John Carmack:You need a 3.68 Teraflop GPU to match PS4's GPU

[QUOTE="GioVela2010"][QUOTE="pcgamingowns"] yea a piece of shit not capable of gaming try owning 5 gaming rigsAMD655

My nan has 2 massive rigs that reach her knees, and?

[QUOTE="AMD655"]

[QUOTE="GioVela2010"] I have more than 5GioVela2010

My nan has 2 massive rigs that reach her knees, and?

Wow, i think you should go masturbate the outrageous hormone splurge is ridiculous.

Stay mad.

 [QUOTE="ronvalencia"]

The dummy is you. I gave you hints with PCI-E version 3.0 16 lanes being used as "CPU I/O bus" before I'll post a practical example of a HSA enabled PC with a discrete Radeon HD 7800 and no copy process.

 

Do you have practical programming on AMD HSA?

 

Your "than those on PC which i suspect are zero" statement is a joke.

 

savagetwinkie

[QUOTE="tormentos"]

 

You are as dumb as Ronvalencia HSA is a design,without hUMA is not true HSA,so tell me what PC has hUMA now.?

 

There are more games on consoles been done now in HSA maner,than those on PC which i suspect are zero.

ronvalencia

The dummy is you. I gave you hints with PCI-E version 3.0 16 lanes being used as "CPU I/O bus" before I'll post a practical example of a HSA enabled PC with a discrete Radeon HD 7800 and no copy process.

Do you have practical programming on AMD HSA?

Your "than those on PC which i suspect are zero" statement is a joke.

:lol: its not true HSA without HUMA? HUMA is for APU's with a shared memory pool,  With dedicated memory for the cpu and gpu as Ron has stated they will use the PCI-E lane to allow the gpu and cpu to communicate. AMD has partially resident textures aka PRT with all GCN based gpu's which means that even if a gpu has less memory then console version is allocating PRT stream's small bits of these massive textures into the GPU as needed, giving compatible games a virtually endless supply of  texture data it can apply to the game world without over taxing the buffer because of repeat textures.

[QUOTE="GioVela2010"][QUOTE="AMD655"]

 

My nan has 2 massive rigs that reach her knees, and?

AMD655

Wow, i think you should go masturbate the outrageous hormone splurge is ridiculous.

Stay mad.

[QUOTE="AMD655"]

[QUOTE="GioVela2010"] WTF is your point dumbdumb? He told me to try owning 5 gaming rigs. I simply answered him. I didnt ask your pleb ugly self about your nan's massive rigs. GioVela2010

Wow, i think you should go masturbate the outrageous hormone splurge is ridiculous.

Stay mad.

??

I am not a mirror unfortunately.

[QUOTE="GioVela2010"][QUOTE="AMD655"]

 

Wow, i think you should go masturbate the outrageous hormone splurge is ridiculous.

 

Stay mad.

AMD655

??

I am not a mirror unfortunately.

feeding the trolls is punishable by loss of brain cells

[QUOTE="AMD655"]

[QUOTE="GioVela2010"] Stay uglywis3boi

??

I am not a mirror unfortunately.

feeding the trolls is punishable by loss of brain cells

[QUOTE="pcgamingowns"]try owning 5 gaming rigsConanTheStoner

Why?

[QUOTE="ConanTheStoner"]

[QUOTE="pcgamingowns"]try owning 5 gaming rigsMonsieurX

Why?

lol aliens colonial marines on pc looks better than these shitty ps4/fakebox launch games

lol aliens colonial marines on pc looks better than these shitty ps4/fakebox launch games

pcgamingowns

[QUOTE="AM-Gamer"]

[QUOTE="04dcarraher"] The x1950pro was not even unified shader based it had a fixed 8 vertex and 36 pixel processors and only operated at 248 GFLOPS and used GDDR3 it was not more advanced then the Xenos in the 360 and was developed in 2005 and finished in mid 2006 AMD had to hold back the x1800 series months to fix a design flaw, and was still released before the 8800GT. As you know the PS3 is using a gimped G70 chip aka 7800, that its performance lies around 7800GS levels. Until the Cell was used to off load shader intensive jobs the PS3 was sub par even compared to a normal 7800. Comparing one gpu architecture to another that are not even related (based on same foundation) is not the way to compare their processing abilities. also should note that AMD stopped theoretical performances with their 6000 series. However your purposely ignoring the fact the PS4 gpu is GCN based which means you can directly compare its processing performance with other GCN based gpu's. and that means that the PS4 gpu falls behind the 7870 and falls flat on its face when compared to 7870XT, 7950 or 7970GE. pcgamingowns

Yes because Windows overhead and being able to code right to the metal means nothing, please STFU ill take JC word over you. 

