Well, it all depends on how it's marketed, it could have a huge snowball effect.
If you award people 'hero points' and display a leaderboard showing how many lives each gamer has saved... or break it down by institution-- like college or business. Especially if the interface is really cool. Maybe have a hall of fame of cool-looking protein folds you could download.
It could become quite a competition to not play your PS3, particularly if launch titles turn out to be as good as launch titles typically are.
I thought that there must be some problem with the system if they're unable to get all the CPUs under full load.
It's actually really easy to do if your memory system isn't meant to service 8 cores. And the article pretty much backs this up, every time the quad cores fail to shine it's blamed on the memory. But to me, the really interesting aspect of this is that they always blame FB-DIMM, which gains bandwidth by sacrificing latency. They even go so far as to suggest:
if Apple were to release a Core 2 (Conroe/Kentsfield) based Mac similar to the Mac Pro, it could end up outperforming the Mac Pro by being able to use regular DDR2 memory.
So, I think regular DDR2 @ 667 = 5.4 GB/s... divided amongst 8 cores is just 677 MB/s per core. It seems insane to think that would work (maybe it would, maybe my numbers are wrong also). If you want to attack latency but simply can't give up the bandwidth, wouldn't the SMP model work better-- just swap out the L2-miss stalled thread, and run the other full bore. Now you've reduced the problem to distributing your register bank among active threads. Well, I think that's how video cards do it, and memory latency is their enemy #1.
In any event, there you have it. The performance pendulum has left Ghz, is briefly swinging toward more cores, but appears headed now toward memory systems. Does anyone else think it's funny that L1 is still just 32kb? (oughta be enough for anybody).
Not so far in future, there will be more bandwidth and then the limiting factor could be the speed/size of the memory. Or, it could be the power envelop of the entire system, Or, it could be back to the raw performance of CPUs.
Ah, but the future is now. Cell has already addressed these issues: 25.6 GB/s main memory bandwidth, 256 kb of L1 cache per core, OoO sacrificed to minimize heat, maximal raw performance of CPUs in FP, integer, FP, load/store, FP, and main memory transfer (DMA engine) without any silliness like 8 GP registers (128). Even when multiple Cells are hooked together, it's over a 35 GB/s IOIF port.
Also for onboard multitasking, you forgot about being latency-bound by atomic operations, which is something that would be really bad with separated L2 caches. This issue is also elegantly handled by cell by having only a single bus-snooped L2.
It must be frustrating for the hardware guys. You address all the bottlenecks in a pretty uniform way, and they still criticize: "But... uh, the software guys need a refresher course in hypertasking..."
The One positive PS3 has is its hardware. Its the software, sales, and marketing departments that are killing PS3. Did you watch E3? That was the longest continuous WTF I've ever seen. Powerpoint? Massive Crabs? Card monsters? The devkits they're giving developers are supposedly a total nightmare, with SDKs that take days to get working... And then sales goes and picks the magic number we all know.
But the hardware underneath it all is brilliant. Cell is a pretty sweet piece of chip for video games. RSX is good enough. Blu-Ray has 25 GB of storage, which should be enough for the next 5 years of games. A hard drive removes the artificial limitation of streaming bandwidth from building seamless worlds. Wireless controllers, Wifi internet, 1080p... tilt is tilt, even if a gimmick-- all these options are there for devs.
If only the rest of Sony could get their shit together, the box would sell itself.
The flop, of course, came from floating point operation. Even then it's vague--is it single, double or double-extended?
I was thinking the same thing. Running the numbers, 256 GFlop * 16,000 => 4.096 PFlop @ single precision. So if IBM means SP flops, something is slowing its theoretical max down by 2.5x. But Cell's DP perf yields 18.2 * 16,000 =>.292 PFlop @ DP. So that's not it either.
