Apple only cares about device drivers for the specific hardware it will be using in Macintosh branded computers because that's all Mac OS X is goign to run on. They are not releasing their software for white box PCs.
At some point, if Apple sales don't improve, Intel will make the same decisions that IBM made. "How much of our sales does Apple contribute to? 0.005%? We need them about as much as Wal-Mart needs Rubbermaid."
The whole point is that Apple's needs are aligned with where Intel is going anyways. That's the beauty of going with the commodity architecture.
The biggest issue with PowerPC is that Apple was the only real customer for comsumer machines using the architecture. IBM only uses it for their servers. Sony, Microsoft, and Nintendo are only using it for their console. And virtually everyone else is using it as an embedded processor. The market for x86 chips, on the other hand, is very much driven by the needs and trends of the consumer market.
Don't buy the "but Apple could have moved to Cell!" rubbish. All indications thus far are that it would be an exceptionally poor general purpose processor. The PPE core that the operating system would run on is far far slower than the existing G4 and G5 lines, despite the additional clock speed, due to significantly fewer execution units (2 on the PPE core v. 8 on the G5) and the lack of branch prediction on the chip. The "workstations" that have been mentioned in the past are most likely going to be heavily geared toward specific workloads.Cringly's "this isn't about technology" assertion really falls flat once you take that into account.
He also brings up the choice of Intel over AMD, which is not all that hard to understand either. Intel has massive massive fab capabilities, and is much less likely to have production issues. If they were going to use desktop chips in their initial production designs, this might not be a concern, but given that it seems pretty clear the first machines will be portables and consumer machines, they'll likely be using chips from the Pentium M line (yonah will be out then, and include the move to a 65nm process, sse3, and dual cores). Though AMD has a chip in this line (Turion) their limited penetration in the notebook market means fairly low production levels.
Of course, AMD probably did play a part in Apple's decision as well. If Intel ever fails to deliver on their roadmap, there's another major player in the market they can turn to.
The current single core Dothan clocked at 1.7GHz (Pentium-M 735) is rated at 21W. Documentation from IBM on the 970 core states that minimum core voltage on the 90nm unit is 1.0V, and quotes consumption of 27W when clocked at 1.25GHz.
Now back to real reality. Clockspeed doesn't mean anything on a modern CPU. IntOps, Flops, and SIMD mean everything. Oh, and having more gpr's than say, a $5 atmel AVR.
Good point. Of course you don't seem to get that it's in favour of the Pentium M, which is clocked slower than the G5, and is faster clock for clock.
I'd consider it less annoying than incredibly loud ring tone (as often as not crappy music) followed by HI JOSIE! OMG DID YOU HOOK UP WITH THAT GUY ROB AT THE HORSE'S ASS PUB LAST NIGHT? OMG HE HAD THE CUTEST BUTT! BLAH BLAH BLAH!.
[Yes, Mr. Lameness Filter. I *know* that all caps is like YELLING. That's the friggin' point in this case. s/Lameness/Lame/)
Texting is less disruptive for short self-contained communication; e.g. "heading to the office, be there in five", "don't forget to pick up the kids on your way home" or "be sure to drink your ovaltine".
The tap method is easier in a couple ways. The guide for the code is right on the keypad, for one; if you don't recall how to make an S, just look. And the sequence follows the alphabet, which most any native speaker will have learned when they were five or younger.
Honestly, SPARC and SPARC64 isn't really a great distinction for Sun hardware. You're better off looking at the architacture - sun4c, sun4m, sun4u, etc. There is a difference aside from the kernel here, the version of the ISA the code is optimized for.
SPARC v7, for example, lacks an integer multiply instruction, meaning it needs to be implemented in hardware. This is painfully slow, and Debian compiling for that target means their ssh implementation takes seven or eight times longer than it would if were compiled for SPARC v8.
Mind you, v8 is still 32-bit. It's not until sun4u (SPARC v9) that they went 64-bit, but that architecture was introduced about a decade ago. Anything else is going to be a major niche product at this point, and anyone who's displaced by this can either stick with Debian 3.0, install NetBSD, or go on Ebay and upgrade to a newer machine (seriously, you can pick up a sun4u machine for as little as $20 + s/h a lot of the time).
Why invest in an all new line using it's own developed Sun hardware based on AMD Opteron chips? Or the new SPARC Throughput Computing chips (http://www.sun.com/processors/throughput/)?
Why scrap years of development work on the UltraSparc V and Gemeni chips?
The whole "throughput computing chip" is just Sun going along with industry trends. Multi-core chips were in the works at Sun, Intel, IBM, and AMD long before Sun started pushing "throughput computing".
I won't make any bold predictions about Sun, but it'll be interesting to see where things go over the years. Focus on original Opteron designs and the resurrection of Solaris x86 make for interesting speculation with regard to transitions though.