[QUOTE="pcgamingowns"]

[QUOTE="AM-Gamer"]

Yes because Windows overhead and being able to code right to the metal means nothing, please STFU ill take JC word over you. 

xboxiphoneps3

[QUOTE="ronvalencia"]

The dummy is you. I gave you hints with PCI-E version 3.0 16 lanes being used as "CPU I/O bus" before I'll post a practical example of a HSA enabled PC with a discrete Radeon HD 7800 and no copy process.

Do you have practical programming on AMD HSA?

Your "than those on PC which i suspect are zero" statement is a joke.

savagetwinkie

Did you miss this post?

http://fabricengine.com/2012/07/gpu_computation_technology_preview/

This 3rd party app running on beta HSA on Radeon HD 7800.

The AMD HSA technology platform has the goal of providing a heterogeneous computation platform in which both CPU and GPU cores access and manipulate memory identically. HSA will enable complex data structures with pointer indirection to be shared between the CPU and GPU. Not only will no copying of data between different memory spaces be necessary, but the pointers imbedded in a complex data structure will be usable without change on both CPU and GPU cores.

In collaboration with AMD, the Fabric Engine development team has extended the KL compiler and Fabric Engine Core execution environment to support GPU computation on high-end AMD GPUs. The primary means by which this preliminary work was possible was the availability of an LLVM back end for AMD GPU hardware.

The animated scene was run on a workstation with an AMD A10-5800K APU with both integrated graphics and a discrete Radeon HD 7800 card; however, only the discrete card was used for GPU computation and OpenGL rendering for these tests.

--------------------------------------------------------

Notice the concurrent HSA compute and OpenGL on discrete Radeon HD 7800. The devs didn't even bother using the IGP in AMD A10-5800K APU.

PS; No copying would be involved when the right work allocation was given to the right processor unit from the beginning or GDDR5 is treated like hUMA with PCI-E bus acting like PS4's CPU I/O bus.

With AMD HSA ecosystem,

1. The CPU can directly access GPU's memory address locations (located in GDDR5 physical space) i.e. PCI-E version 3.0 16 lane (32 GB/s full duplex) acts like "<20GB/s CPU I/O bus" in the PS4. Radeon HD 7800's GDDR5 acts like GDDR5 memory in PS4. Can be treated like a hUMA.

2. GPU can directly access CPU's memory address locations (located in DDR3 physical space) i.e. the reverse of the above process. The DX11.2 tiled resource/AMD PRT would be important to JIT load textures from the larger DDR3 memory pool.

Current AMD Gaming Evolved PC box can be configured to act like PS4. The main physical difference is the non-single chip solution with the PC i.e. the PC hardware is not cheap compared to PS4.

AMD Gaming Evolved PC box act like X1 and PS4.

Btw, AMD Trinity's PCI-E version 2.0 16 lane bus (16 GB/s full duplex) would be lower than "<20 GB/s CPU I/O bus" in PS4.

Radeon HD 7800 is acting like an APU with it's GDDR5 memory and it's connected to an external CPU (connected via PCI-E) i.e. AMD Piledriver dual module is too large to be glued with Radeon HD 7800.

[QUOTE="pcgamingowns"]

lol aliens colonial marines on pc looks better than these shitty ps4/fakebox launch games

mitu123

[QUOTE="MonsieurX"][QUOTE="ConanTheStoner"]

Why?

pcgamingowns

[QUOTE="nameless12345"]

 

 

Consoles are more advanced than PCs as far as unified CPU/GPU (APU) and RAM tech is regarded.

But one has to consider that most PC gamers rather use a discrete GPU + CPU combo as it's more powerful than the best APUs out there.

But eventually we will see high-powered APUs on the PC too.

tormentos

Oh for sure i did  not argue that.

PC will stay on top even without HSA,but the PS4 will be able to do some thing that the 7870 GPU will not in the end,thanks to having efficient design,HSA, and several other advantages it has over PC.

