Next Gen Console discussion ( PS4 'n Xbox One'n Wii U )

[Extreme Gamer]

Read properly. Amd was the pioneer of developing the "HSA" with its APU and making it more mainstream. Any system architecture that has many cores (dissimilar in nature or having vector + scalar cores) is an heterogeneous architecture. The word "HSA" is derived from the founding members starting with ARM, AMD, QUALCOMM, SAMSUNG, TEXAS INSTRUMENTS,IMAGINATION AND MEDIATEK. Its not constrained with amd or a nomenclature like you think.

HSA Explained: How Heterogenous System Architecture will improve computing - GPU Science

Maxwell will have ARM 57 cores with general nvidia gpu cores. They will be "HSA" even though there is no AMD here. Cell is an heterogeneous unit and in today's terms, it can be said it complies with HSA model.

Don't make a laughing stock of yourself again mate.

Where is the first class GPU in cell? SPEs can hardly be called a first class GPU. Cell is primarily a CPU with heterogenous compute capabilities. Apart from that, what about the unified memory pool (not the same as unified memory access control/fully coherent DMA)?

The name HSA does not derive from the founding members. The name was coined by them. To use the name HSA, Cell has to conform to the HSA standard by getting approval from the HSA foundation.

If you are developing a product that implements the HSA API and HSAIL Finalizer you will need to submit your product for compliance testing before you can use the HSA name or logo with your product or call your product compliant or conformant with HSA Specifications. This conformance requirement will ensure consistency across implementing products for developers and their applications.

The program is in the process of formal creation, so please check back to get more details on how to validate your product(s) in the future.

Read the HSA's IOMMUv2 specification and realize how mistaken you are. Cell does not even comply with this, let alone the rest of the spec. I don't think you bothered to read post #445 properly. If you did, re-read it.

If the Maxwell architecture complies with the HSA standard and the foundation approves of it, I will agree that it is an HSA design. Until the full HSA spec sheet is out, you can't say for sure that it complies.

You're the one who is on the way to becoming a laughing stock, with your post stating that "HSA is programming model" you really showed your true colours. The college you studied in should be blacklisted if this is the BS they spew. HSA is strictly a hardware design complying with the heterogenous compute hardware design principle. Rather, it is the programming model that develops around the hardware design to take advantage of said hardware design.

For the lazybones, this is what vickybat wrote:

HSA is nothing but a programming model and of course it complies with heterogeneous computing.

 

[Cilus]

*domino.research.ibm.com/comm/research.nsf/8be45dd2101a4d5e85256aee006d6b6a/d87d07769001062585256ffd0046bea0/body_main/0.A16!OpenElement&FieldElemFormat=gif

Extreme Gamer, here you will find the Unified Memory Access through EIB. And SPU can directly fetch data from memory without any help of PPU.

HSA is the term used by AMD to describe to the Heterogeneous Computing Model in such a way that Programming Models can be easily implemented to take benefit of Heterogeneous system and for this the Foundation is trying to specify some universal standards for its implementation. It does not mean that it is invented by AMD only. Any heterogeneous system which use a Unified Memory architecture and an instruction set architecture (ISA) which enables to dispatch tasks dynamically to LCU (Legacy Compute Unit) which is the conventional CPU and TCU (Throughput Compute Engine) or parallel Processing unit. For mainstreaming it, AMD has used couple of nomenclatures like LCU, TCU, Wavefront and Work Items and defines their property under HSA implementation. But this does not mean that they weren't there before that, just were not known by these names.
Cell processor does use an unified 256 bit memory bus, known as Elemental Interconnect Bus EIB) which is directly connected to both SPU and PPU. Similarly, both PPU and SPU use a subset of 64 bit Power VMX instruction set, resulting of same type of job queue for both SPE and PPU. So definitely it is a HSA design, you believe it or not.

 

[vickybat]

Where is the first class GPU in cell? SPEs can hardly be called a first class GPU. Cell is primarily a CPU with heterogenous compute capabilities. Apart from that, what about the unified memory pool (not the same as unified memory access control/fully coherent DMA)?

The name HSA does not derive from the founding members. The name was coined by them. To use the name HSA, Cell has to conform to the HSA standard by getting approval from the HSA foundation.



Read the HSA's IOMMUv2 specification and realize how mistaken you are. Cell does not even comply with this, let alone the rest of the spec. I don't think you bothered to read post #445 properly. If you did, re-read it.

If the Maxwell architecture complies with the HSA standard and the foundation approves of it, I will agree that it is an HSA design. Until the full HSA spec sheet is out, you can't say for sure that it complies.

You're the one who is on the way to becoming a laughing stock, with your post stating that "HSA is programming model" you really showed your true colours. The college you studied in should be blacklisted if this is the BS they spew. HSA is strictly a hardware design complying with the heterogenous compute hardware design principle. Rather, it is the programming model that develops around the hardware design to take advantage of said hardware design.