It's long been rumored that a post-PS3 Cell is in development that can pipeline DP flops. Its max theoretical DP perf would still be half of SP because it's just 2 DP values per 128-bit register instead of 4. AND, if you figure they lower the GHz to 3.2 to cut the heat output in half, you arrive at the magical number... 1.638 PFlop.
So can we take this as evidence that there now exists a Cell that performs DP calculations pipelined?
this company thinks so little of the mindless drones working for them that they don't even have the common courtesy to force their overworked, underpaid manager to take them into the back and fire them personally.
In what universe does this 'personal firing' by the 'overworked underpaid manager' benefit anyone? Does Mr Overworked enjoy firing? Does the fired employee benefit from some half-assedly personalized and prolonged epilogue to their membership in the Radio Shack Family?
No, it's all just a total logistical nightmare that's all bullshit anyway. Just fire people in hordes, make sure they can collect their unemployment, write a form-letter-of-recommendation and cut the bullshit down to the absolute, positive minimum.
If there's one thing the RT raytracing community is good at, it's explaining how good it works in theory. Take some numbers, extrapolate a little one dimension, then another and BOOM-- The Future. There are several problems with raytracing in real-time:
1) Static Objects Only. The huge majority of computation time is traversing a spatial subdivision structure. It happens that K-d trees offer the best characteristic (typically, fewest primitive per leaf for a given memory limit). However, these are really heinous to dynamically update. You can cheaply re-create it with median partitioning, but your trees are crappy. You can do a much nicer SAH (surface area heuristic), but to do this per frame blows out your CPU budget.
2) Bandwidth. Even if you could update your subdivision structure very cheaply, that structure still needs to be propogated out to all the CPUs participating in the raytrace. For the 1.87 MTri model they list on page 6, their spatial structure was 127 MB. Say you have a bandwidth of 6 GB/s, it takes 20ms just to transfer the structure (and there are other problems here). So your ceiling is 50 Fps before you trace your first ray.
3) Slower than a GPU. Even though they give you some little graph showing that raytracing (a static model, with static partitioning) beats a GPU at a MTri in the frame, this is very deceiving. The GPU pipeline works such that zillions of sub-pixel triangles simply can't get into pixel shaders fast enough, and force the pixel shader to be run many times extra. Double the resolution, however and the GPU won't take a cycle longer... with raytracing, performance will halve. So they found a bottleneck in the GPU which is totally unrepresentative of a game in every single sense, and said LOOK! BETTER! (in theory).
4) Hey, Where's my Features? All the cool things about raytracing (nice shadows, refraction, implicit surfaces, reflection, subsurface scattering) all get tossed out the window to make it real-time! What's the point, then? Given all the pixel shader hacks invented to make a GPU frame look interesting, the quality that can be achieved in a real-time raytrace is sadly tame. Especially when you consider that quality is the supposed advantage of raytracing.
Your heart is in the right place, but your response lacks foresight. Next-gen is not "being crammed down consumers' faces," because there is still tons of quality discount current-gen product floating around. Sony is even still publishing first party titles, and there are a few good third-party games coming. Your thought seems to emphasize this choice between early-adopting or throwing up your arms and quitting.
Just Wait. Price reductions are for people like you-- more sense than money, skeptical of frenzied impulse buys. Unhappy in long lines, Just wait and check out what you missed in the current-gen for almost free. But realize that the next-gen is necessary... 300 Mhz processors and 32kb of L1 cache only last so long.
the only thing that really would have made Dell change their tune is customer demand.
Well, that has been there for several years now. Perhaps because of the AMD lawsuit, Intel has had to tone down their vendor-bribing, and the revenue from AMD demand now outweighs the check Intel cuts to Dell.
If only DVD were big enough, I would agree with you. 6 GB is a pretty big limitation, and if you actually use all of it you're in a world of no fun with layer changes and attrocious inner-circle read speeds.
But the reality is that we're on the verge of really huge space-consuming technology: off-line procedural textures, baked detail maps on top of multiple layers of material maps, procedural geometry. These are not run-time jobs. Erosion for instance, can consume hours of CPU time, and artists always want the last tweak before an asset gets finalized.