The PS3 GPU beat the 7800GTX,even that the GPU inside the PS3 is actually weaker than a 7800GTX,with almost half the bandwidth,lower ROP count and half the memory.

But the PS3 has the Cell to offload some of the work to and is capable of handling shader intensive games because of this.

The 7800GTX lacks unified shaders so the comparison is not exactly a fair one.

On non shader intensive games the 7800GTX easily beats the PS3.

The PS4 is not even out so we have no way of knowing how well it can perform and how much HSA will help in performance.

[QUOTE="mitu123"]

[QUOTE="pcgamingowns"]

lol aliens colonial marines on pc looks better than these shitty ps4/fakebox launch games

MonsieurX

It's funny to see PC gamers here pretending as knowledgable game developers with crediblity. :lol: No amount of lies can save your overpaid PC.

It's funny to see PC gamers here pretending as knowledgable game developers with crediblity. :lol: No amount of lies can save your overpaid PC.

Mr_BillGates

Much more funny to see a consolite twisting the developer's words... can you show me where Carmack explicitly stated what's written in the thread title?

[QUOTE="tormentos"]

[QUOTE="nameless12345"]

Consoles are more advanced than PCs as far as unified CPU/GPU (APU) and RAM tech is regarded.

But one has to consider that most PC gamers rather use a discrete GPU + CPU combo as it's more powerful than the best APUs out there.

But eventually we will see high-powered APUs on the PC too.

RyviusARC

Oh for sure i did not argue that.

PC will stay on top even without HSA,but the PS4 will be able to do some thing that the 7870 GPU will not in the end,thanks to having efficient design,HSA, and several other advantages it has over PC.

The PS3 GPU beat the 7800GTX,even that the GPU inside the PS3 is actually weaker than a 7800GTX,with almost half the bandwidth,lower ROP count and half the memory.

But the PS3 has the Cell to offload some of the work to and is capable of handling shader intensive games because of this.

The 7800GTX lacks unified shaders so the comparison is not exactly a fair one.

On non shader intensive games the 7800GTX easily beats the PS3.

The PS4 is not even out so we have no way of knowing how well it can perform and how much HSA will help in performance.

More info on HSA, DirectCompute issues PC games from http://www.slideshare.net/zlatan4177/gpgpu-algorithms-in-games

Both DX11.2 and HSA for Windows will lower the API overheads.

It's funny to see PC gamers here pretending as knowledgable game developers with crediblity. :lol: No amount of lies can save your overpaid PC.

Mr_BillGates

Mr_BillGates

K.

[QUOTE="RyviusARC"]

[QUOTE="tormentos"]

 

Oh for sure i did not argue that.

 

PC will stay on top even without HSA,but the PS4 will be able to do some thing that the 7870 GPU will not in the end,thanks to having efficient design,HSA, and several other advantages it has over PC.

The PS3 GPU beat the 7800GTX,even that the GPU inside the PS3 is actually weaker than a 7800GTX,with almost half the bandwidth,lower ROP count and half the memory.

ronvalencia

But the PS3 has the Cell to offload some of the work to and is capable of handling shader intensive games because of this.

The 7800GTX lacks unified shaders so the comparison is not exactly a fair one.

On non shader intensive games the 7800GTX easily beats the PS3.

The PS4 is not even out so we have no way of knowing how well it can perform and how much HSA will help in performance.

More info on HSA, DirectCompute issues PC games from http://www.slideshare.net/zlatan4177/gpgpu-algorithms-in-games

Both DX11.2 and HSA for Windows will lower the API overheads.

But most of the API overhead seems to effect the CPU.

[QUOTE="ronvalencia"]

[QUOTE="RyviusARC"]

But the PS3 has the Cell to offload some of the work to and is capable of handling shader intensive games because of this.

The 7800GTX lacks unified shaders so the comparison is not exactly a fair one.

On non shader intensive games the 7800GTX easily beats the PS3.

The PS4 is not even out so we have no way of knowing how well it can perform and how much HSA will help in performance.

RyviusARC

More info on HSA, DirectCompute issues PC games from http://www.slideshare.net/zlatan4177/gpgpu-algorithms-in-games

Both DX11.2 and HSA for Windows will lower the API overheads.

