Wednesday's PS3 4.30 Firmware Update to Remove Folding@home Client

[Rob Williams]

If you were starting to get the idea from Sony that it loves finding features to remove from its PS3, I think you've got it all wrong. Well, if you ignore the fact that with the next firmware update, Life and the Folding@home client will be removed. Past that, I don't think Sony goes out of its way to remove features at all!

Read the rest of our post and then discuss it here!

 

[marfig]

They could remove the ability to play games too. That one won't be missed.

 

[madmat]

Read through the blog comments, it seems the sheep are more concerned about some meaningless trophies than something beneficial to society. Of course that doesn't really surprise me.

 

[Rob Williams]

Read through the blog comments, it seems the sheep are more concerned about some meaningless trophies than something beneficial to society. Of course that doesn't really surprise me.

Yup... noticed that too. I kept reading and was thinking, "Come on! At least one person..."

 

[Brett Thomas]

It's a shame to see F@H pulled...but at the same time, as the average power bill (particularly in the US) is ridiculous and our energy consumption goes higher and higher, another power drain isn't a horrible thing to see go away...particularly since the PS3 version was ridiculously inefficient.

The track record of things getting removed from the PS3 is becoming legendary, however. "Here, play PS2/PS1 games...no, don't. Here, try linux. No, don't..." I had the old 60GB version with the Emotion engine chip in it...and it recently died. I hesitate to replace it, as the PS2 backwards compatibility was a HUGE part of my love for it. Now, it's becoming a shadow of itself and makes a much more difficult sell for me to buy a PS4. I've actually never seen a CONSOLE with such regressive featuresets.

 

[marfig]

I've actually never seen a CONSOLE with such regressive featuresets.

Or anything else for that matter. It's not easy to think of a device event remotely tied to the concept of technology that has become less feature rich, than more.

It's become clear to anyone right now that the whole concept behind a multi-purpose console device is something that SONY regrets having tried. For copyright reasons, or whatever else gives them bad dreams at night, they simply don't want it done anymore. What is absolutely surprising is seeing that they didn't just take it like a man and made sure the next version of the console didn't have none of it, and instead chose to remove the features that were advertised at the time of sale. I'm also quite amazed at the loophole in current consumer protection laws that allowed them to get away with it.

Don't mean to offend, but really this is a console brand that no elucidated consumer should ever consider anymore. I understand the huge investment some people made in PS3 games, but really, if gamers keep forgetting they are first and foremost consumers, they will keep reacting to the gaming industry as drug addicts who can take anything that is thrown at them.

 

[DarkStarr]

This is why I am so glad that FF is no longer PS only I play my Final Fantasy TWIMTBP on Xbox 360. (Too bad I can't run it on PC though..... would be SO MUCH EYE CANDY!)

 

[Kougar]

I'm obviously a folder (as if my sig wasn't enough of a hint), but honestly this isn't a big deal. The PS3 client has never been capable of much computational work... it's equivalent to running the SMP client on a Pentium D basically. Sure you get work done, but the power draw is high and the amount of work accomplished doesn't compare to a modern CPU or GPU client.

While PPD is not yet directly comparable across platforms as a measure of scientific work done (this is about to change, finally) a GTX 450 and any other low-end or old GPU could fold double the PPD or more than the PS3. A higher end GTX 480 or a Core i7 2600 would fold 15-25x the PPD at the low end, often more.

 

[marfig]

Note that the PS3 client had the potential to be a whole more more efficient and powerful than the GPU client. The reason this hasn't happened was because it wasn't anywhere near mature. This is a F@H problem, not anything to do with the capabilities of a PS3 hardware. So, for all that matters, removing support for folding is a decision that impacts on the PS3 feature set.

Had F@H decided to remove their PS3 client and SONY would be off the hook. It wasn't their problem; They offered all the functionality, F@H just chose not to take advantage of it. But this isn't what happened.

Incidentally, does go to show one of the limitations of the development model imposed by F@H. It's understandable they will not open source the clients. This would open a pandora box of potential hacked clients that could lie to the server and ruin the whole concept of distributed computing. But in contrast we are left with clients that are only updated at the convenience of the small developer teams that do have access to the source code. This invariably ends up producing situations like the PS3 client which, having the potential to become a powerhouse, has been instead one of the least efficient they had to offer.

 

[Kougar]

While I would be quick to agree that there was probably significantly more optimizations Pande Group could have done (based on how their software is developed & updated on other platforms) I'm not so sure about the processor aspect.

Cell only has a single actual "core", and eight SPE's branching off that core that handle the execution. One SPE is disabled for fab yield purposes, and a second SPE is dedicated purely to the PS3 OS and can't be utilized by devs. So that just leaves six SPEs left to play with.

I'm no expert, but highly-specialized SPEs with a severe memory access limitation doesn't seem to offer that much to work with. Here's the Link I was reading on it. What is amusing to me is that even the Xenon tri-core in the Xbox 360 supposedly only offers 1/3rd the floating-point multiple performance of a Pentium 4, (cell has one of these that handles the loading of the SPEs).

 

[DarkStarr]

Problem is Pande Group didn't deal with coding the PS3s F@H Client. Sony did, they decided they didn't want to do it anymore.

