That cost is wrong, covered this repeatedly.
No, you simply ignored all the manufacturing costs and all the ancillary costs which I listed, and as I expected you are focused on development costs. $7, $7, $7. :sneaky: Ya, OK, Thanks for proving you have no idea about how the semi-conductor industry works at all and that your knowledge is strictly from a programming perspective. I didn't know it cost $0 to pay for the wafers at a foundry, $0 to test them at the foundry, $0 to ship them and assemble them inside a PS3, $0 costs to test the working units inside a PS3, etc.
Maybe you should convince NV and AMD know that it cost them $7-8 of real cash flow per Tahiti XT/Kepler GK104 per each HD7970/GTX680 sold. Good one.
It had less RAM and a slower GPU.
Yes, thanks for proving further that the Cell was a waste of $ since the other components in PS3 were more limiting. In other words the Cell was too expensive, and power hungry and didn't offer any performance advantage for games over a Core 2 Duo, and as a package couldn't outperform Tri-Core Xenon + R500 in the Xbox360 after
6 years of programmers learning how to optimize the Cell to its maximum potential.
Sony is an IP owner for Cell, they make the processor themselves. The R&D cost associated with that is less then $7 per chip.
Like I said, goes 1 ear, leaves the other. Using your faulty logic, it costs NV $7-8 to manufacture a GK104 and sell it in GTX680. How much it cost to actually sell the Cell CPU to Sony is not what it cost to develop the design alone. That's basics that you can't even get right. I suggest you enroll in some business courses in school to understand how manufacturing, technology and semi-conductor companies make money and what underlying costs they have. Since understanding how companies function and make cash flow is my line of work, I do not need to explain myself since you can't even grasp the basic idea of cost of goods sold as well as ancillary costs outside of R&D. The cost to actually deliver the chip inside the PS3 was not $7. You are delusional if you believe that or just ignorant to how a technology/business company actually functions. Time to leave your programming cubicle and pick up some business books.
$400 cheaper then the closes BluRay play that was still inferior due to it not having Cell.
BluRay capability has nothing to do with the Cell processor. Completely irrelevant to the point being made that the Cell was too expensive, inefficient to code for and not much faster for games than a Tri-core Xbox360 CPU, much less a Core 2 Duo CPU.
Nope. At low resolutions a system becomes both CPU and GPU limited, exactly the same way for PCs. The only debate is which component is the more limiting, depending on the game. Either way your argument has failed on both accounts:
1) IF the consoles are primarily GPU limited, then the PS3's Cell didn't aid in graphics;
2) IF the consoles are both CPU and GPU limited at both resolutions, the Cell's superiority didn't allow it to post higher minimum framerates in many CPU limited situations, such as Blighttown in Dark Souls, etc.
So you have not proven how the Cell was beneficial overall given its expensive cost and power inefficiency.
Cell was going to be the graphics processor in the original PS3, I have noted that would have been moronic against 2005 GPUs.
It was moronic to use the Cell in the first place. Sony would have gotten a faster console had they gone with a more modern version of AMD's GPU and reused the exact same Tri-core CPU from Xbox360. Instead, they spent more $ on a barely faster CPU and paired it with a slow GPU. Fail 2/2.
Yup, he said PS3 is horribly inefficient to code for and much slower than modern PCs. You fail to read those quotes, but keep posting your opinion. 100s of other professionals in the industry, including Carmack, agree that PS3's processing power is exhausted. You say the Cell is faster than modern x86 CPUs. That contradicts that idea that the Cell's power is exhausted. Does not compute.
In Vector code, which is where Cell kills POWER.
Useless for real world gaming performance of PS3 console. This hardly showed up in real games against the 360. Thus, irrelevant.
A random person's research paper vs.
The person who designed the Cell: ~
80W of power.
PS3 uses 240W of power vs. 180W in 360, yet doesn't have better graphics. GeForce 7950 Mobile used less than 60W of power, thus the difference is mainly attributable to the Cell = fail.
1 Core i7 Nehalem is faster in games than the entire Tri-Core PowerPC-based Xenon CPU. 4 Core i7 3770K has a 77W TDP with an APU. Cell and Tri-Core PowerPC CPUs were slower overall than the PowerPC CPU used in G5 by Apple. Apple abandoned PowerPC achitecture for Core 2 Duo in 2005, citing superior performance of Intel's CPUs, superior performance/watt and superior power consumption.
Since,
Xbox360 Tri Core, Cell in PS3 < PowerPC G5 processor < Core 2 Duo < 1 Core i7 Nehalem, i7 3770K is faster than all of these processors, while consuming less power then the Cell did. :thumbsup:
Any chip must be produced. Sony owns the IP for Cell. The cost to acquire this IP was $7.
The cost to manufacture / produce a semi-conductor chip is not a "Sunk Development Cost" as you seem to allude to. The cost of the Cell's IP is only 1 aspect of what the real net cash outlay for each Cell chip that went inside PS3 was. Is the cost to manufacture GK104s out of a 300mm wafer $7-8 per chip? I didn't think so.
That was an estimate based on what a comparable CPU would cost them if they bought it from another company. Sony owns Cell.
Incorrect. Someone has to pay for manufacturing: wafer costs, yields, testing of CPU chips, sorting them based on clock speeds, defects, ship them to a factory, assemble them inside a PS3, test the console to make sure the CPU works inside the console. All of these are costs Sony had to pay for per each console sold = $230 at launch. These costs in aggregate can be summarized as manufacturing costs for the Cell.