But most of the API overhead seems to effect the CPU.

http://www.dsogaming.com/news/directx-11-2-revealed-available-only-on-windows-8-1-xbox-one/

Low latency present API. Basically api support for the system to tell the app when its the best time to start rendering and showing the content on screen, to allow the shortest latency possible. By using this api they were able to reduce latency from 3 frames (at 60fps) to less than 1.

AMD's HSA solution also reduces compute GpGPU latency returns to less than 16 ms i.e. 60 ms latency = four frame latency.

Both AMD and MS are tacking the latency issues on the PC.

 

5870's stream processors are not symmetrical i.e. 4 simple and 1 complex within VLIW5 group. You can't compare Radeon HD VLIW's FLOPs with Radeon HD GCNs.

 

 

ronvalencia

Irrelevant that is an advantage of the 7000 series over it,is one of the reason is more refine and efficient,which is my whole point new hardware is more efficient than old one.

[QUOTE="tormentos"]

 

You are as dumb as Ronvalencia HSA is a design,without hUMA is not true HSA,so tell me what PC has hUMA now.?

 

There are more games on consoles been done now in HSA maner,than those on PC which i suspect are zero.

ronvalencia

The dummy is you. I gave you hints with PCI-E version 3.0 16 lanes being used as "CPU I/O bus" before I'll post a practical example of a HSA enabled PC with a discrete Radeon HD 7800 and no copy process.

Do you have practical programming on AMD HSA?

Your "than those on PC which i suspect are zero" statement is a joke.

Standalone GPUs can benefit from HSA, but not to the extent of an APU.  A dedicated GPU will have its own attached memory as well as shared memory with the main CPU.  Due to the communication latency issues of writing from main memory to the video cards memory, it is not nearly as seamless as what an APU can accomplish.  It makes sense that such a setup would benefit a solution with a shared memory pool as well as a shared memory controller.  Everything else involves more latency and differing amounts and types of memory.

http://www.pcper.com/reviews/Processors/AMD-Details-hUMA-HSA-Action

OWNED....

Now this come from AMD it self have a nice day.

[QUOTE="ronvalencia"]

[QUOTE="RyviusARC"]

 

But the PS3 has the Cell to offload some of the work to and is capable of handling shader intensive games because of this.

The 7800GTX lacks unified shaders so the comparison is not exactly a fair one.

On non shader intensive games the 7800GTX easily beats the PS3.

The PS4 is not even out so we have no way of knowing how well it can perform and how much HSA will help in performance.

RyviusARC

More info on HSA, DirectCompute issues PC games from http://www.slideshare.net/zlatan4177/gpgpu-algorithms-in-games

Both DX11.2 and HSA for Windows will lower the API overheads.

But most of the API overhead seems to effect the CPU.

 

:lol: its not true HSA without HUMA? HUMA is for APU's with a shared memory pool,  With dedicated memory for the cpu and gpu as Ron has stated they will use the PCI-E lane to allow the gpu and cpu to communicate. AMD has partially resident textures aka PRT with all GCN based gpu's which means that even if a gpu has less memory then console version is allocating PRT stream's small bits of these massive textures into the GPU as needed, giving compatible games a virtually endless supply of  texture data it can apply to the game world without over taxing the buffer because of repeat textures.

04dcarraher

Standalone GPUs can benefit from HSA, but not to the extent of an APU.  A dedicated GPU will have its own attached memory as well as shared memory with the main CPU.  Due to the communication latency issues of writing from main memory to the video cards memory, it is not nearly as seamless as what an APU can accomplish.  It makes sense that such a setup would benefit a solution with a shared memory pool as well as a shared memory controller.  Everything else involves more latency and differing amounts and types of memory.

http://www.pcper.com/reviews/Processors/AMD-Details-hUMA-HSA-Action

This is a devastating blow to yours and Ronvalencias theory.

A damn drive will not make GDDR5 memory and DDR3 memory work as unified memory over a PCI-e Bus that is ridiculous,why do you think HSA is not on PC yet.? If it was that easy AMD would have release a driver and problem solve  millions of PC are already HSA.

HSA is a design not something you enable with a driver.

HSA creates an improved processor design that exposes the benefits and capabilities of mainstream programmable compute elements, working together seamlessly. With HSA, applications can create data structures in a single unified address space and can initiate work items on the hardware most appropriate for a given task. Sharing data between compute elements is as simple as sending a pointer. Multiple compute tasks can work on the same coherent memory regions, utilizing barriers and atomic memory operations as needed to maintain data synchronization (just as multi-core CPUs do today).

http://developer.amd.com/resources/heterogeneous-computing/what-is-heterogeneous-system-architecture-hsa/

If there were any more doutbs..

