Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
5-Fold Increase in Video Encoding with OpenCL-like Technology
- Thread starter MacRumors
- Start date
- Sort by reaction score
What the $**% are people bitching about in regards for PPC? You did not buy those PPC G5 Macs for future support, you bought those machines to do work at that time. Computers are not time machines, they are to help you do work, nothing else. They do not have a time limit, they do not die in 2-3 years just because the OS has moved on. They will continue to work for more than 4-5 years with the same software.
People should consider upgrading to a newer OS as a bonus, not a goddamn given right.
WTF is wrong with you people?
People should consider upgrading to a newer OS as a bonus, not a goddamn given right.
WTF is wrong with you people?
tea leaves? You're suggesting the fact that not a single beta build of Snow Leopard runs on PCC is just tea leaves?
That was when Intel had a more limited design. Intel has dumped that approach now. AMD took their eye off the ball for a moment. Looks like they are getting back in the game though.
Maybe when Apple makes the transition to CPU+GPU combo chips AMD may get a shot providing that for them (if they have something competitive at that point.).
This parallel processing with communication overhead between the processing elements works OK when you have problems that require more computation than communication. Where the communication element is higher this kind of decomposition isn't going to work so hot. Not all applications are constrained by code blocks that are highly parallel and/or sit in a chunk of code that is highly parallel for a long period of time.
Secondly, coprocessors on cards are somewhat limited by how much power they can draw over the PCI bus on a Mac Pro that is limited. [ if could run a separate power drop from the power supply to the card that might change but Apple's boxes don't afford for that. ]
Not so sure AMD is going to fill that niche. They'll have a better chance if can come up with GPGPUs that get stuffed into iMacs (or something else of high units, something priced to "move", to be successful in earning much.).
I'll admit to not bothering to keep up with all the model numbers, but I suspect you're comparing desktop class processors (3GHz dual cores) with mobile GPUs. I think you're also ignoring the fact that the GPU comes free, and is much better optimized for a lot of stream processing that CPUs just don't handle well. You're also looking at the results of one application under test and trying to over generalize.
Your initial point (not quoted above) is correct-- it's for specialized use. I've got computer vision algorithms that I couldn't run real time on my MacBook Pro, that now run real time with about 13% CPU loading just by moving some key image transforms into Core Image and leaving the CPU to handle control logic. Even the mobile class GPUs can pack a lot of punch.
Seeded in only x86-64 ? I can imagine Apple could shaft a group of folks by going 64-bit only for the core and dropping the G5 (even though they are 64 bit). That would make it easier on them. However, it doesn't follow their past policies. Certainly they can throw all the folks who bought on the Intel transition under the bus. What I've reacting to is folks trying to spin that as Apple's normal policy to legacy support. It hasn't been (e.g., if recall correctly, didn't abnormally shrink support on the 68K to PowerPC transition. )
Apple considers stuff over 5 years vintage. That's reasonable. However, that is 2 years more than 3. Four years is close to being a judgment call. If Snow Leopard drags out the release day closer to Christmas..... even closer to 5.
If Apple throws the G5 folks under the bus and not the 32 only Intel folks that is bogus. Pragmatically both are different builds than a x86-64. If is the "number of builds" is a problem why not reduce costs of both? Apple can pull that stunt. However, if they ever get to the point where would to make another transition, then their credibility is going to be crap that they won't shaft folks who buy products close to the transition point. That is what happens to companies that don't stick spirit of their published policies.
Several folk have pointed out that customer knew Apple was in a transition. That ignores the point that Apple also knew. Indeed, they knew more than the customers since Apple does not give forward information in many cases. The transition was going to have costs. To not have accounted for those was Apple's problem. All platform migrations cost extra money. If making the migration didn't give you a cost savings to pay for the migration then shouldn't have done it. Likewise they could have covered some increased costs in the price of the machines when sold (ignoring that folks would pay for the OS upgrade).