Games have been small up til now because it takes too much artist time to go through and build 6 GB of assets, but that era is over. The future is in achieving uniqueness throughout the world, and that 'unique' part consumes a ton of memory. Of course, on the XBox, it looks like worlds will always be low-res heightmaps with tiling textures:\
Previous labs-on-a-chip existed in nature only for nanoseconds, as the lab would instantly eat the chip. The breakthrough here is configuring a lab on a very small chip, while showing the lab a much bigger chip, thus holding its mesmerized gaze indefinitely.
Re:PAX to fill void?
on
The End of E3?
·
· Score: 2, Insightful
Remember, E3 is NOT supposed to be a convention. Its a PRESS event.
Exactly. And this is a huge smackdown of the gaming press. This is the big publishers saying very simply "We don't need to impress you. It's much cheaper to buy you." And it's true! Publishers that spent huge $$$ on E3 can instead spawn publicity that times with game releases much better. They don't have to waste time on demos of half-finished games that will be shown side-by-side to titles almost ready to ship.
Or they can go directly to retailers with game demos. Imagine more console set-ups at your local gamestop or best buy. Next-Gen consoles all have downloadable demos, so publishers can hit gamers directly. E3 just makes no sense anymore (did it ever?). This is just further marginalization of a gaming press whose credibility has long been withering.
It's actually 2000 GFlops for a few hundred dollars.
And calling this a Petaflop supercomputer is similarly misleading, for roughly the same reason. The PS3 gets its 2 TF from the GPU, which can process 384 flops per cycle in an architecture built specifically to shade pixels. Likewise this MDGrape-3 is built at the hardware level to solve the n-Body problem, and that's it.
The Xbox 360 premium is functionally closest to the PS3 Core, except that it doesn't ship with an HD-DVD drive. The PS3 will see a much more even split between people who want Core and Premium.
Many people don't need the 60GB hard drive, HDMI out, built-in WiFi, etc.
Most folks I know dont even bother to look at the new games on the shelves because they assume they cant run it
I personally don't look at new games because they're such a pain to use. Download the new drivers, download a bios update, install the latest patch, fight through the DRM... ugh. Not to mention I only run a 6600GT, so your other point bears weight as well.
Personally, I feel like this will be a big blow to developers. MS is trying to squeeze out Nvidia for being "the friend of my enemy" in a different war. Nvidia will support XP to its dying end, while ATi will be pressured by MS to force users into a Vista purchase. In the end, everyone loses as consumers don't know what to buy, card makers don't know what drivers to support, developers don't know what DX version to target... and everyone buys a Wii.
Why Slashdot would link an Inquirer story is beyond me. Maybe Slashdot is for entertainment purposes only, but "News for Nerds" ought to be supported by some attempt at Fact. The Inquirer is just a machine meant to cause a ruckus for the purpose of page hits... any ounce of partiality or balance of truth be damned if it detracts from the hit count.
Linking stories from the front page is just feeding it. It's not news.
In fact, my major complaint is that the controller is wired, which won't be the case on the PS3. Compare a wireless controller that plays PS3/2/1 games and operates the CD/DVD/Blu-Ray drive versus the quantity and magnitude of your other remotes. Wireless controller wins in my book.
And more importantly, who actually wants more components in the entertainment system? If the PS3 weren't coming out so soon, I would buy a slimline just to cut down on space! If a PS3 can replace everything but the stereo, it has won my heart and wallet. And if it allows MP3s to be copied onto the hard-drive, and all I need is to supply a TV, receiver and speakers... $600 is acceptable.
Not buying it will not hinder game playing at all.
That's an odd way to look at it... How about "Game playing is hindered whether you buy it or not." Streaming off a DVD is huge pain for developers. It is at least doable in the low-res low-poly current-gen market, but the fact that all PS3 devs can design knowing a hard drive is installed opens up a lot of possibilities there.
