cpu's because they're priced many times a couple dollars more expensive than in USA that's not much of a problem, but the GPU market is a very different story. Most people buy HD 7850 at most (above 300 US dollars) 7870? extra 70-100 dollars 7950? 450 dollars and so on.

Thats not really AMDs fault. Chances are most import products in that country are marked up.

 

[Skysnake]

The answer to #9: it's 28nm HP. HPM is for mobile solutions.

Of course, but there was a lot of rumors out there, that AMD use the 28nm HPM prpcess from TSMC and not the normal HP process to save energy.

So very nice to know, that AMD use the same process for Tahiti and Hawaii.

The answer to the FCAT question, FCAT is not the end-all-be-all of frame pacing. There are many provable scenarios where the observed performance is stutter-free, whereas FCAT results suggest that it should be stuttering like crazy. If I recall correctly, Tomb Raider is an instance of this. People should use their own eyeballs vs. relying on an automated test, because the automated testing absolutely does not tell the whole story.

What's clear, however. is that we had an issue with the consistency in frame delivery, and we've largely resolved that problem. The remaining scenarios will be resolved this quarter with a driver update that intelligently and algorithmically normalizes frame times.
So also no real fundamental solution of the problem.

Too poor, that nVidia more like it to blame AMD than to work with them hand in hand to give the developers some API calls so that they can solve the problem with the divergent timeing of the ingametime and the displaytime in there engines -.-

This is one of the reasons why i don't like nVidias company politic. If it helps to sell hardware, they just score off somebody. And that with a longtime plan... The same with there proprietary "standards". They don't care if it is bad for the whole industry or not. They just look if they sell more cards...

I hope AMD don't go the same path in the future. Mantle can be something like this...

TressFX ist the right way, and i really really really hope all the people out there understands soon, that proprietary is no choice in the long term view. Proprietary kills advancement.

I am really interested in OpenCL 2.0. And btw it is not bad to spread out, that the Kepler GPU don't support OpenCL official I think nVidia will repent there decision in the future.