For the lazybones, this is what vickybat wrote:

Everybody knows here what you do. And you learn to mind your language from now on.

This isn't a journalism forum. Put proper facts in a sane manner. By programming model, i meant cuda, open-cl, openmp, c++AMP etc. Without them, HSA makes no sense.
No sense at all. You being a CS student should understand this. Personal insults aren't tolerated and i'm reporting your post for insulting my college.
Btw i'm not even a CS graduate but Electronics and telecommunications. You should be inherently knowing these things better than me. But i suppose that's not the case.

Its not about the architecture only. You have to program accordingly so that gpu or any vector processor can perform general operations. I gave you example before. Re-read them.
A programming model is what makes HSA successful today. Or else all those gpu's will be doing rendering only using graphics api's. If you indeed are a CS student, you will understand this now
and won't make meaningless posts.

There you go, Cilus just explained beautifully why cell is an HSA implementation in pure standards. Stop baseless arguments.

 

[heidi2521]

*domino.research.ibm.com/comm/resea...dy_main/0.A16!OpenElement&FieldElemFormat=gif


Extreme Gamer, here you will find the Unified Memory Access through EIB. And SPU can directly fetch data from memory without any help of PPU.

HSA is the term used by AMD to describe to the Heterogeneous Computing Model in such a way that Programming Models can be easily implemented to take benefit of Heterogeneous system and for this the Foundation is trying to specify some universal standards for its implementation. It does not mean that it is invented by AMD only. Any heterogeneous system which use a Unified Memory architecture and an instruction set architecture (ISA) which enables to dispatch tasks dynamically to LCU (Legacy Compute Unit) which is the conventional CPU and TCU (Throughput Compute Engine) or parallel Processing unit. For mainstreaming it, AMD has used couple of nomenclatures like LCU, TCU, Wavefront and Work Items and defines their property under HSA implementation. But this does not mean that they weren't there before that, just were not known by these names.
Cell processor does use an unified 256 bit memory bus, known as Elemental Interconnect Bus EIB) which is directly connected to both SPU and PPU. Similarly, both PPU and SPU use a subset of 64 bit Power VMX instruction set, resulting of same type of job queue for both SPE and PPU. So definitely it is a HSA design, you believe it or not.

Dude, Heterogenous Computing != HSA. Heterogenous computing is just an idea. HSA is a full fledged specification. The HSA foundation is not trying to define what heterogenous computing is, just writing and maintaining the HSA spec.

Refer to post *www.thinkdigit.com/forum/consoles/...ion-ps4-n-xbox720-n-wii-u-16.html#post1903046 to understand why Cell is not HSA compliant. Reading the IOMMU v2 spec would also help.

 

[Extreme Gamer]

HSA is the term used by AMD to describe to the Heterogeneous Computing Model in such a way that Programming Models can be easily implemented to take benefit of Heterogeneous system and for this the Foundation is trying to specify some universal standards for its implementation. It does not mean that it is invented by AMD only. Any heterogeneous system which use a Unified Memory architecture and an instruction set architecture (ISA) which enables to dispatch tasks dynamically to LCU (Legacy Compute Unit) which is the conventional CPU and TCU (Throughput Compute Engine) or parallel Processing unit. For mainstreaming it, AMD has used couple of nomenclatures like LCU, TCU, Wavefront and Work Items and defines their property under HSA implementation. But this does not mean that they weren't there before that, just were not known by these names.
Cell processor does use an unified 256 bit memory bus, known as Elemental Interconnect Bus EIB) which is directly connected to both SPU and PPU. Similarly, both PPU and SPU use a subset of 64 bit Power VMX instruction set, resulting of same type of job queue for both SPE and PPU. So definitely it is a HSA design, you believe it or not.

The foundation is trying to specify universal standards for HSA, not for heterogenous computing. Please, try to distinguish between the two. Heterogenous computing is a design idea, that different kinds of processors can be used on the same chip. Heterogenous computing does not define that memory has to be unified for the entire chip. The different components can use their own chips.

Also, the fact that there is no first class GPU on the Cell does not change. Like I said, Cell is definitely heterogenous compute compliant, but it is not HSA compliant. What about remapping addresses for devices that do not support 64-bit addressing, providing page granularity control of device access to system memory? This is all from the IOMMUv2 spec.

Look at page 63 of the IOMMUv2 spec PDF. Does the Cell comply with that? Similarities of Cell with HSA are superficial at best and nonexistent at worst.

Everybody knows here what you do. And you learn to mind your language from now on.

This isn't a journalism forum. Put proper facts in a sane manner. By programming model, i meant cuda, open-cl, openmp, c++AMP etc. Without them, HSA makes no sense.
No sense at all. You being a CS student should understand this. Personal insults aren't tolerated and i'm reporting your post for insulting my college.
Btw i'm not even a CS graduate but Electronics and telecommunications. You should be inherently knowing these things better than me. But i suppose that's not the case.