If the numbers say that a G5 only population was too small for an upgrade then fine. That's would be a more reasonable call. However, yelping about 3 year "pro class" computers ($2,500-$3,000+ machines ) are not worthy of software support is crap. Would be yet another reason why Apple doesn't do well with businesses (in addition to limited forward guidance) Especially on an OS release that was primarily aimed at incremental improvements and quality improvements. (although I suspect Apple will change the tune on that come WWDC.)
It might compare not so well actually.
I kind of didn't want to blow the steam off by dumping my personal experience here. It seems that few people here have experience with Nero MP4. It was from the beginning targeted at real-time encoding.
Two years ago Nero could do MP4 encoding of 480i content in real-time - as it was streamed over network. But the price for such high performance is overall bad quality of the video. On TV it looked OK, but on computer screen ... I'd rather wait for HandBrake to finish.
For streaming Nero is excellent. For movie encoding - plainly bad.
"Suggesting" isn't quite strong enough, but yes. Why should one assume its absence constitutes evidence that PPC won't be in the final release, let alone an official statement from Apple to that effect?
Edit: I guess I just want this little bit of conjecture to be properly labeled. People are propagating it as fact, but it's just supposition.
Incorrect Grammar
>reduces video encoding time by up to five times
You can't reduce anything by 5 times. Once you reduce it one time, there is nothing left. What he should have said is "reduces video encoding time to one fifth of what it was".
Sorry about the nitpick. This type of phrase is a pet peeve.
>reduces video encoding time by up to five times
You can't reduce anything by 5 times. Once you reduce it one time, there is nothing left. What he should have said is "reduces video encoding time to one fifth of what it was".
Sorry about the nitpick. This type of phrase is a pet peeve.
Call me when there is absolutely any evidence that Snow Leopard will run on PPC machines.
Thanks for the input. I have done numerous comparisons on my computer with badaboom vs handbrake of HD content. Handbrake wins every time in quality. It actually makes me sad. I paid $29 for badaboom and the only thing I will use it for is my iphone. When I want to encode for any larger screen I use handbrake and wait the extra time. As of right now CUDA encoding has pretty horrible quality.
No one is making that assertion. You might want to go back and read the context.
The purpose of betas is to shake out bugs before going to production. If no one runs the PPC stuff then can't report the bugs. Your also not allowing for folks to test out their universal binary aspects. Ideally both sides of the universal binary should perform the same but sometimes there are generation differences. [ But if the objective is to nuke universal binaries, then there is reason not to deploy a PPC beta. ]
However, yes. Sometimes just do betas on one platform when releasing on one platform. Helps turn the crank faster if have smaller regression suite to execute. Depends upon how stable the code is. What some folks consider "beta" is not what others do. [ some say beta is feature complete but buggy. Others beta is something that doen't blow up very often and has many features stubbed out. ] In a "alpha"-to-"beta" state would make sense to limit the complexity. In the last 3-4 months, before release should have something more complete though.
The thing that kills me, about the PPC-whiners, is this:
If your g5 is so great that you're mad at not getting Snow Leopard for it, then WHY are you prepared to be instantly dissatisfied with your G5 the day the OS is released?
If you and your G5 NEED Snow Leopard, then let's face it, your G5 sucks. If it's worth keeping, it's worth keeping.
That 3-5 years of use/progressive-obsolescence did not go down the toilet. If you're happy with your G5 in 2009, how can an incompatible upgrade make you bat an eye? If you're so desperate for a faster machine, then get one, instead of the sense of entitlement that Apple owes you a cheap/pirated upgrade.
Oh, yeah. One more for the clueless...10.6 is 10.5, OPTIMIZED FOR INTEL MACHINES.
Spoiled brats.
If your g5 is so great that you're mad at not getting Snow Leopard for it, then WHY are you prepared to be instantly dissatisfied with your G5 the day the OS is released?
If you and your G5 NEED Snow Leopard, then let's face it, your G5 sucks. If it's worth keeping, it's worth keeping.