Oddly, we haven't had a redundant power supply blow on us in years, but we've had them save our ass regardless because of loose cords, blown inverters, and sloppy electricians.
1.) Redhat. Slightly cheaper than windows, but as yet unproven for things like active directory integration. Unsure whether you'll still be able to get updates in 5 years (thanks redhat 9.0).
Red Hat Enterprise and Desktop products now come with a guaranteed support cycle. The current version is set for maintainance until 2012. That is what the subscription cost is paying for.
That convention is strictly American, and is often eschewed in technical and scientific circles for reasons of clarity.
Adoption of this convention is a historical accident to begin with. Commas and periods were susceptible to becoming disloged in early typesetting equipment, so printers started tucking them inside quotes and parantheses when they could.
Poor black levels compared to other LCDs, certainly. Poor black levels compared to a CRT, on the other hand, are a hurdle that hasn't been overcomed. Even the best LCDs aren't going to do much better than 0.5 cd/m^2, whereas CRTs can be had with levels as low as 0.01 cd/m^2.
The only caveat is that CRTs are more susceptible to ambient light, so the advantage "disappears" in certain environments.
OLED having good contrast does fuck all for LCD....they just have phospor for burn-in.
Monitors haven't been particularly susceptible to burn-in since the monochrome days.
I suppose they're 'perfectly fine' in the same way that 19th-century telephone voice quality and NTSC video are 'perfectly fine'.
Both of those are demonstratably inferior by pretty much every metric to newer technologies. In the case of LCD and CRT there are distinct advantages to both.
Sure, LCD has been closing the gap on it's deficiencies for a while now, but thin-tube CRT are right around the corner now, from several manufacturers. And you had additional competition from OLED, especially on small screen applications.
None of these technologies are going anywhere for quite some time.
LCDs still have comparatively poor black levels, contast, and viewing angles compared to CRTs. Some people don't seem to mind, but even decent ones are damn hard on my eyes over time.
Unfortunately, there aren't any really good measures of x86 to PPC performance out there, that I'm aware. Some exceedingly amatuer treatments posted on the web here and there, and the ones that Apple has been publishing regularly since the G3 days. The Apple benchmarks are dubious insofar as they eschew any established industry accepted metrics of performance, and focus on a limited set of applications.
Look at their choices - Photoshop CS, Aftereffects, Lightwave 3D. And in two of the tests they used different software on the PC and Mac - a beta version Final Cut Pro (Mac) vs a release version of Adobe Premier (PC), and Logic Pro (Mac) vs teinberg Cubase SX (PC). One glaring common factor here is that all the applications are ripe for optimization using SIMD instructions, which is an Apple strong point.
Well, for starters, a typical LCD draws about 10 watts, so that's draw apparently is a wash.
The next question is - which Celeron? Assuming you don't have a cheapie "desktop processor in a notebook" model, that still leaves a range from 5W (90nm 800MHz Celeron-M ULV) to 35W (130nm 2.5GHz Mobile Celeron).
Hard drive performance at this point is really a matter of what you're willing to spend. Just looking at Seagate's offerings, the difference between their 4200RPM drive and 7200RPM is a whopping.3 watts.
A 1.4GHz Pentium M is speced to draw a peak of 10W, considerably less than the 1.4GHz G4 used in the "high end" Mac Mini. Even if you bump up to a 2.1GHz chip, peak power consumption is 21W - still at least as good as the G4 (haven't found peak draw figures, but the reported "typical" figure is 19W).
As per IBM: "Single precision floating point computation is geared for throughput of media and 3D graphics objects."
The problem isn't that single precision isn't adequate for a lot of tasks. The problem is that single percision mode sacrifices IEEE compliance in order to maximize throughput for the targeted application.
I wouldn't be surprised if they go on to release a different impelementation with some or all SPUs geared toward more general purpose tasks, but it won't have the raw throughput of the current chip.
Geodes are pretty much aimed only at embedded applications rather than desktops, and Transmeta is transitioning out of the hardware business.
Popularity of Pentium-M as a desktop chip (30W, 14W for the low voltage chip) is on the rise, and their ULV chips (7W) are gaining popularity in the SBC market.
Via's line of Eden and C3 processors are also worthy units, with worst case draw ranging from 2.5W to 14W. The upcoming "Luke" chip, which combines the CPU and north bridge on the same die, is currently running at 10W, but is expected to drop to 3.5W when they move to a 90nm process. And that's with an integrated graphics process and MPEG-2/4 acceleration.
Slashdot Mirror
Apple only cares about device drivers for the specific hardware it will be using in Macintosh branded computers because that's all Mac OS X is goign to run on. They are not releasing their software for white box PCs.
I'm not the original poster, but how about Apple as a source?