HSA is not something you turn up with a drive,a driver will not magically change how hardware is strocture on PC in fact do you know what HSA stand for.?

Heterogeneous System Architecture.

[QUOTE="ronvalencia"]

[QUOTE="tormentos"]

You are as dumb as Ronvalencia HSA is a design,without hUMA is not true HSA,so tell me what PC has hUMA now.?

There are more games on consoles been done now in HSA maner,than those on PC which i suspect are zero.

tormentos

The dummy is you. I gave you hints with PCI-E version 3.0 16 lanes being used as "CPU I/O bus" before I'll post a practical example of a HSA enabled PC with a discrete Radeon HD 7800 and no copy process.

Do you have practical programming on AMD HSA?

Your "than those on PC which i suspect are zero" statement is a joke.

Standalone GPUs can benefit from HSA, but not to the extent of an APU. A dedicated GPU will have its own attached memory as well as shared memory with the main CPU. Due to the communication latency issues of writing from main memory to the video cards memory, it is not nearly as seamless as what an APU can accomplish. It makes sense that such a setup would benefit a solution with a shared memory pool as well as a shared memory controller. Everything else involves more latency and differing amounts and types of memory.

http://www.pcper.com/reviews/Processors/AMD-Details-hUMA-HSA-Action

OWNED....

Now this come from AMD it self have a nice day.

What's OWNED about it? That text was from the author's article.

You are not understanding the 3rd party HSA app was minimising DDR3 memory usage, hence no copy process for the Radeon HD 7800.

Also under HSA, CPU has direct access (read/write) to GPU's memory locations i.e. PCI-E connection is acting like PS4's CPU I/O bus i.e. you wouldn't have "writing from main memory to the video cards memory".

The statement "writing from main memory to the video cards memory" is a copy process and it's in conflict with 3rd party HSA Radeon HD 7800 example.

[QUOTE="RyviusARC"]

[QUOTE="ronvalencia"]

More info on HSA, DirectCompute issues PC games from http://www.slideshare.net/zlatan4177/gpgpu-algorithms-in-games

Both DX11.2 and HSA for Windows will lower the API overheads.

 

savagetwinkie

But most of the API overhead seems to effect the CPU.

That is if the CPU is bottlenecking it.

 

What's OWNED about it? That text was from the author's article.

 

You are not understanding the 3rd party HSA app was minimising DDR3 memory usage, hence no copy process for the Radeon HD 7800.

Also under HSA, CPU has direct access (read/write) to GPU's memory locations i.e. PCI-E connection is acting like PS4's CPU I/O bus i.e. you wouldn't have "writing from main memory to the video cards memory".

The statement "writing from main memory to the video cards memory" is a copy process and it's in conflict with 3rd party HSA Radeon HD 7800 example.

 

ronvalencia

""AMD Details hUMA: HSA in Action!""

From the same link.

Keep the denial going,this was from AMD presentation..:lol:

Clear as day,only you in your magical world woudl think that a driver for HSA will actually make GDDR5 and DDR3 in 2 different polls work as hUMA over a PCI-e bus..

My reaction to arguments over technical specs:

[QUOTE="ronvalencia"]

What's OWNED about it? That text was from the author's article.

You are not understanding the 3rd party HSA app was minimising DDR3 memory usage, hence no copy process for the Radeon HD 7800.

Also under HSA, CPU has direct access (read/write) to GPU's memory locations i.e. PCI-E connection is acting like PS4's CPU I/O bus i.e. you wouldn't have "writing from main memory to the video cards memory".

The statement "writing from main memory to the video cards memory" is a copy process and it's in conflict with 3rd party HSA Radeon HD 7800 example.

tormentos

""AMD Details hUMA: HSA in Action!""

From the same link.

Keep the denial going,this was from AMD presentation..:lol:

Clear as day,only you in your magical world woudl think that a driver for HSA will actually make GDDR5 and DDR3 in 2 different polls work as hUMA over a PCI-e bus..

It doesn't say much. LOL. The 3rd party HSA app on Radeon HD 7800 is a practical example in action.

You can avoid the copy process if you don't write to main memory in the first place.

