Desktops that will have Cell?
- Thread starter Giraffe
- Start date
-
- Tags
- Power Supplies
Action_Man
Splendid
- Jan 7, 2004
- 3,857
- 0
- 22,780
xxsk8er101xx
Splendid
- Dec 31, 2007
- 3,967
- 4
- 22,815
mpjesse
Splendid
- Dec 31, 2007
- 7,214
- 0
- 25,810
a processor is only as good as it's compiler.
in other words, no. programmers won't take the time to learn how to write new machine language. nor should they... AMD and Intel are doing just fine competing against one another. the last thing we need is another new processor.
the current generation and variation of cell will never see the light of day on a desktop.
in other words, no. programmers won't take the time to learn how to write new machine language. nor should they... AMD and Intel are doing just fine competing against one another. the last thing we need is another new processor.
the current generation and variation of cell will never see the light of day on a desktop.
I think we must look at this in terms of the initial idea for the Cell CPU
Sony's vision for the first generation cell CPU is PS3 and LCD TV systems for home entertainment. They want to integrate all on a common platform that can be used for networking around homes and streaming content directly (this is one of the Cell's strongest points - see the demo of Cell streaming 32 video sequences at once).
Second generation CPU's will be found in other additional appliances around the home and office, again for similar reasons.
As this is currently Sony and Toshiba's sole vision it is likely other companies will jump on the bandwagon and introduce their very own 'Cell' style CPU's with similar specs and uses, in similar appliances, perhaps even utilising the initial blueprint for the CPU.
It's the start of a new generation of 'intelligent' home entertainment systems, and fair dos to Sony, they are at the forefront. It's only a matter of time before our TV's and PC's decide for us what channel to watch or station to choose.
As for the Cell and it's threat to the humble PC architecture, well, I personally think we will see a massive shift in the market over the next 5 years. The simple, stand alone PC base unit will become a thing of the past, and as we all install electrical induction loops and wireless networks around our homes, seperate processing boxes (such as base unit) will become irrelevent...
The individual neccessities will become key... i.e. we need a monitor to look at... so everything will become housed within that space-frame... for example, if the processor in the 42" lcd downstairs can handle the process you are trying to commit on the monitor upstairs, then why would you need another box or processor in the working room? it's all about space and energy saving now - its the future... and Sony have cleverly started it with the Cell CPU
To summarise, it's unlikely we will see desktops with cell because by the time of Cell's true maturity, desktops will be a thing of the past
Sony's vision for the first generation cell CPU is PS3 and LCD TV systems for home entertainment. They want to integrate all on a common platform that can be used for networking around homes and streaming content directly (this is one of the Cell's strongest points - see the demo of Cell streaming 32 video sequences at once).
Second generation CPU's will be found in other additional appliances around the home and office, again for similar reasons.
As this is currently Sony and Toshiba's sole vision it is likely other companies will jump on the bandwagon and introduce their very own 'Cell' style CPU's with similar specs and uses, in similar appliances, perhaps even utilising the initial blueprint for the CPU.
It's the start of a new generation of 'intelligent' home entertainment systems, and fair dos to Sony, they are at the forefront. It's only a matter of time before our TV's and PC's decide for us what channel to watch or station to choose.
As for the Cell and it's threat to the humble PC architecture, well, I personally think we will see a massive shift in the market over the next 5 years. The simple, stand alone PC base unit will become a thing of the past, and as we all install electrical induction loops and wireless networks around our homes, seperate processing boxes (such as base unit) will become irrelevent...
The individual neccessities will become key... i.e. we need a monitor to look at... so everything will become housed within that space-frame... for example, if the processor in the 42" lcd downstairs can handle the process you are trying to commit on the monitor upstairs, then why would you need another box or processor in the working room? it's all about space and energy saving now - its the future... and Sony have cleverly started it with the Cell CPU
To summarise, it's unlikely we will see desktops with cell because by the time of Cell's true maturity, desktops will be a thing of the past
chuckshissle
Splendid
- Feb 2, 2006
- 4,579
- 0
- 22,780
No it's gonna be years to get the Cell to work with a desktop. Sure it is possible and they have plans, even Nvidia, ATI and Texas instruments have plans to produce cpus. But Intel and AMD is just fine so don't be expecting the Cell to be avialable to desktop pcs anytime soon.