At some point, if Apple sales don't improve, Intel will make the same decisions that IBM made. "How much of our sales does Apple contribute to? 0.005%? We need them about as much as Wal-Mart needs Rubbermaid."
The whole point is that Apple's needs are aligned with where Intel is going anyways. That's the beauty of going with the commodity architecture.
The biggest issue with PowerPC is that Apple was the only real customer for comsumer machines using the architecture. IBM only uses it for their servers. Sony, Microsoft, and Nintendo are only using it for their console. And virtually everyone else is using it as an embedded processor. The market for x86 chips, on the other hand, is very much driven by the needs and trends of the consumer market.
Don't buy the "but Apple could have moved to Cell!" rubbish. All indications thus far are that it would be an exceptionally poor general purpose processor. The PPE core that the operating system would run on is far far slower than the existing G4 and G5 lines, despite the additional clock speed, due to significantly fewer execution units (2 on the PPE core v. 8 on the G5) and the lack of branch prediction on the chip. The "workstations" that have been mentioned in the past are most likely going to be heavily geared toward specific workloads.Cringly's "this isn't about technology" assertion really falls flat once you take that into account.
He also brings up the choice of Intel over AMD, which is not all that hard to understand either. Intel has massive massive fab capabilities, and is much less likely to have production issues. If they were going to use desktop chips in their initial production designs, this might not be a concern, but given that it seems pretty clear the first machines will be portables and consumer machines, they'll likely be using chips from the Pentium M line (yonah will be out then, and include the move to a 65nm process, sse3, and dual cores). Though AMD has a chip in this line (Turion) their limited penetration in the notebook market means fairly low production levels.
Of course, AMD probably did play a part in Apple's decision as well. If Intel ever fails to deliver on their roadmap, there's another major player in the market they can turn to.
The logo changes never really register for me unless someone points them out. I only noticed the current one after a friend pointed it out on IRC.
The current single core Dothan clocked at 1.7GHz (Pentium-M 735) is rated at 21W.
Documentation from IBM on the 970 core states that minimum core voltage on the 90nm unit is 1.0V, and quotes consumption of 27W when clocked at 1.25GHz.
Now back to real reality. Clockspeed doesn't mean anything on a modern CPU. IntOps, Flops, and SIMD mean everything. Oh, and having more gpr's than say, a $5 atmel AVR.
Good point. Of course you don't seem to get that it's in favour of the Pentium M, which is clocked slower than the G5, and is faster clock for clock.
Doh. s/hardware/software/
I'd consider it less annoying than incredibly loud ring tone (as often as not crappy music) followed by HI JOSIE! OMG DID YOU HOOK UP WITH THAT GUY ROB AT THE HORSE'S ASS PUB LAST NIGHT? OMG HE HAD THE CUTEST BUTT! BLAH BLAH BLAH!.
[Yes, Mr. Lameness Filter. I *know* that all caps is like YELLING. That's the friggin' point in this case. s/Lameness/Lame/)
Texting is less disruptive for short self-contained communication; e.g. "heading to the office, be there in five", "don't forget to pick up the kids on your way home" or "be sure to drink your ovaltine".
The tap method is easier in a couple ways. The guide for the code is right on the keypad, for one; if you don't recall how to make an S, just look. And the sequence follows the alphabet, which most any native speaker will have learned when they were five or younger.
Honestly, SPARC and SPARC64 isn't really a great distinction for Sun hardware. You're better off looking at the architacture - sun4c, sun4m, sun4u, etc. There is a difference aside from the kernel here, the version of the ISA the code is optimized for.
SPARC v7, for example, lacks an integer multiply instruction, meaning it needs to be implemented in hardware. This is painfully slow, and Debian compiling for that target means their ssh implementation takes seven or eight times longer than it would if were compiled for SPARC v8.
Mind you, v8 is still 32-bit. It's not until sun4u (SPARC v9) that they went 64-bit, but that architecture was introduced about a decade ago. Anything else is going to be a major niche product at this point, and anyone who's displaced by this can either stick with Debian 3.0, install NetBSD, or go on Ebay and upgrade to a newer machine (seriously, you can pick up a sun4u machine for as little as $20 + s/h a lot of the time).
Why invest in an all new line using it's own developed Sun hardware based on AMD Opteron chips? Or the new SPARC Throughput Computing chips (http://www.sun.com/processors/throughput/)?
Why scrap years of development work on the UltraSparc V and Gemeni chips?
The whole "throughput computing chip" is just Sun going along with industry trends. Multi-core chips were in the works at Sun, Intel, IBM, and AMD long before Sun started pushing "throughput computing".
I won't make any bold predictions about Sun, but it'll be interesting to see where things go over the years. Focus on original Opteron designs and the resurrection of Solaris x86 make for interesting speculation with regard to transitions though.