Its not about the architecture only. You have to program accordingly so that gpu or any vector processor can perform general operations. I gave you example before. Re-read them.
A programming model is what makes HSA successful today. Or else all those gpu's will be doing rendering only using graphics api's. If you indeed are a CS student, you will understand this now
and won't make meaningless posts.

There you go, Cilus just explained beautifully why cell is an HSA implementation in pure standards. Stop baseless arguments.

How did I insult your college? I said IF, so it looks like you made the BS up yourself. And as a CS student, I do know more than you. It is that obvious.

Even if HSA does not make sense without CUDA, OpenCL etc (which is not necessarily true- without them it is a PITA to program, but it isn't impossible), does not mean that HSA itself becomes a programming model. A programming model might make HSA succesful, but that does not make HSA a programming model itself.

Don't deny that you didn't call HSA a programming model. You indeed did, and dead5 and I just showed that that was completely false.

 

[vickybat]

How did I insult your college? I said IF, so it looks like you made the BS up yourself. And as a CS student, I do know more than you. It is that obvious.

Even if HSA does not make sense without CUDA, OpenCL etc (which is not necessarily true- without them it is a PITA to program, but it isn't impossible), does not mean that HSA itself becomes a programming model. A programming model might make HSA succesful, but that does not make HSA a programming model itself.

Don't deny that you didn't call HSA a programming model. You indeed did, and dead5 and I just showed that that was completely false.

Oh so you know how to code in binary and machine language? Coding is not journalism buddy. General purpose computing was absent before cuda,open-cl and others.
If it was possible, how come nobody did it?? Your posts don't show a single thing that points you knowing more than me. Instead of posting links, go through them properly and put things into perspective. Its not a PITA, its nearly impossible to code in a logical sense. Eventually you have to come with a programming model to work up on a particular architecture. Open-cl, cuda and all are created to code on vector processor part of gpu only. It ignores the rasterizer. X86 can also be considered as a programming model. The instructions sets are nothing but programs that enable execution units to perform operations. Assembly language is the most generic definition.

With programs only, they distinguish between an integer and a float type. FMA is also an instruction that can do simultaneous multiply and add without breaking them into smaller parts. Programming model is like a global term and can be associated with any architecture and not only HSA. You and dead5 i guess aren't smart enough to comprehend this. To comprehend what cilus and i were trying to say from the previous two pages.
Programming models were developed specially to make HSA possible that you see today. Refrain from posting off-topic from now on. You could get into trouble. Keep this as a console thread only.
Enough posting has been done from our side to make you and your brother understand. If you indeed are a CS student, learn to behave and act like one.

 

[heidi2521]

Oh so you know how to code in binary and machine language? Coding is not journalism buddy. General purpose computing was absent before cuda,open-cl and others.
If it was possible, how come nobody did it?? Your posts don't show a single thing that points you knowing more than me. Instead of posting links, go through them properly and put things into perspective. Its not a PITA, its nearly impossible to code in a logical sense. Eventually you have to come with a programming model to work up on a particular architecture. Open-cl, cuda and all are created to code on vector processor part of gpu only. It ignores the rasterizer. X86 can also be considered as a programming model. The instructions sets are nothing but programs that enable execution units to perform operations. Assembly language is the most generic definition.

With programs only, they distinguish between an integer and a float type. FMA is also an instruction that can do simultaneous multiply and add without breaking them into smaller parts. Programming model is like a global term and can be associated with any architecture and not only HSA. You and dead5 i guess aren't smart enough to comprehend this. To comprehend what cilus and i were trying to say from the previous two pages.
Programming models were developed specially to make HSA possible that you see today. Refrain from posting off-topic from now on. You could get into trouble. Keep this as a console thread only.
Enough posting has been done from our side to make you and your brother understand. If you indeed are a CS student, learn to behave and act like one.

LOL Wut?! So you mean to say that the IBM PC, the Amiga, The Commodore etc. were all specialized computers?

Define "Programming Model". None of what I think it could mean makes sense with what you have typed.

I won't bother replying to most of what you have typed because Extreme Gamer is doing a wonderful job of dissecting it.

Oh, and the classic vickybat "Please stay on topic" when you start to lose an argument.

 

[Extreme Gamer]

Oh so you know how to code in binary and machine language? Coding is not journalism buddy. General purpose computing was absent before cuda,open-cl and others.
If it was possible, how come nobody did it?? Your posts don't show a single thing that points you knowing more than me. Instead of posting links, go through them properly and put things into perspective. Its not a PITA, its nearly impossible to code in a logical sense. Eventually you have to come with a programming model to work up on a particular architecture. Open-cl, cuda and all are created to code on vector processor part of gpu only. It ignores the rasterizer. X86 can also be considered as a programming model. The instructions sets are nothing but programs that enable execution units to perform operations. Assembly language is the most generic definition.