 

[marfig]

Here's the Link I was reading on it.

Interesting read, thank you. I did only a diagonal read of it but it was immediately apparent there was a bias towards game performance analysis on such areas as rendering and physics calculations, which really do not match the type of calculations performed on the PS3 by the F@H client.

For starters here's the essential read: The F@H FAQ on the PS3. There's plenty to read there. But it quickly becomes obvious the nature of the WIP that was taking place and the huge potential of this client considering that, like no other, it bridged the current gap between CPU folding and GPU folding.

In particular we should take into account future console generations (considering the already big performance of the PS3 client on a vast array of WU which makes it currently the faster client F@H has to offer, despite on some specialized WUs the GPU client being much better). These new generations have the potential to further improve console folding.

We also learn from that FAQ that, at the time of writing, close to a million PS3s were contributing to the project. SONY decision greatly reduces F@H folding capabilities in the future, and will have a big impact on its current output. It's a blow to a scientific project like no other I've seen recently. All in the name of god knows what.

 

[Kougar]

I dunno, in the FAQ while they state that it is more flexible, it can't run at the same speed as the PC calculations. "although not at the same speed increase relative to PC's". PPD performance of a PS3 was somewhere around 800-1,000 PPD, which doesn't come close to the 30,000-45,000 capable of a stock 3770K, or 25,000 points per day of a GTX 580.

Vijay Pande recently stated they have finally advanced their cores to the point that Pande Group can do the same work on the CPU client as they can on the GPU client. So the specialized nature of the PS3 isn't needed anymore. Along with this they are finally going to merge the GPU & CPU point return scales relative to a single benchmarking machine, so PPD can finally be directly comparable regardless of CPU or GPU client.

 

[Tharic-Nar]

You are also comparing a 7-year-old CPU with a brand spanking new one too (architectural difference aside), so I'm not surprised that the PS3 is now deemed 'inefficient'. The fact that the CPU is faster than a GPU in this case isn't a surprise either - CPUs have smarter pipelines and are specifically designed for this kind of workload of advanced scientific calculation - probably something that even a Tesla card would have problems with, but who knows.

While there are millions of PS3 out there, the F@H client was disabled by default, users had to manually activate it themselves... if they even knew it existed. It wouldn't surprise me if the number of PS3 folders was just a few tens of thousand.

 

[Rob Williams]

I'm obviously a folder (as if my sig wasn't enough of a hint), but honestly this isn't a big deal. The PS3 client has never been capable of much computational work... it's equivalent to running the SMP client on a Pentium D basically. Sure you get work done, but the power draw is high and the amount of work accomplished doesn't compare to a modern CPU or GPU client.

This is true, and one of the reasons I'd rather just leave the PS3 to rot in the corner than to turn it on and add a ton of heat to the room. It's hard to beat a CPU+GPU combo nowadays.

I'm no expert, but highly-specialized SPEs with a severe memory access limitation doesn't seem to offer that much to work with.

Good point. I ran into similar performance limitations when running Linux on it as well... having access to just 256MB of RAM or whatever it was, was a bit restrictive.

Problem is Pande Group didn't deal with coding the PS3s F@H Client. Sony did, they decided they didn't want to do it anymore.

Really? I don't remember reading that anywhere.

I dunno, in the FAQ while they state that it is more flexible, it can't run at the same speed as the PC calculations. "although not at the same speed increase relative to PC's". PPD performance of a PS3 was somewhere around 800-1,000 PPD, which doesn't come close to the 30,000-45,000 capable of a stock 3770K, or 25,000 points per day of a GTX 580.

Not that it matters much, if I recall I earned about 4K PPD on the PS3 when I used to run it. I am not sure if something changed over the years to decrease that or not.

 

[Kougar]

You are also comparing a 7-year-old CPU with a brand spanking new one too (architectural difference aside), so I'm not surprised that the PS3 is now deemed 'inefficient'. .

Hence exactly why it should be discontinued. It doesn't compare to modern hardware. I can run F@H on a Pentium 4 and suck 150 watts out of the wall, but at best it would get the same PPD as the PS3.

Comparing the PS3 to an E6700, a cheap dual core, no HT chip that hit the market before the PS3 arrived. PPD results I've googled up fall around the 2,200 PPD mark in Linux.

Not that it matters much, if I recall I earned about 4K PPD on the PS3 when I used to run it. I am not sure if something changed over the years to decrease that or not.

Are you sure it was 4,000 PPD? I am basing my facts off what I could google up, but there is no single reference table for hardwarePPD anymore. If you remember getting 4K then I'll take your word on it. All the reference threads & sites I knew about no longer exist or were only for bigadv projects.

4K Is a lot better than 1K though, regardless. 4K would allow it to at least beat the first generation of Conroe-based dual-core chips from 2006.

 

[Rob Williams]

4K Is a lot better than 1K though, regardless. 4K would allow it to at least beat the first generation of Conroe-based dual-core chips from 2006.

I just remember it being a lot more than 1K, and 4K seemed familiar. I almost feel like booting it up just to check, but I assume there wouldn't even be a workload for it.

 

[Kougar]

There might be. As long as they have projects left ready to go they'd be sending them.