That 3-5 years of use/progressive-obsolescence did not go down the toilet. If you're happy with your G5 in 2009, how can an incompatible upgrade make you bat an eye? If you're so desperate for a faster machine, then get one, instead of the sense of entitlement that Apple owes you a cheap/pirated upgrade.
Oh, yeah. One more for the clueless...10.6 is 10.5, OPTIMIZED FOR INTEL MACHINES.
Spoiled brats.
Approximately FIFTEEN PERCENT of active Macs are PPC, today, and dropping.
Check out this graph for 2009:
http://att.macrumors.com/attachment.php?attachmentid=164381&d=1238160682
And this thread, where I read it:
https://forums.macrumors.com/posts/7355635/
:"Hmm, spend millions of dollars to further optimize the already good OS for -15%, and dropping, of Macs, or make a streamlined OS for the +85%, that is only possible because the shackles of backwards-compatibility are removed? Decisions, decisions..."
Check out this graph for 2009:
http://att.macrumors.com/attachment.php?attachmentid=164381&d=1238160682
And this thread, where I read it:
https://forums.macrumors.com/posts/7355635/
:"Hmm, spend millions of dollars to further optimize the already good OS for -15%, and dropping, of Macs, or make a streamlined OS for the +85%, that is only possible because the shackles of backwards-compatibility are removed? Decisions, decisions..."Very good point, but...
...this could be a deliberate choice: Apple could decide to limit the flow of bugs to those that they're most concerned about, so that they have a controlled amount of bug reports to sift through. We already know that they don't throw everything into the dev seeds right away. How else could one ever hope for unannounced features in the final product? Yet they do happen, don't they.
One thing I'll say for certain is that if PPC support is in the final product, the PPC whiners are going to take credit for it, as if Apple caved to the fury of irate posters on a rumors site. It will be just like how the iPhone jailbreakers love to take credit for the iPhone SDK. (As if.)
Nice link! I would be surprised if Apple left even 10%, let alone 15%, of their upgrade revenue sitting on the table when it's so easy to pick up.
With that mentality it is a mistake for Apple to release any retail version of Mac OS X. Why bother print up all those DVDs with updates? Everyone should just stick with the OS that came with the computer and whatever updates/bugfixes they get through Software Updates.
You should expect future support when you buy the OS as part of your system purchase. All the bug/security fixes should come with the software you acquired. If Apple supported OS bug fixes two operating systems back that might be reasonable. Apple can't do both though.
Not support bug fixes two OS back and not support keeping a window keep getting upgrades.
It is not the hardware that is the core issue here is the support for the software ( which is included in the price when buy the machine or buy a follow on update. ).
There might be less of an outcry if Apple came out and said would extend the window for Leopard longer. So that when the follow on to SnowLeopard comes that Leopard would get appropriate security fixes (or something for the folks they are about to cut short. ). They'd miss out on the extra goodies that SL has, but doesn't change the window on OS support they very reasonably have had when bought the machine.
If vintage/legacy come every 5 and OS come every 2 then there are two windows depending upon the overlap between the hardware and software cycles. The problem with this pace of every 2 years for the OS is that it is a bit less than 1/2 of the hardware primary lifetime. That leads to a reasonable expectation that there should be 2 upgrade opportunities before on unsupported software.
what is a "bonus" if they are paying for it? It isn't like the OS update is free. They drop cold hard cash for it, which means they paid for it. If they wre goddamn given rights you wouldn't have to pay for it.
If they knew in advance where the show stopper bugs were in the first place then why don't they fix them? The purpose of testing should be to seek out what you do NOT know about the correct functioning of the code. You can only be considered about show stopper bugs if you know what they are.
What you are talking about is more so a process in which they are testing to see just how many defects they can ship without people complaining greatly about it. (low defects on the primary targetd platform and the primary targeted apps ). Defects they mayalready know exist, but not so sure want to extend the release date to fix. Or defects can somewhat blow-off till after ship date that not particularly keen to find.