With programs only, they distinguish between an integer and a float type. FMA is also an instruction that can do simultaneous multiply and add without breaking them into smaller parts. Programming model is like a global term and can be associated with any architecture and not only HSA. You and dead5 i guess aren't smart enough to comprehend this. To comprehend what cilus and i were trying to say from the previous two pages.
Programming models were developed specially to make HSA possible that you see today. Refrain from posting off-topic from now on. You could get into trouble. Keep this as a console thread only.
Enough posting has been done from our side to make you and your brother understand. If you indeed are a CS student, learn to behave and act like one.

Unlike you, I don't post links without going through them. Obviously coding is not journalism. Everybody knows that even if your fantasies indicate otherwise.

Have you heard of ENIAC? At it's core it was a general-purpose computer. You did have to manually reconfigure (rewire) it every time you changed the tasks and objectives for the system, but that does not take away that fact. General purpose computing didn't exist before OpenCL and CUDA? Really? Have you ever heard of Fortran? It was released ~50 years before anyone even heard of CUDA

Learn this: GPGPU != GPC [general purpose computing] in the same way HSA != Heterogenous computing. GPGPU is an implementation of GPC.

And PITA and Nigh impossible are not mutually exclusive.

Vickybat, I think you should stick to your area of expertise, because an architecture can never be a programming model. Instruction set defines what assembly language can be used to program the back-end (i.e. the compiler) for any programming language, it does not define the architecture alone.

And FYI, don't insult my brother and I with attacks such as we're not smart enough. It is quite obvious who understands better and who does not.

And you need to rephrase "With programs only, they distinguish between an integer and a float type," because your point is not clear and frankly, that sentence is not making any sense.

And you are gravely mistaken when you say "Programming model is like a global term and can be associated with any architecture and not only HSA."

And you were the one who went off-topic. You've started to lose so now you're asking us to go back on topic.

Since it is obvious that you're definition of programming model differs from the globally accepted one, could you please post your definition here?

Another fact: neither CUDA nor OpenCL were developed with HSA in mind. Could you name a single one that was? Because if "programming models were designed to make HSA possible today", you must be knowing quite a few that were designed with HSA in mind.

I am acting as a responsible CS student who is trying to prevent misinformation from spreading, TYVM.

 

[Cilus]

1st of all, stop trying to prove you know it better because you study some particular subject in a famous university. It does not prove anything about our knowledge. Lets take your example. I am sure that you went to a very reputed school and they also teach Physics very well. But that didn't stop you to claim here that you can see changes in 200 Frames per second in games and you see it in your Monitor with the 580 2 GB SLI. Redeemer?
But you had forgotten that as per Physics, Human eye can only see differences up to 75 Frames per second, 80 Frames at max. You had also forgotten the fact that a 60 Hz display can only show 60 frames per second and 200 Frames per second can't be displayed on it. Those claims leaded us to believe that you're some kind of Mutant. But did anybody ask you about your schooling or level of education? No.

So, from now stop mentioning that you know better because you are a computer science student.

Also, the fact that there is no first class GPU on the Cell does not change. Like I said, Cell is definitely heterogenous compute compliant, but it is not HSA compliant. What about remapping addresses for devices that do not support 64-bit addressing, providing page granularity control of device access to system memory? This is all from the IOMMUv2 spec.

I didn't get the point...why should we consider devices which do not support 64 Bit addressing when every component of CELL supports it?? Please explain. AFAIK, both SPE and PPE support Power PC 64 bit instruction and addressing models. But since SPEs are used for some special parallel processing like single precision FPU operations, it only support a subset of complete POWER 64 Instruction set whereas PPE support the complete set. But both support the 64 it memory addressing and general LOAD-STORE instructions. In fact, CELL supports lots of features of HSA which are still not present in current generation architecture. So why I have to consider devices not supporting 64 bit addressing which is non-existence in CELL?

One thing, I agree with you that CELL was never used as a true HSA design because no such software based programming was used to develop applications for CELL. While developing, the developers need to be very careful about how they are scheduling the workloads in a CELL and Cell lacks the hardware support for automatic assignment at runtime.

regarding CUDA and OPenCL, they were designed for GP-GPU programming in mind but can very easily used with minor modifications for HSA computing. That is the main reason that AMD is promoting OPenCL as an universal language for HSA programming. Just specifying some standard in hardware developement make HSA successful, you need to leverage it with supporting programming models and OPENCL is perfect example of it. So even in software field, you need to have a programming model too. For example, current generation Trinity and Llano don't support UMA but still they are considered as HSA unit and that can be implemented through well designed programming. As per you, currently there isn't any HSA chip available as neither of the CPU and GPU do have unified memory access and PST is the only one to do that. Try to get the idea 1st rather than pointing me some slide numbers.