That, coupled with the extra storage space available on Blu-Ray means that Sony developers don't have to worry about "legacy" next-gen systems in 2008,9, and so on.
The Core 2 Duo benchmarks we ran were not completed in our own labs and we have used some unfamiliar tests in order to establish how well the new Core architecture performs. This was because we were not allowed to tweak the system or install our own benchmarks - the machine was built and configured by Intel engineers.
So take these results with an even smaller grain of salt. Goddamn benchmarks.
The point about context switching is a good one. Not only do all the registers need to be saved, but the entire 256 kb of local store! That's a hugely non-trivial feat, but I think performance applications will be written to avoid context switches entirely.
The RAM is XDR. The IOIF (to talk to other Cells) connection is 2 FlexIO ports. The bus itself (called the EIB ) is something like 300 GB/s. I agree that peak is never achievable, but it should be possible to get around 18 GB/s or so.
Slashdot Mirror
If you award people 'hero points' and display a leaderboard showing how many lives each gamer has saved... or break it down by institution-- like college or business. Especially if the interface is really cool. Maybe have a hall of fame of cool-looking protein folds you could download.
It could become quite a competition to not play your PS3, particularly if launch titles turn out to be as good as launch titles typically are.
It's actually really easy to do if your memory system isn't meant to service 8 cores. And the article pretty much backs this up, every time the quad cores fail to shine it's blamed on the memory. But to me, the really interesting aspect of this is that they always blame FB-DIMM, which gains bandwidth by sacrificing latency. They even go so far as to suggest:
So, I think regular DDR2 @ 667 = 5.4 GB/s... divided amongst 8 cores is just 677 MB/s per core. It seems insane to think that would work (maybe it would, maybe my numbers are wrong also). If you want to attack latency but simply can't give up the bandwidth, wouldn't the SMP model work better-- just swap out the L2-miss stalled thread, and run the other full bore. Now you've reduced the problem to distributing your register bank among active threads. Well, I think that's how video cards do it, and memory latency is their enemy #1.
In any event, there you have it. The performance pendulum has left Ghz, is briefly swinging toward more cores, but appears headed now toward memory systems. Does anyone else think it's funny that L1 is still just 32kb? (oughta be enough for anybody).
Ah, but the future is now. Cell has already addressed these issues: 25.6 GB/s main memory bandwidth, 256 kb of L1 cache per core, OoO sacrificed to minimize heat, maximal raw performance of CPUs in FP, integer, FP, load/store, FP, and main memory transfer (DMA engine) without any silliness like 8 GP registers (128). Even when multiple Cells are hooked together, it's over a 35 GB/s IOIF port.
Also for onboard multitasking, you forgot about being latency-bound by atomic operations, which is something that would be really bad with separated L2 caches. This issue is also elegantly handled by cell by having only a single bus-snooped L2.
It must be frustrating for the hardware guys. You address all the bottlenecks in a pretty uniform way, and they still criticize: "But... uh, the software guys need a refresher course in hypertasking..."
But the hardware underneath it all is brilliant. Cell is a pretty sweet piece of chip for video games. RSX is good enough. Blu-Ray has 25 GB of storage, which should be enough for the next 5 years of games. A hard drive removes the artificial limitation of streaming bandwidth from building seamless worlds. Wireless controllers, Wifi internet, 1080p... tilt is tilt, even if a gimmick-- all these options are there for devs.
If only the rest of Sony could get their shit together, the box would sell itself.
I was thinking the same thing. Running the numbers, 256 GFlop * 16,000 => 4.096 PFlop @ single precision. So if IBM means SP flops, something is slowing its theoretical max down by 2.5x. But Cell's DP perf yields 18.2 * 16,000 => .292 PFlop @ DP. So that's not it either.