The volume problem you reference can be controlled by limiting your the size of the beta population so that you don't get lots of duplicate reports. However, if you make the beta population too small you won't get broad coverage to sniff out the bugs. If the objective is about quality you want to have the bugs in the bug tracking system; makes for a harder beta demographic selection problem. If you just want the bug code low enough to get over "ship fever" then you'll look to shotcut the breadth of the testing.
Also depends a bit on how much major defects only show up on one side of universal binary code. Only Apple would likely have the numbers on those. If the majority of the defects so up on both sides of Universal binary then only really need to build and ship out one half of the half of it to get good coverage. But again that goes back to careful selection of demographics for the beta.
PPC support isn't going to "wink in" at the very last moment. If it is not there on the DVD they hand out at WWDC and they hint at a release date in October (or earlier ) ... then likely not going to be there later. I think what will show up at WWDC will be the real beta (where beta means "feature complete but not free of defects.... please go find any remaining show stopper bugs... otherwise we are going to release this." )
Who says they aren't? Surely you don't think that dev seeds are their only source of testing. I'm sure they have vast labs full of diverse hardware exactly for this purpose. If I were them, I would measure the defect rate in the lab and only push things to dev seeds when I'm likely to only learn things that I don't already know.
Lol, no...what I'm talking about is software engineering.
On a side-note...people keep using the word 'beta'...is that the official designation of these "seeds", or are we all playing fast-and-loose with the language, here? If it's the latter, then it would probably be best to not make assertions based on the connotations of the word.
Approximately FIFTEEN PERCENT of active Macs are PPC, today, and dropping.
Check out this graph for 2009:
http://att.macrumors.com/attachment.php?attachmentid=164381&d=1238160682
And this thread, where I read it:
https://forums.macrumors.com/posts/7355635/
Depends if they want Snow Leopard to be a pure 64-bit OS release. Those numbers would be even more skewed toward Intel if threw out the non 64-bit machines on both sides. There are certainly more than an extremely small number of 32-bit Intel, but on the PowerPC side most likely the G4s are the very dominate chip present in that subcategory.
Dude, I am not talking about minor version updates. That’s reasonable expectation and common sense to expect future updates in that area. Nobody here is talking about some minor update. 10.5.x is always going to be free and Apple will continue to release a few more 10.5.x updates after Snow Leopard as well. Nobody is saying that Apple still stop supporting Leopard. What everybody is saying that they want Apple to support their machine in a NEW MAJOR release. That’s simply unreasonable. People should not have that kind of expectations at all in the first place. If people are going to spend $4,000 on a computer expecting it to be supported with a new MAJOR releases of their OS for the next 5 years, then they are making a major mistake and wasting their money, technology will be refreshed much faster than the software, and it would make far more sense to spend 2000$ on a machine and get a new one in 2 years with a new OS support than it does the other way. People who spend 4000$ on a computer are usually buying it for the hardware to do something that can be only be done with that specific hardware, not because of some OS features that may show up randomly in some major OS update 2-3 years down the way.
In this thread, people are talking about a major feature (OpenCL) that only comes in a major version release, Snow Leopard. Nobody who bought PPC machines 3-5 years ago, even thought about OpenCL or Snow leopard. To see people bitching about this crap, it is just retarded.
I would say the same thing about people complaining about Vista and W7 not working on their 3-5 years old PC. It doesn't matter that they bought a PC that was 5000$ 3 years ago, they should not ever expected that Vista and W7 to work on those machines. They should expect only that Microsoft would support XP with SP and minor updates for free and MS has been doing that for what now, 7-9 years? That was much nicer for MS to do that but they are only doing that because Vista turned out to be totally crap. And if Vista and W7 turns out that it'll work on their machines, that's a bonus and they will be happy for pay the cash for the new OS. That's the bonus part, not a right to expect every new OS to support their machines.