 

[Extreme Gamer]

1st of all, stop trying to prove you know it better because you study some particular subject in a famous university. It does not prove anything about our knowledge. Lets take your example. I am sure that you went to a very reputed school and they also teach Physics very well. But that didn't stop you to claim here that you can see changes in 200 Frames per second in games and you see it in your Monitor with the 580 2 GB SLI. Redeemer?

But you had forgotten that as per Physics, Human eye can only see differences up to 75 Frames per second, 80 Frames at max. You had also forgotten the fact that a 60 Hz display can only show 60 frames per second and 200 Frames per second can't be displayed on it. Those claims leaded us to believe that you're some kind of Mutant. But did anybody ask you about your schooling or level of education? No.

I never claimed that I could distinguish between 200Hz and 150Hz. I said that the "magical" 60hz cap that the folks in that thread had posted was incorrect. The eyes are analogue, not digital with refresh rates etc, light is received continuously in the eyes and no matter the average framerate, if the instantaneous framerate is not constant, the eyes will definitely perceive that feed to be less smooth than one where the instantaneous framerate is constant. Instantaneous framerate is equivalent to the delay between each frame here. Whenever I look at a 200hz video, all I see is a very abnormal video that is not at all pleasing to watch.

So, from now stop mentioning that you know better because you are a computer science student.

After mentioning it once, I did not mention it again until Vickybat brought it up. So please don't be biased against me and see what Vickybat wrote before posting like this.

I didn't get the point...why should we consider devices which do not support 64 Bit addressing when every component of CELL supports it?? Please explain. AFAIK, both SPE and PPE support Power PC 64 bit instruction and addressing models. But since SPEs are used for some special parallel processing like single precision FPU operations, it only support a subset of complete POWER 64 Instruction set whereas PPE support the complete set. But both support the 64 it memory addressing and general LOAD-STORE instructions. In fact, CELL supports lots of features of HSA which are still not present in current generation architecture. So why I have to consider devices not supporting 64 bit addressing which is non-existence in CELL?

The point is Cell being compliant with the HSA spec, which it is not. The current generation architecture is an APU but it is not HSA. Jaguar will be the first. If cell can't do it, it does not conform to the HSA specification.

One thing, I agree with you that CELL was never used as a true HSA design because no such software based programming was used to develop applications for CELL. While developing, the developers need to be very careful about how they are scheduling the workloads in a CELL and Cell lacks the hardware support for automatic assignment at runtime.

Again, HSA and HC are not the same. But we agree that Cell is not HSA

regarding CUDA and OPenCL, they were designed for GP-GPU programming in mind but can very easily used with minor modifications for HSA computing. That is the main reason that AMD is promoting OPenCL as an universal language for HSA programming. Just specifying some standard in hardware developement make HSA successful, you need to leverage it with supporting programming models and OPENCL is perfect example of it. So even in software field, you need to have a programming model too. For example, current generation Trinity and Llano don't support UMA but still they are considered as HSA unit and that can be implemented through well designed programming. As per you, currently there isn't any HSA chip available as neither of the CPU and GPU do have unified memory access and PST is the only one to do that. Try to get the idea 1st rather than pointing me some slide numbers.

True, but that isn't the point. Vickybat said that OpenCL and CUDA were designed to make HSA possible, which is what I'm disagreeing with. HSA was thought up independent of OpenCL or CUDA. Nobody is talking about the success of HSA, so stop diverting the topic please. I know very well what I'm talking about.

Replies in bold.

I don't post links without knowing what is in them. Please actually read what is in them before replying in this manner. It is rude of you to dismiss them without even checking the contents.

And why are you only replying to me without even responding to the utter BS that Vickybat posted? I don't think any software engineer would agree with someone calling a system architecture a programming model. Programming models are developed around sys-archs. So far, I have not posted anything inaccurate or incorrect while Vickybat has gotten a free run. I even bother to post relevant examples in the form of links and those get disregarded. Instead of diverting the directions my points take, why don't you respond directly against them if they're incorrect? So don't go into the success or non-success of HSA because that is not the point being debated.

 

[NoasArcAngel]

w

Replies in bold.
I never claimed that I could distinguish between 200Hz and 150Hz. I said that the "magical" 60hz cap that the folks in that thread had posted was incorrect. The eyes are analogue, not digital with refresh rates etc, light is received continuously in the eyes and no matter the average framerate, if the instantaneous framerate is not constant, the eyes will definitely perceive that feed to be less smooth than one where the instantaneous framerate is constant. Instantaneous framerate is equivalent to the delay between each frame here. Whenever I look at a 200hz video, all I see is a very abnormal video that is not at all pleasing to watch.

The point is Cell being compliant with the HSA spec, which it is not. The current generation architecture is an APU but it is not HSA. Jaguar will be the first. If cell can't do it, it does not conform to the HSA specification.