It would be horrible at anything you throw at it, any piece of code that you throw at that processor will run like shit. The design of that processor is soo bad. It makes no sence it was created to be cheap and thats it, the only reason why it recieves soo much spot light is because sony hypes this shitty processor up. The only thing that would be running your desktop would be the single ppe core, which is in-order, has a 512 kb cache its a 2 issue core, it has to scheduel all of the operations for those other 9 cores. Those 9 cores have to wait for the ppe core to send them/ scheduel an operation. Those 9 cores are being feed off of one pipeline they also only have a 256 kb cache which isnt adequate for any operation of any magnitude, there will be alot of cache misses in both the ppe core and the spes. What ibm doesnt tell you about the performance of the cell, is that they take the MAX/ PEAK performance of one spe, then they multiply that number by the ammount located on the chip, then they add the gflop performance of the ppe core. What they arent telling you that calculation is ridiculously flawed, as they arent factoring in a real world environnement, where the spes have to wait for the ppe to scheduel them an operation, and many other things. To sum this all up, its a poor performing proccessor, with a horrendously flawed design and is made to be cheap and thats it.
sunangel
Distinguished
- Feb 27, 2006
- 221
- 0
- 18,680
It has been discussed many times here. Don't be or get discouraged by the remarks of non-knowers. I provided you with a link that may give you a clue to the future of the Cell processor. Below is a snipet. To sum up, LBNL, after thorough testing, confirmed the processing power is 8x greater than any current processor of any company. STI (Sony, Toshiba, IBM) with minor modifications will be able to bring this to workstations. I am not sure if this means desktops, but since the two words are interchanged so often I am certain of its meaning. Please not the part about doubling the current effects of the processors' power.
"... The authors argue that Cell's three-level memory architecture, which decouples main memory accesses from computation and is explicitly managed by the software, provides several advantages over mainstream cache-based architectures. First, performance is more predictable, because the load time from an SPE's local store is constant. Second, long block transfers from off-chip DRAM can achieve a much higher percentage of memory bandwidth than individual cache-line loads. Finally, for predictable memory access patterns, communication and computation can be effectively overlapped by careful scheduling in software.
Overall results demonstrate the tremendous potential of the Cell architecture for scientific computations in terms of both raw performance and power efficiency," the authors wrote. While their current analysis uses hand-optimized code on a set of small scientific kernels, the results are striking. On average, Cell is eight times faster and at least eight times more power efficient than current Opteron and Itanium processors, despite the fact that Cell's peak double precision performance is fourteen times slower than its peak single precision performance. If Cell were to include at least one fully utilizable pipelined double precision floating point unit, as proposed in their Cell+ implementation, these speedups would easily double...."
Source: Lawrence Berkeley National Laboratory
http://www.cs.berkeley.edu/~samw/projects/cell/CF06.pdf
"... The authors argue that Cell's three-level memory architecture, which decouples main memory accesses from computation and is explicitly managed by the software, provides several advantages over mainstream cache-based architectures. First, performance is more predictable, because the load time from an SPE's local store is constant. Second, long block transfers from off-chip DRAM can achieve a much higher percentage of memory bandwidth than individual cache-line loads. Finally, for predictable memory access patterns, communication and computation can be effectively overlapped by careful scheduling in software.
Overall results demonstrate the tremendous potential of the Cell architecture for scientific computations in terms of both raw performance and power efficiency," the authors wrote. While their current analysis uses hand-optimized code on a set of small scientific kernels, the results are striking. On average, Cell is eight times faster and at least eight times more power efficient than current Opteron and Itanium processors, despite the fact that Cell's peak double precision performance is fourteen times slower than its peak single precision performance. If Cell were to include at least one fully utilizable pipelined double precision floating point unit, as proposed in their Cell+ implementation, these speedups would easily double...."
Source: Lawrence Berkeley National Laboratory
http://www.cs.berkeley.edu/~samw/projects/cell/CF06.pdf
Are you serious ??????? wtf. Dude my single core amd would rip the cell apart. Any dual core would rip the cell apart, it is not a powerful core. Analyse its archetecture for yourself, look at a few documents that arent IBM biased. And arent throwing unrealistic flawed calculations at you.