You allocate the required job to the right processor in the first place.

-------------

The presentation is correct if you write data on the main memory and then transfer it to the GPU.

The presentation is not applicable if you directly write to GPU's main memory.

[QUOTE="savagetwinkie"][QUOTE="RyviusARC"]

 

But most of the API overhead seems to effect the CPU.

RyviusARC

That is if the CPU is bottlenecking it.

[QUOTE="tormentos"]

[QUOTE="ronvalencia"]

What's OWNED about it? That text was from the author's article.

You are not understanding the 3rd party HSA app was minimising DDR3 memory usage, hence no copy process for the Radeon HD 7800.

Also under HSA, CPU has direct access (read/write) to GPU's memory locations i.e. PCI-E connection is acting like PS4's CPU I/O bus i.e. you wouldn't have "writing from main memory to the video cards memory".

The statement "writing from main memory to the video cards memory" is a copy process and it's in conflict with 3rd party HSA Radeon HD 7800 example.

ronvalencia

""AMD Details hUMA: HSA in Action!""

From the same link.

Keep the denial going,this was from AMD presentation..:lol:

Clear as day,only you in your magical world woudl think that a driver for HSA will actually make GDDR5 and DDR3 in 2 different polls work as hUMA over a PCI-e bus..

It doesn't say much. LOL. The 3rd party HSA app on Radeon HD 7800 is a practical example in action.

You can avoid the copy process if you don't write to main memory in the first place.

You allocate the required job to the right processor in the first place.

-------------

The presentation is correct if you write data on the main memory and then transfer it to the GPU.

The presentation is not applicable if you directly write to GPU's main memory.

[QUOTE="savagetwinkie"][QUOTE="RyviusARC"]

 

But most of the API overhead seems to effect the CPU.

RyviusARC

That is if the CPU is bottlenecking it.

Even with the extra DirectX overhead, CPUs are so much more powerful on the PC than on consoles so a decent CPU rarely bottlenecks games. And if they do, it's usually because the game is poorly optimized for multi-core processors. Case in point: Assassin's Creed 3, which only uses one core.

The fact that next gen consoles have weak CPUs comprised of 8 Jaguar cores (the same as are being used in tablets and low power netbooks and notebooks) means that in order for developers to get decent performance out of them they are going to be forced to start multi-threading their game engines to make use of multiple cores. This means that the CPUs already available on the PC will start to be better utilized, again negating any CPU advantage the consoles might have due to less overhead.

ITT people pretending to know more than carmack.

[QUOTE="ronvalencia"]

[QUOTE="tormentos"]

""AMD Details hUMA: HSA in Action!""

From the same link.

Keep the denial going,this was from AMD presentation..:lol:

Clear as day,only you in your magical world woudl think that a driver for HSA will actually make GDDR5 and DDR3 in 2 different polls work as hUMA over a PCI-e bus..

savagetwinkie

It doesn't say much. LOL. The 3rd party HSA app on Radeon HD 7800 is a practical example in action.

You can avoid the copy process if you don't write to main memory in the first place.

You allocate the required job to the right processor in the first place.

-------------

The presentation is correct if you write data on the main memory and then transfer it to the GPU.

The presentation is not applicable if you directly write to GPU's main memory.

WTF? HSA enables direct access to GPU's memory space by maping the it's memory with the host CPU's virtual address space. This enables 64bit pointers to be use across the linear address space and direct access to the GPU.

The "Either way you still have to copy memory over the PCI-e to get it ot the GPU" statement is not true for all cases i.e. copying data between different memory spaces is NOT necessary.

PC's PCI-E bus can act like PS4's CPU I/O bus.

http://fabricengine.com/2012/07/gpu_computation_technology_preview/

This 3rd party app running on beta HSA on Radeon HD 7800.

The AMD HSA technology platform has the goal of providing a heterogeneous computation platform in which both CPU and GPU cores access and manipulate memory identically. HSA will enable complex data structures with pointer indirection to be shared between the CPU and GPU. Not only will no copying of data between different memory spaces be necessary, but the pointers imbedded in a complex data structure will be usable without change on both CPU and GPU cores.

In collaboration with AMD, the Fabric Engine development team has extended the KL compiler and Fabric Engine Core execution environment to support GPU computation on high-end AMD GPUs. The primary means by which this preliminary work was possible was the availability of an LLVM back end for AMD GPU hardware.