And why are you only replying to me without even responding to the utter BS that Vickybat posted? I don't think any software engineer would agree with someone calling a system architecture a programming model. Programming models are developed around sys-archs. So far, I have not posted anything inaccurate or incorrect while Vickybat has gotten a free run. I even bother to post relevant examples in the form of links and those get disregarded. Instead of diverting the directions my points take, why don't you respond directly against them if they're incorrect? So don't go into the success or non-success of HSA because that is not the point being debated.

Tell me one thing... at what frames per second do eyes perceive your normal vision? There is nothing as instantaneous framerate. anything above 60fps and the eye can not make out the difference. If you have a 200hz display it means that the screen is being refreshed 200 times a second, doesnt mean that a 60fps video is playing at 200fps.

The cell is not HSA compliant, but that does not mean that it shares the same ideas with the HSA model of computing. If you are going into so much technical detail then x86 is really x86_64.

programming models need not be always designed about sys arch, in the future and even in the past sys arch has been designed about programming models. You dont even bother to read others replies and jump the gun, so why should we treat you differently?

LOL Wut?! So you mean to say that the IBM PC, the Amiga, The Commodore etc. were all specialized computers?

Define "Programming Model". None of what I think it could mean makes sense with what you have typed.

I won't bother replying to most of what you have typed because Extreme Gamer is doing a wonderful job of dissecting it.

Oh, and the classic vickybat "Please stay on topic" when you start to lose an argument.

computers at that time werent specialized, likewise today your HSA compliant pc is not a roadrunner.

The main circuit board in an IBM PC is called the motherboard (IBM terminology calls it a planar). This mainly carries the CPU and RAM, and it has a bus with slots for expansion cards. On the motherboard are also the ROM subsystem, DMA and IRQ controllers, coprocessor socket, sound (PC speaker, tone generation) circuitry, and keyboard interface. The original PC also adds to this the cassette interface.
The bus used in the original PC became very popular, and it was subsequently named ISA. While it was popular, it was more commonly known as the PC-bus or XT-bus; the term ISA arose later when industry leaders chose to continue manufacturing machines based on the IBM PC AT architecture rather than license the PS/2 architecture and its MCA bus from IBM. The XT-bus was then retroactively named 8-bit ISA or XT ISA, while the unqualified term ISA usually refers to the 16-bit AT-bus (as better defined in the ISA specifications.) The AT-bus is an extension of the PC-/XT-bus and is in use to this day in computers for industrial use, where its relatively low speed, 5 volt signals, and relatively simple, straightforward design (all by year 2011 standards) give it technical advantages (e.g. noise immunity for reliability).


Quadram Quadboard.
A monitor and any floppy or hard disk drives are connected to the motherboard through cables connected to graphics adapter and disk controller cards, respectively, installed in expansion slots. Each expansion slot on the motherboard has a corresponding opening in the back of the computer case through which the card can expose connectors; a blank metal cover plate covers this case opening (to prevent dust and debris intrusion and control airflow) when no expansion card is installed. Memory expansion beyond the amount installable on the motherboard was also done with boards installed in expansion slots, and I/O devices such as parallel, serial, or network ports were likewise installed as individual expansion boards. For this reason, it was easy to fill the five expansion slots of the PC, or even the eight slots of the XT, even without installing any special hardware. Companies like Quadram and AST addressed this with their popular multi-I/O cards which combine several peripherals on one adapter card that uses only one slot; Quadram offered the QuadBoard and AST the SixPak.

This is why the ibm pc was indeed specialized in the true form.

oh, and this is not a debate of losing or winning, if it was ... you my friend have lost it even before it began. So lets not try to keep score... or you will end up playing mario kart which is so hardcore ( you are a small kid, no wonder mario kart is so hardcore for you ... also i never heard of valve employing underage kids )

You're trolling, right?

*en.wikipedia.org/wiki/Heterogeneous_computing

*www.thinkdigit.com/forum/attachments/consoles/10487d1368417755-next-gen-console-discussion-ps4-n-xbox720-n-wii-u-lolwiki.jpg

Heterogeneous computing systems present new challenges not found in typical homogeneous systems. The presence of multiple processing elements raises all of the issues involved with homogeneous parallel processing systems, while the level of heterogeneity in the system can introduce non-uniformity in system development, programming practices, and overall system capability. Areas of heterogeneity can include:[7]
ISA or instruction set architecture
Compute elements may have different instruction set architectures, leading to binary incompatibility.
ABI or application binary interface
Compute elements may interpret memory in different ways. This may include both endianness, calling convention, and memory layout, and depends on both the architecture and compiler being used.
API or application programming interface
Library and OS services may not be uniformly available to all compute elements.
Low-Level Implementation of Language Features
Language features such as functions and threads are often implemented using function pointers, a mechanism which requires additional translation or abstraction when used in heterogeneous environments.
Memory Interface and Hierarchy
Compute elements may have different cache structures, cache coherency protocols, and memory access may be uniform or non-uniform memory access (NUMA). Differences can also be found in the ability to read arbitrary data lengths as some processors/units can only perform byte-, word-, or burst accesses.
Interconnect
Compute elements may have differing types of interconnect aside from basic memory/bus interfaces. This may include dedicated network interfaces, Direct memory access (DMA) devices, mailboxes,

Dead5... my god...