JonathanDeane
Distinguished
- Mar 28, 2006
- 1,469
- 0
- 19,310
I think the thing Cell will bring to computing is some advancements in multi threaded programming and maybe more physics in PC games when they do the normal movie based games that hit every single platform. Thats a good thing, but I dont think you will be able to buy a Cell based PC any time soon (maybe Sony will make one in Japan running Linux ???) with Linux I could see some use for it but those Linux guys can get it running on a toaster....
Multithreading on the cell will prove very slow. The ppe doesnt have an adequate cache size to actually scheduel operations for 9 cores, plus do all of the general purpose code that the spes cannot actually do. There isnt enough cache to do this efficiently, plus that core either way isnt powerful enough to do so. This core is sevral steps back from what we already have, just because it runs at 3.2 ghz doesnt mean anything. It probably has a deeper pipeline to get this 3.2 ghz clock speed, plus its only 2 issue. That whole processor is horrendously designed. What kind of quak wrote that article ?? this is something new to me... I was sure hardware would be better suited to manage memory then software... i mean this is all new to me software being faster then the hardware that runs it..... because you know that software is going to use up clock cycles running this software code...honestly. Also power and efficiency ??? the cell was burning up, was burning the hardware around the ppe core and the motherboard, sure.... thats really good power and efficiency.
sunangel
Distinguished
- Feb 27, 2006
- 221
- 0
- 18,680
Yes Mike995, you catching on quickly. This is why Rambus's new memory is used. Because of the limited caches sizes on the chip, the throughput and speed of the DRAM counters this problem. In effect, you can have an unlimited cache size, limited only by the amount of DRAM.
I know what your thinking. The memory controller will slow it down. Ah-Ha! The SPEs have direct memory access coordinated by the PPC chip.
I know what your thinking. The memory controller will slow it down. Ah-Ha! The SPEs have direct memory access coordinated by the PPC chip.
sunangel
Distinguished
- Feb 27, 2006
- 221
- 0
- 18,680
The following is directly from IBM's website and refers to the use of local storage (DRAM) as a form of cache.
... "To further reduce power, the team opted for a heterogeneous configuration with a novel SIMD-centered architecture. This configuration combines the flexibility of an IBM 64-bit Power Architecture™ core with the functionality and performance-optimized SPU SIMD cores. In this organization, the SPU accelerators operate from a local storage that contains instruction and data for a single SPU. This local storage is the only memory directly addressable by the SPU.
The SPU architecture was built to:
-provide a large register file,
-simplify code generation,
-reduce the size and power consumption by unifying resources, and
-simplify decode and dispatch."...
http://domino.research.ibm.com/comm/research.nsf/pages/r.arch.innovation.html
... "To further reduce power, the team opted for a heterogeneous configuration with a novel SIMD-centered architecture. This configuration combines the flexibility of an IBM 64-bit Power Architecture™ core with the functionality and performance-optimized SPU SIMD cores. In this organization, the SPU accelerators operate from a local storage that contains instruction and data for a single SPU. This local storage is the only memory directly addressable by the SPU.
The SPU architecture was built to:
-provide a large register file,
-simplify code generation,
-reduce the size and power consumption by unifying resources, and
-simplify decode and dispatch."...
http://domino.research.ibm.com/comm/research.nsf/pages/r.arch.innovation.html
Eviltwin17
Distinguished
- Feb 21, 2006
- 520
- 0
- 18,990
no clue, but if it ever happens it will cost you at least $5000... its like taking 4 or 5 amd fx processors and jamming them into one.
akscottbot
Distinguished
- Apr 23, 2006
- 47
- 0
- 18,530
I think the answer to Cell viability as a desktop processor are right there in front of you.
While has a tremendous amount of execution resources, its not a "smart" processor. It has no OOO execution, and no branch prediction except for in the PPE. It very much resembles an original pentium cpu with 8 DSPs that you might see in specialized A/V equipment straped on, all running at modern clock speeds. This adds up to a processor with incredible potential for crunching numbers. In scientific research and multimedia environments there is no processor that could touch Cell.