....

The animated scene was run on a workstation with an AMD A10-5800K APU with both integrated graphics and a discrete Radeon HD 7800 card; however, only the discrete card was used for GPU computation and OpenGL rendering for these tests.

You are assuming writing to CPU's memory space is a necessary requirement i.e. your POV breaks the purpose for direct GPU memory access.

You can stay with the current Direct3D and old driver model for in-direct access to GPU's memory space, while Intel (e.g. Instant Access) , AMD (e.g. HSA) and NVIDIA (e.g. DirectGPU) works out a method to have the real DirectX.

DirectX is an oxymoron from having direct access to the hardware.

ITT people pretending to know more than carmack.

painguy1

http://www.pcgamer.com/2013/02/20/crytek-next-gen-consoles-pc/

"Crytek: Its impossible for next-gen consoles to compete with PCs"

Crytek's Cevat Yerli > ID's John Carmack.

The real question is how much was did Sony have to pay this struggling "legend" to say that?

[QUOTE="savagetwinkie"][QUOTE="ronvalencia"]

It doesn't say much. LOL. The 3rd party HSA app on Radeon HD 7800 is a practical example in action.

You can avoid the copy process if you don't write to main memory in the first place.

You allocate the required job to the right processor in the first place.

-------------

The presentation is correct if you write data on the main memory and then transfer it to the GPU.

The presentation is not applicable if you directly write to GPU's main memory.

ronvalencia

WTF? HSA enables direct access to GPU's memory space by maping the it's memory with the host CPU's virtual address space. This enables 64bit pointers to be use across the linear address space and direct access to the GPU.

The "Either way you still have to copy memory over the PCI-e to get it ot the GPU" statement is not true for all cases i.e. copying data between different memory spaces is NOT necessary.

PC's PCI-E bus can act like PS4's CPU I/O bus.

http://fabricengine.com/2012/07/gpu_computation_technology_preview/

This 3rd party app running on beta HSA on Radeon HD 7800.

The AMD HSA technology platform has the goal of providing a heterogeneous computation platform in which both CPU and GPU cores access and manipulate memory identically. HSA will enable complex data structures with pointer indirection to be shared between the CPU and GPU. Not only will no copying of data between different memory spaces be necessary, but the pointers imbedded in a complex data structure will be usable without change on both CPU and GPU cores.

In collaboration with AMD, the Fabric Engine development team has extended the KL compiler and Fabric Engine Core execution environment to support GPU computation on high-end AMD GPUs. The primary means by which this preliminary work was possible was the availability of an LLVM back end for AMD GPU hardware.

....

The animated scene was run on a workstation with an AMD A10-5800K APU with both integrated graphics and a discrete Radeon HD 7800 card; however, only the discrete card was used for GPU computation and OpenGL rendering for these tests.

You are assuming writing to CPU's memory space is a necessary requirement i.e. your POV breaks the purpose for direct GPU memory access.

You can stay with the current Direct3D and old driver model for in-direct access to GPU's memory space, while Intel (e.g. Instant Access) , AMD (e.g. HSA) and NVIDIA (e.g. DirectGPU) works out a method to have the real DirectX.

DirectX is an oxymoron from having direct access to the hardware.

ITT people pretending to know more than carmack.

painguy1

 

It doesn't say much. LOL. The 3rd party HSA app on Radeon HD 7800 is a practical example in action.

 

You can avoid the copy process if you don't write to main memory in the first place.

You allocate the required job to the right processor in the first place.

-------------

 

The presentation is correct if you write data on the main memory and then transfer it to the GPU.

 

The presentation is not applicable if you directly write to GPU's main memory.

ronvalencia

Standalone GPUs can benefit from HSA, but not to the extent of an APU.  A dedicated GPU will have its own attached memory as well as shared memory with the main CPU.  Due to the communication latency issues of writing from main memory to the video cards memory, it is not nearly as seamless as what an APU can accomplish.  It makes sense that such a setup would benefit a solution with a shared memory pool as well as a shared memory controller.  Everything else involves more latency and differing amounts and types of memory.

http://www.pcper.com/reviews/Processors/AMD-Details-hUMA-HSA-Action

OWNED....

Now this come from AMD it self have a nice day.

Read again and again and again until you understand that HSA more than anything is a design.