@ extreme_gamer you have been banned once already for posting on the hardware thread. Do you want another ban?


*lup.lub.lu.se/luur/download?func=downloadFile&recordOId=1890491&fileOId=1890492

The Cell Broadband Engine Architecture (CBEA) is a
heterogeneous processor architecture, originally developed for
the PlayStation 3TM by Sony, Toshiba and IBM. The processor
in the PlayStation 3TM is equipped with one general-purpose
PowerPC Processor Unit (PPU) and seven1 Synergistic Processor Units (SPU) [4].

I'm just gonna' just gonna point out that heterogenous computing != HSA and let y'all argue.

HSA is but a subset of heterogenous computing.

or combining scalar processing on the CPU with parallel processing on the GPU while enabling high bandwidth access to memory and high application performance at low power consumption. HSA defines interfaces for parallel computation utilizing CPU, GPU, and other programmable and fixed function devices, and support for a diverse set of high-level programming languages, thereby creating the next foundation in general purpose computing.

Heterogeneous computing may not be HSA in the true sense but the roots of HSA have been derived from this. Hence 90% of the functionality of the HSA is basically a heterogeneous computer running some standardized protocols. This is what HSA is .

Extreme_Gamer are you contradicting your own brother?

Where is the first class GPU in cell? SPEs can hardly be called a first class GPU. Cell is primarily a CPU with heterogenous compute capabilities. Apart from that, what about the unified memory pool (not the same as unified memory access control/fully coherent DMA)?

The name HSA does not derive from the founding members. The name was coined by them. To use the name HSA, Cell has to conform to the HSA standard by getting approval from the HSA foundation.



Read the HSA's IOMMUv2 specification and realize how mistaken you are. Cell does not even comply with this, let alone the rest of the spec. I don't think you bothered to read post #445 properly. If you did, re-read it.

If the Maxwell architecture complies with the HSA standard and the foundation approves of it, I will agree that it is an HSA design. Until the full HSA spec sheet is out, you can't say for sure that it complies.

You're the one who is on the way to becoming a laughing stock, with your post stating that "HSA is programming model" you really showed your true colours. The college you studied in should be blacklisted if this is the BS they spew. HSA is strictly a hardware design complying with the heterogenous compute hardware design principle. Rather, it is the programming model that develops around the hardware design to take advantage of said hardware design.

For the lazybones, this is what vickybat wrote:

and if you did not know, the ps3 rsx gpu is nothing but a 7800 series gpu from nvidia. that does qualify as a gpu doesnt it? or now we have to show you that the rsx is a gpu and ......

cell does comply with HSA, not in the true sense but with a bit of work you can get HSA to run on the ps3

*www.ll.mit.edu/HPEC/agendas/proc06/Day3/05a_Steinsaltz_Pres.ppt

^ these guys show you how .

HSA is a programming model. It always was it always will be, thats like saying that open gl is not a programming model because you had AGP and PCI express cards back in the day.


*www.thinkdigit.com/forum/attachments/consoles/10490d1368427381-next-gen-console-discussion-ps4-n-xbox720-n-wii-u-omggg.jpg

Extreme_Gamer this is really the limit, are you typing with your eyes closed ? or has the hacker taken over your system once again?

and WAN intrusions? REALLY? !

 

[Cilus]

Regarding your post of education level, why to post it in the 1st place?
2ndly, I still have the posts where you have mentioned that you can see the difference in 200FPS, saved it for some fun. DOn't mind. No cap and other stuffs were mentyioned over there. And if the instantaneous FPS is over 85 Frames per second and above, human eye hardly finds any difference.

 

[NoasArcAngel]

Regarding your post of education level, why to post it in the 1st place?
2ndly, I still have the posts where you have mentioned that you can see the difference in 200FPS, saved it for some fun. DOn't mind. No cap and other stuffs were mentyioned over there. And if the instantaneous FPS is over 85 Frames per second and above, human eye hardly finds any difference.

1 second = 1000 mili seconds

shortest notable dark period = 16 ms which implies if you have 85fps. = 1000 / 85 = 11.76 ms . Therefor anything over 60fps ( 1000/60 ) wont matter















Extreme_Gamer enjoy your vacation

 

[vickybat]

Spoiler

Unlike you, I don't post links without going through them. Obviously coding is not journalism. Everybody knows that even if your fantasies indicate otherwise.

Facepalm won't be your saving grace buddy. You really have no clue on anything, let alone computing.

Have you heard of ENIAC? At it's core it was a general-purpose computer. You did have to manually reconfigure (rewire) it every time you changed the tasks and objectives for the system, but that does not take away that fact. General purpose computing didn't exist before OpenCL and CUDA? Really? Have you ever heard of Fortran? It was released ~50 years before anyone even heard of CUDA

Learn this: GPGPU != GPC [general purpose computing] in the same way HSA != Heterogenous computing. GPGPU is an implementation of GPC.