The manufacturing cost per each Cell chip fell
$37.73 by November-December 2009:
Those are real direct costs to Sony per each Cell chip by end of 2009. Your assessment that the Cell only costs Sony $7 in cash flows is incorrect based on lack of understanding how technology or semi-conductor businesses work.
The actual programmers you quoted said that POWER is better under ideal circumstances, not even the entire Cell processor, just the light alternative of it.
That's not what they said.
Oles Shishkovstov: You can calculate it like this: each 360 CPU core is approximately a quarter of the same-frequency Nehalem (i7) core. Add in approximately 1.5 times better performance because of the second, shared thread for 360 and around 1.3 times for Nehalem, multiply by three cores and you get around 70 to 85 per cent of a single modern CPU core on generic (but multi-threaded) code. Bear in mind though that the above calculation will not work in the case where the code is properly vectorised. In that case 360 can actually exceed PC on a per-thread per-clock basis.
1) He said on a per-thread per clock basis, not overall performance.
2) Ideal vectorized code is meaningless since it's again just theoretical performance. You love talking about theoretical performance. The only thing that counts is real world performance. Theoretical performance is good for marketing and geeky discussions. In the real world using "generic code" (i.e., translated: modern code that's easy to optimize), the Tri-Core 360 CPU is 70-85% of the speed of 1 Core i7 CPU.
Now if 1000 people spend 1000 years writing ideal vectorized code for the Cell, it should outperform Core i7 CPU on a per core, per clock basis. Such ideals are meaningless since that's not realistic, cost effective or has been reproduced in real world games on PS3.
Your claims went from ludicrous to absurd: so now just a fraction of the Cell is better than the entire modern x86 processor under ideal circumstances? What are those ideal circumstances, does the programmer have to work on Mars to achieve those results for 10 years? They were never achieved in the real world, which means a theoretical ideal is meaningless. A PowerPC G5 has not outperformed a modern x86 Core 2 Duo processor in any real world game as far as I am aware. The Cell is computationally slower to a modern Core i7 CPU and Intel makes the fastest CPUs on the planet for games. The Cell won't be in PS4 because AMD's modern x86 CPUs are even faster than the Cell.
Do I think it would be a mistake for Sony to abandon Cell? Yes.
I am glad you don't work for Sony then because their engineers and management team have moved on to the year 2012, and you are still stuck in 2006 fairy tales dreams of a programmer who can only think in theoretical FLOP terms and vectorized ideal code.
The only reason people seem to think it will that I have been able to find is Kotaku.
No, it's because every major developer probably begged Sony to ditch the Cell and never use such a terrible design again. They also were very vocal against the Cell from the beginning. Hopefully, Sony's management wakes up. Can't go wrong with any choices really: x86 AMD Fusion or Intel Core i3 CPU would all vastly outperform the Cell SPE setup for next generation games. No matter what Sony chooses, even a more modern PowerPC CPU-derivative with actual cores and none of the SPE non-sense is going to be miles faster.
No, it wasn't. Cerb stated it wasn't a direct derivative, Cerb was 100% correct. It is based on the POWER architecture, it isn't a direct derivative.
What are you even arguing about. I said the Tri-Core CPU in the 360 and the Cell are based on the PowerPC core, with some modifications in the form of SPEs for the Cell and so on. What derivative did I say it's based on? I don't recall naming a specific PowerPC #?
Again, as simple as can be:
The Cell Processor contains 9 processors on a single chip. One is a conventional PowerPC processor, with standard level one (32+32K) and level two (512K) caches and transparent direct access to system memory. To conserve chip space and power this PowerPC is simpler than other common processors. It does not provide hardware support for branch prediction or out of order execution. This makes it perform worse than one would expect given its clock speed (3.2 Ghz in the Playstation 3). The expectation is that the PowerPC will be used in a supervisory role and the majority of processing will be delegated to the SPEs.
The main 3 computation cores in the 360 and the Cell's single core are both based on an IBM PowerPC CPU architecture of that time, which happens to be
slower than the PowerPC CPU architecture of the G5 Apple processor. You are putting words into my mouth that I said it was an exact derivative of "what"? All I said is it's a derivative of a PowerPC CPU architecture of that time. That's just a fact. What you can't seem to grasp for 16 pages now is that the PowerPC CPU architecture of that time was vastly inferior in gaming performance to AMD's or Intel's x86 CPUs on a per core per thread basis in the real world (i.e., not ideal circumstances that only IBM/Sony engineers pushing Cell dreamed up of). The main reason Sony went for this design is the management team wasn't technically knowledgeable enough to understand the flaws in the Cell CPU architecture/design. The same people probably would have chosen the Bulldozer FX-8150 8-core CPU against an Ivy Bridge Core i3 if you told them the FX8150 is a real 8-core CPU vs. a dual-core HT "slower" Intel i3 processor. You would have thrown a bunch of fake marketing slides, showed superior floating point performance of the FX8150 and sold the same Sony execs BS of how an 8-core CPU with immense floating point capabilities would beat a dual-core Intel CPU for games.
You didn't understand that article. The POWER architecture IBM developed is still very much alive on their server offerings for large businesses and continues to evolve, but the specific Cell architecture (i.e., PowerPC computational core supported by SPEs) for the
consumer space is very much dead. So, no Cell in PS4! Don't worry, since you are so smart, maybe you can take the Cell out of the PS3 and mod it into the PS4 to get better graphics next generation. Maybe get a better after-market cooler so you can overclock the Cell since it uses so little power to begin with.
So would you say an Atom can be compared by performance per MHz to Ivy Bridge? No. Why not?
Facebook
Twitter