That being said Cell could never be considered a good choice for the desktop. It is too specialized. Cell is not designed to run instruction heavy code, but is better with heavy math. Code needs to be written directly for Cell, as there is no previous instruction language that can be utilized for it and compilers are very immature at this point. Furthermore the Cell is poor at handling OS code. A good example is the fact that the only Cell based servers being sold at this point use a Xeon to run Linux on the machine, while number crunching is offloaded to the Cells as assymetrical co-processors.
Cell could never handle office applications or inet/email any better that a celeron; it strenghs would be wasted. As for gaming, the benefits are limited as well. Gpu power is becoming more and more important compared to Cpu power in gaming, even more so with the development of dx10.
Probably the biggest obstacle to a Cell based desktop is lack of windows support. No processor will ever find success on the average desktop, if it doesn't have windows, at least not in the forseeable future.
And do you also know the fastest component in your computer is the cpus main cache. You also must understand how fast that cache needs to be, and how fast the cache will need to be in a highly multithreaded situation. And you must understand how slow acessing the ram bus will be especially through software managed memory control. We are hitting brick walls having a memory controler on a nortbridge which is hardware controled, software using up clock cycles, using up memory to run and also being run by hardware which isnt specialised to manage memory will prove to be horrendously slow. Can you imagine how slow the access from the cell to the rambus will be in terms of cpu caching. That capability will be useless at best. Oh and your spes memory access through the ppe is also soo slow, you arent understanding that the ppe core cannot handle 9 cores calling for information. You arent understanding how bottlenecked this archetecture is. This cpu is horrible, and I will tell you one more time, that even my cpu would tear that POS to threads.
crazypyro
Distinguished
- Mar 4, 2006
- 325
- 0
- 18,780
I have read in a magazine, and it has been said that a developer can use Visual C to write code for the Cell. Sony has announced a variant of Linux will be used on the Cell or will be made available for purchase to install on the PS3 to allow for people to do there own programming of Apps, games, or whatever your little coding hearts desire.
Oh, i almost forgot IBM is working on a compiler that lets developers write code for it just as if there writing code for a regular 64-Bit IBM PowerPC core, and it will automatically split the code to utilize the SPE's on the Cell. Its been said that prgramming of the cell is no different than programming for a G5, the only difference is developers have to the ability to split code and offload it to a SPE, which is where the learning curve is at with the PS3 right, how to write code to efficiently utilize all the SPE's.
There will be a viable Cell CPU for a desktop system when the Cell+ gets developed and comes out, as its main PowerPC core is supposed to have all those snazzy features you need to be a desktop chip, plus its SPE's. But it will be held back from being a desktop chip untill Rambus's XDR memory can become mainstream, and someone puts Windows Vista on it.
Oh, i almost forgot IBM is working on a compiler that lets developers write code for it just as if there writing code for a regular 64-Bit IBM PowerPC core, and it will automatically split the code to utilize the SPE's on the Cell. Its been said that prgramming of the cell is no different than programming for a G5, the only difference is developers have to the ability to split code and offload it to a SPE, which is where the learning curve is at with the PS3 right, how to write code to efficiently utilize all the SPE's.
There will be a viable Cell CPU for a desktop system when the Cell+ gets developed and comes out, as its main PowerPC core is supposed to have all those snazzy features you need to be a desktop chip, plus its SPE's. But it will be held back from being a desktop chip untill Rambus's XDR memory can become mainstream, and someone puts Windows Vista on it.
TRENDING THREADS
-
-
-
News US sanctions transform China into legacy chip production juggernaut — production jumped 40% in Q1 2024- Started by Admin
- Replies: 31
-
RTX 4070 vs RX 7900 GRE faceoff: Which mainstream graphics card is better?
- Started by Admin
- Replies: 70
-
Question I have been stuck between NVMe and SATA SSD. What should I do now?- Started by maniac2556
- Replies: 19
-
Question New pc build r9 7900x3d rtx 4080 super no post only ram rgb turns on- Started by Harvey Durward
- Replies: 5
-
Question 1TB HDD 80% Fragmented, Windows 10 Optimise Drives Program Doesn't Help
- Started by sdfbvcxbf
- Replies: 7
Facebook
Twitter
Instagram