Take up journalism coz that suits you. No wonder your posts are a far cry of anything that's remotely technical. Why the hell are you mentioning ENIAC in this context lol.
Fortran was one of the first imperative languages ( i hope you understand what imperative means, if you have any experience with programming) developed to perform scientific calculations from a high level perspective, ideally for ibm 703. The idea was to avoid coding in assembly language. Still till this day, fortran is meant for general purpose computing for HPC applications that use a grid layout of computers or rather supercomputers. The correct term would be grid or distributed computing. Its still used for X86 architecture rather than vector processors. Like i said before, fortran just like any other high level language is for complex instructions and not used for gpgpu. It supports vector data type but isn't used in GPGPU which you confuse with GPC ( general purpose computing). High performance computing and parallel computing are different. GPGPU has to use a special programming model to harness the SIMD type architecture vector processors have. Multicore processors and vector processors are different. Fortran is used for the former. You are seriously confused. And btw, HSA is heterogeneous computing only.

Again foolish of you for bringing fortran for no reason.

And PITA and Nigh impossible are not mutually exclusive.
Vickybat, I think you should stick to your area of expertise, because an architecture can never be a programming model. Instruction set defines what assembly language can be used to program the back-end (i.e. the compiler) for any programming language, it does not define the architecture alone.
And FYI, don't insult my brother and I with attacks such as we're not smart enough. It is quite obvious who understands better and who does not.
And you need to rephrase "With programs only, they distinguish between an integer and a float type," because your point is not clear and frankly, that sentence is not making any sense.

This post begins with another journalist like sentence.
Btw, i do stick with my area of expertise and you already are starting to feel it don't you. A programming model defines an architecture, any architecture whatsover. Without it, no architecture is going to do anything. Think it technically, not like a journalist who tries to find words from a dictionary all the time. Lady ada lovelace ( the first computer programmer in the world), devised an algorithm to perform general purpose scientific calculations on the "Analytical Engine" created by "Charles Babbage". Without her work, the analytical engine would have been nothing but a pile of junk. Every programming model forms the backbone of an architecture. Disagreeing with this statement, you prove that you have no programming knowledge at all.

And btw, nobody tried to insult you or your brother. Both of you made fools out of yourselves coz of your own misdoings.

And you are gravely mistaken when you say "Programming model is like a global term and can be associated with any architecture and not only HSA."
And you were the one who went off-topic. You've started to lose so now you're asking us to go back on topic.
Since it is obvious that you're definition of programming model differs from the globally accepted one, could you please post your definition here?
Another fact: neither CUDA nor OpenCL were developed with HSA in mind. Could you name a single one that was? Because if "programming models were designed to make HSA possible today", you must be knowing quite a few that were designed with HSA in mind.

I am acting as a responsible CS student who is trying to prevent misinformation from spreading, TYVM.

Oh yeah i'm mistaken. lol Go back to the study table and try to learn things and not copy definitions from the web. You have a history of doing so if i remember.
And wait, i'm losing to you?? That's the biggest understatement of the century. Go back and reading what programming is all about. HSA is too deep for your current state of mind to imbibe anything. You can't even differentiate between parallel and grid computing. Arguing with you is not only a waste of time for me but everyone in the forum. And, you are anything but a responsible CS student. The title is earned, not self proclaimed. lol

CUDA and open-cl were designed with GPGPU in mind and to harness vector processing capability of GPU's. Ultimately they are the founding pillars of HSA and form the programming model for HSA along with X86 model standards. They work in unity and not as separate entities.

Replies in bold.I hope this ends here. Enjoy your vacation and comeback like a responsible person.

@Thread

Lets keep this thread clean guys and prepare ourselves for the advent of next-gen consoles starting from the xbox next reveal on may 21st and ps4 at E3.
I hope all interested members will show up here with all sorts of news and materials related to next gen consoles and what they bring in store. Will try my best to post the entire coverage here.

Time for some news:

Nordic retailer lists Gran Turismo 6 for PS4

Wii U goes mobile

This could be interesting and a different turn on things.

PS4 Watch Dogs is 'a simulation that the current generation doesn't have the processing power to do'

 

[NoasArcAngel]

me too, will try to update this thread as soon as possible.

 

[RCuber]

I thought tdf changed ..

 

[NoasArcAngel]

I thought tdf changed ..

it did. Tdf is a stage, where some people play their part and depart.

 

[darkv0id]

^ Stuff like this used to happen before?

 

[NoasArcAngel]

^ Stuff like this used to happen before?

you mean ?....

 

[vickybat]

Sony Said It Never Considered 'Always On' for PlayStation 4

Fingers crossed on Microsoft's decision.

Wii U Does Not Support EA's Frostbite Engine