It's long been rumored that a post-PS3 Cell is in development that can pipeline DP flops. Its max theoretical DP perf would still be half of SP because it's just 2 DP values per 128-bit register instead of 4. AND, if you figure they lower the GHz to 3.2 to cut the heat output in half, you arrive at the magical number... 1.638 PFlop.
So can we take this as evidence that there now exists a Cell that performs DP calculations pipelined?
Good Lord! Another
I see through your 'acme' shenanigans.
In what universe does this 'personal firing' by the 'overworked underpaid manager' benefit anyone? Does Mr Overworked enjoy firing? Does the fired employee benefit from some half-assedly personalized and prolonged epilogue to their membership in the Radio Shack Family?
No, it's all just a total logistical nightmare that's all bullshit anyway. Just fire people in hordes, make sure they can collect their unemployment, write a form-letter-of-recommendation and cut the bullshit down to the absolute, positive minimum.
1) Static Objects Only. The huge majority of computation time is traversing a spatial subdivision structure. It happens that K-d trees offer the best characteristic (typically, fewest primitive per leaf for a given memory limit). However, these are really heinous to dynamically update. You can cheaply re-create it with median partitioning, but your trees are crappy. You can do a much nicer SAH (surface area heuristic), but to do this per frame blows out your CPU budget.
2) Bandwidth. Even if you could update your subdivision structure very cheaply, that structure still needs to be propogated out to all the CPUs participating in the raytrace. For the 1.87 MTri model they list on page 6, their spatial structure was 127 MB. Say you have a bandwidth of 6 GB/s, it takes 20ms just to transfer the structure (and there are other problems here). So your ceiling is 50 Fps before you trace your first ray.
3) Slower than a GPU. Even though they give you some little graph showing that raytracing (a static model, with static partitioning) beats a GPU at a MTri in the frame, this is very deceiving. The GPU pipeline works such that zillions of sub-pixel triangles simply can't get into pixel shaders fast enough, and force the pixel shader to be run many times extra. Double the resolution, however and the GPU won't take a cycle longer... with raytracing, performance will halve. So they found a bottleneck in the GPU which is totally unrepresentative of a game in every single sense, and said LOOK! BETTER! (in theory).
4) Hey, Where's my Features? All the cool things about raytracing (nice shadows, refraction, implicit surfaces, reflection, subsurface scattering) all get tossed out the window to make it real-time! What's the point, then? Given all the pixel shader hacks invented to make a GPU frame look interesting, the quality that can be achieved in a real-time raytrace is sadly tame. Especially when you consider that quality is the supposed advantage of raytracing.
And c'mon. It's Gameplay that counts anyway :P
Your heart is in the right place, but your response lacks foresight. Next-gen is not "being crammed down consumers' faces," because there is still tons of quality discount current-gen product floating around. Sony is even still publishing first party titles, and there are a few good third-party games coming. Your thought seems to emphasize this choice between early-adopting or throwing up your arms and quitting.
Just Wait. Price reductions are for people like you-- more sense than money, skeptical of frenzied impulse buys. Unhappy in long lines, Just wait and check out what you missed in the current-gen for almost free. But realize that the next-gen is necessary... 300 Mhz processors and 32kb of L1 cache only last so long.
Well, that has been there for several years now. Perhaps because of the AMD lawsuit, Intel has had to tone down their vendor-bribing, and the revenue from AMD demand now outweighs the check Intel cuts to Dell.
But the reality is that we're on the verge of really huge space-consuming technology: off-line procedural textures, baked detail maps on top of multiple layers of material maps, procedural geometry. These are not run-time jobs. Erosion for instance, can consume hours of CPU time, and artists always want the last tweak before an asset gets finalized.
Games have been small up til now because it takes too much artist time to go through and build 6 GB of assets, but that era is over. The future is in achieving uniqueness throughout the world, and that 'unique' part consumes a ton of memory. Of course, on the XBox, it looks like worlds will always be low-res heightmaps with tiling textures :\
Previous labs-on-a-chip existed in nature only for nanoseconds, as the lab would instantly eat the chip. The breakthrough here is configuring a lab on a very small chip, while showing the lab a much bigger chip, thus holding its mesmerized gaze indefinitely.