Oddly, we haven't had a redundant power supply blow on us in years, but we've had them save our ass regardless because of loose cords, blown inverters, and sloppy electricians.
Via isn't an IBM subsidiary, though Via is contracting IBM to fab their chips.
we can already see AMD moving away with their 64-bit systems
Erm, embracing and extending the x86 ISA is moving away from it?
1.) Redhat. Slightly cheaper than windows, but as yet unproven for things like active directory integration. Unsure whether you'll still be able to get updates in 5 years (thanks redhat 9.0).
Red Hat Enterprise and Desktop products now come with a guaranteed support cycle. The current version is set for maintainance until 2012. That is what the subscription cost is paying for.
That convention is strictly American, and is often eschewed in technical and scientific circles for reasons of clarity.
Adoption of this convention is a historical accident to begin with. Commas and periods were susceptible to becoming disloged in early typesetting equipment, so printers started tucking them inside quotes and parantheses when they could.
Poor black levels compared to other LCDs, certainly. Poor black levels compared to a CRT, on the other hand, are a hurdle that hasn't been overcomed. Even the best LCDs aren't going to do much better than 0.5 cd/m^2, whereas CRTs can be had with levels as low as 0.01 cd/m^2.
The only caveat is that CRTs are more susceptible to ambient light, so the advantage "disappears" in certain environments.
Only until OLED displays are widely available.
...they just have phospor for burn-in.
OLED having good contrast does fuck all for LCD.
Monitors haven't been particularly susceptible to burn-in since the monochrome days.
I suppose they're 'perfectly fine' in the same way that 19th-century telephone voice quality and NTSC video are 'perfectly fine'.
Both of those are demonstratably inferior by pretty much every metric to newer technologies. In the case of LCD and CRT there are distinct advantages to both.
Sure, LCD has been closing the gap on it's deficiencies for a while now, but thin-tube CRT are right around the corner now, from several manufacturers. And you had additional competition from OLED, especially on small screen applications.
None of these technologies are going anywhere for quite some time.
LCDs still have comparatively poor black levels, contast, and viewing angles compared to CRTs. Some people don't seem to mind, but even decent ones are damn hard on my eyes over time.
Unfortunately, there aren't any really good measures of x86 to PPC performance out there, that I'm aware. Some exceedingly amatuer treatments posted on the web here and there, and the ones that Apple has been publishing regularly since the G3 days. The Apple benchmarks are dubious insofar as they eschew any established industry accepted metrics of performance, and focus on a limited set of applications.
Look at their choices - Photoshop CS, Aftereffects, Lightwave 3D. And in two of the tests they used different software on the PC and Mac - a beta version Final Cut Pro (Mac) vs a release version of Adobe Premier (PC), and Logic Pro (Mac) vs teinberg Cubase SX (PC). One glaring common factor here is that all the applications are ripe for optimization using SIMD instructions, which is an Apple strong point.
Well, for starters, a typical LCD draws about 10 watts, so that's draw apparently is a wash.
.3 watts.
The next question is - which Celeron? Assuming you don't have a cheapie "desktop processor in a notebook" model, that still leaves a range from 5W (90nm 800MHz Celeron-M ULV) to 35W (130nm 2.5GHz Mobile Celeron).
Hard drive performance at this point is really a matter of what you're willing to spend. Just looking at Seagate's offerings, the difference between their 4200RPM drive and 7200RPM is a whopping
A 1.4GHz Pentium M is speced to draw a peak of 10W, considerably less than the 1.4GHz G4 used in the "high end" Mac Mini. Even if you bump up to a 2.1GHz chip, peak power consumption is 21W - still at least as good as the G4 (haven't found peak draw figures, but the reported "typical" figure is 19W).
Jackass.
As per IBM: "Single precision floating point computation is geared for throughput of media and 3D graphics objects."
The problem isn't that single precision isn't adequate for a lot of tasks. The problem is that single percision mode sacrifices IEEE compliance in order to maximize throughput for the targeted application.
I wouldn't be surprised if they go on to release a different impelementation with some or all SPUs geared toward more general purpose tasks, but it won't have the raw throughput of the current chip.
Geodes are pretty much aimed only at embedded applications rather than desktops, and Transmeta is transitioning out of the hardware business.
Popularity of Pentium-M as a desktop chip (30W, 14W for the low voltage chip) is on the rise, and their ULV chips (7W) are gaining popularity in the SBC market.
Via's line of Eden and C3 processors are also worthy units, with worst case draw ranging from 2.5W to 14W. The upcoming "Luke" chip, which combines the CPU and north bridge on the same die, is currently running at 10W, but is expected to drop to 3.5W when they move to a 90nm process. And that's with an integrated graphics process and MPEG-2/4 acceleration.