Exactly. And this is a huge smackdown of the gaming press. This is the big publishers saying very simply "We don't need to impress you. It's much cheaper to buy you." And it's true! Publishers that spent huge $$$ on E3 can instead spawn publicity that times with game releases much better. They don't have to waste time on demos of half-finished games that will be shown side-by-side to titles almost ready to ship.
Or they can go directly to retailers with game demos. Imagine more console set-ups at your local gamestop or best buy. Next-Gen consoles all have downloadable demos, so publishers can hit gamers directly. E3 just makes no sense anymore (did it ever?). This is just further marginalization of a gaming press whose credibility has long been withering.
And calling this a Petaflop supercomputer is similarly misleading, for roughly the same reason. The PS3 gets its 2 TF from the GPU, which can process 384 flops per cycle in an architecture built specifically to shade pixels. Likewise this MDGrape-3 is built at the hardware level to solve the n-Body problem, and that's it.
Many people don't need the 60GB hard drive, HDMI out, built-in WiFi, etc.
On the right there's new and used X360s for £150 or less.
HA! You think the Wii won't sell out? Be prepared to pay $600 on eBay!
If anyone wants an XBox, no need to wait. In stores everywhere.
I personally don't look at new games because they're such a pain to use. Download the new drivers, download a bios update, install the latest patch, fight through the DRM... ugh. Not to mention I only run a 6600GT, so your other point bears weight as well.
Personally, I feel like this will be a big blow to developers. MS is trying to squeeze out Nvidia for being "the friend of my enemy" in a different war. Nvidia will support XP to its dying end, while ATi will be pressured by MS to force users into a Vista purchase. In the end, everyone loses as consumers don't know what to buy, card makers don't know what drivers to support, developers don't know what DX version to target... and everyone buys a Wii.
But was the up-mod for the insight, or for the irony? ;)
Some of our zany mods need to tick it up a few for Informative, just to seal the deal
Why Slashdot would link an Inquirer story is beyond me. Maybe Slashdot is for entertainment purposes only, but "News for Nerds" ought to be supported by some attempt at Fact. The Inquirer is just a machine meant to cause a ruckus for the purpose of page hits... any ounce of partiality or balance of truth be damned if it detracts from the hit count.
Linking stories from the front page is just feeding it. It's not news.
He bore the look of a soulless ghost.
In fact, my major complaint is that the controller is wired, which won't be the case on the PS3. Compare a wireless controller that plays PS3/2/1 games and operates the CD/DVD/Blu-Ray drive versus the quantity and magnitude of your other remotes. Wireless controller wins in my book.
And more importantly, who actually wants more components in the entertainment system? If the PS3 weren't coming out so soon, I would buy a slimline just to cut down on space! If a PS3 can replace everything but the stereo, it has won my heart and wallet. And if it allows MP3s to be copied onto the hard-drive, and all I need is to supply a TV, receiver and speakers... $600 is acceptable.
That's an odd way to look at it... How about "Game playing is hindered whether you buy it or not." Streaming off a DVD is huge pain for developers. It is at least doable in the low-res low-poly current-gen market, but the fact that all PS3 devs can design knowing a hard drive is installed opens up a lot of possibilities there.
That, coupled with the extra storage space available on Blu-Ray means that Sony developers don't have to worry about "legacy" next-gen systems in 2008,9, and so on.
It's kind of an interesting anti-MS combo: Google, Sony, and IBM. All riding the Linux pony.
So take these results with an even smaller grain of salt. Goddamn benchmarks.
The RAM is XDR. The IOIF (to talk to other Cells) connection is 2 FlexIO ports. The bus itself (called the EIB ) is something like 300 GB/s. I agree that peak is never achievable, but it should be possible to get around 18 GB/s or so.