G5 Processor and Bus Slewing

[MacRumors]

A Spymac thread (http://www.spymac.com/forums/showthread.php?threadid=38381) posted this weekend reveals some screenshots and information about a feature of the new PowerMac G5's called "Slewing".

The feature is located in the Energy Saver Control Panel and allows the user to make some modifications to the Processor and Bus performance settings. A concept that is familiar to laptop owners... the G5 can be run in Reduced, Highest and Automatic performance settings.

Details on the various modes are provided in Apple's Developer Documentation (http://developer.apple.com/documentation/Hardware/Developer_Notes/Macintosh_CPUs-G5/PowerMacG5/1Introduction/chapter_2_section_4.html).


To lower power consumption, heat generation, and fan noise, the Power Mac G5 computer incorporates an automatic power management technique called bus slewing. Bus slewing is designed to run at high processor and bus speeds and high voltage when the demand on the processor is high, and to run at low processor and bus speeds and low voltage when the demand on the processor is low.

Speed ranges involved start from 1.3GHz up to the highest rated GHz of the chip (1.6, 1.8, or 2.0). The default slewing option for the PowerMac G5s is Automatic, and the processor and bus speeds are modulated seamlessly to the user.

According to our contacts, there should be no performance loss in Automatic mode.

That being said, at least one user claims an 11 point increase in their Xbench score after changing their Slewing mode from Automatic to Highest. Readers are reminded that XBench (http://www.xbench.com) has been inconsistent in producing reproducible benchmark numbers -- even on the same machine. MacRumors' reader 1stunna (http://forums.macrumors.com/showthread.php?s=&postid=467999&highlight=xbench#post467999) managed to get a 11 point increase in Xbench scores by simply rerunning the test three times, with no other changes to the system. Whether this represents test-to-test variability or a side-effect of the processor automatically ramping up is still up for debate.

Return to Mac Rumors.com (http://www.macrumors.com)

[SeaFox]

As long as "Slewing" doesn't become an auto activating thing in the G5 laptops, like the "Stepping" of Pentium 4M's when you run them on battery power instead of house current, I'll be happy.

It should stay a user control.

Hey! I got to do the first post for this article!

[Makosuke]

Kinda cool to see power saving features like this on a desktop. Aside from making the computer quieter in general (less heat = less fans = less noise), anything that saves the world some power is good for the user's wallet and for the world's oil supplies. Yeah, it's tiny, but if you can do it, why not.

Go Apple efficiency!

[VicMacs]

does this mean that the G5 is loud when running on "highest"? my windtunnel is pretty loud... and although it doesnt bother me I would like to see G5s that are hardly noticeable... I wonder why this option is here... is the G5 a powerjunkie?

WHERE IS PANTHER!?

[scem0]

Well, its good that the option is there.

Im glad apple is making OS X and its connection to the hardware even more customizable.

scem0

[foniks2020]

DO benchmarks take into consideration Cacheing by either processor or hard drive? If they do it is easy to imagine a benchmark showing higher performance after running the test several times. Maybe I misunderstand how cacheing works though.... any comments on that one?

[theipodgod16]

makes logical sence, sort of similar to turning off lights in your house when you arnt in the room huh?

[tduality]

I take the bus slewing as a feature primarily for laptops. But why not use it for the desktops as well since it's there anyway.

Will this be a case of 'premature powerbook rumor of the week'?

[Porsche]

Originally posted by tduality
I take the bus slewing as a feature primarily for laptops. But why not use it for the desktops as well since it's there anyway.

That's my take on it.

[Sol]

I do not see the point of this feature on a desktop system. The G5 PowerMacs are supposed to be quieter than the G4 PowerMacs, even with nine fans spinning at full speed. How much difference can Slewing make?

Having said that, it would be nice to have the option to over-clock the processors from the Energy Saver preference pane. Surely the soundtrack of Doom 3, UT 2003, etc would drown out the noise from the fans in the computer.

[hvfsl]

I wonder if that means there is a way to over-clock the new G5s in the OS, by flashing with a modified bios or something. The G5 can handle upto 1Ghz system bus, so it would be good to overclock the 1.6 and 1.8 versions to 1Ghz bus.

[adamfilip]

yeah thats sorta what i was thinking.. cant wait until some hack comes out that lets you adjust the cpu and fsb speed..

i bet the 1.6 can run at 2ghz..

[strangelogic]

Originally posted by Macrumors
Readers are reminded that XBench (http://www.xbench.com) has been inconsistent in producing reproducible benchmark numbers -- even on the same machine. MacRumors' reader 1stunna (http://forums.macrumors.com/showthread.php?s=&postid=467999&highlight=xbench#post467999) managed to get a 11 point increase in Xbench scores by simply rerunning the test three times, with no other changes to the system.
There has been another point release of Xbench since that was done. It seems to be much more consistent now in version 1.1.1

[cb911]

we've known about this feature for a while, right? so now i guess it's confirmed and it's got a name! Slewing... yeah! :D

[moki]

Originally posted by strangelogic
There has been another point release of Xbench since that was done. It seems to be much more consistent now in version 1.1.1

There are two types of useful benchmarks. One type tests the raw processor performance; but to be used effectively, you'd need to bootstrap the code into running when literally nothing else is running on the machine in question. Obviously this isn't too practical; booting to single user mode and running benchmarks via command line at least controls some of the variables.

Otherwise, you're testing not just the processor's performance on the benchmark, but also the overhead of the OS. You can come *close* to this on Mach by running the benchmark in the realtime band, but you can't eliminate the OS overhead entirely.

The other kind of useful benchmark is running real-world applications on a shipping OS, to test overall system performance. For testing to be meaningful, the RAM and running applications/system version of all of the machines need to be identical, and the tests need to be done at least a dozen times, taking the mean of the results.

Sadly, it appears that Xbench is attempting to run processor benchmarks in real-world conditions, which isn't incredibly useful. To make matters worse, people compare performance on very different computer configurations, under non-controlled conditions.

[Rincewind42]

Originally posted by VicMacs
does this mean that the G5 is loud when running on "highest"? my windtunnel is pretty loud... and although it doesnt bother me I would like to see G5s that are hardly noticeable... I wonder why this option is here... is the G5 a powerjunkie?

No, the G5 is not loud when running on highest. Not unless you have 3 PCI-X cards running maxed out, maxed out RAM, your video card going full bore, both HDs working like crazy and your burning a DVD at the same time. Otherwise it may be slightly louder than using automatic.

WHERE IS PANTHER!?

Ask Think Secret. They seem to know when a new seed is dropped before thus of us that have seed access.

[Rincewind42]

Originally posted by Sol
I do not see the point of this feature on a desktop system. The G5 PowerMacs are supposed to be quieter than the G4 PowerMacs, even with nine fans spinning at full speed. How much difference can Slewing make?

Actually, with all nine fans running at full speed the G5 isn't any quieter than the windtunnels. That said, you shouldn't hear all the fans going at full speed unless your sucking down a lot of power in a hot environment

Having said that, it would be nice to have the option to over-clock the processors from the Energy Saver preference pane. Surely the soundtrack of Doom 3, UT 2003, etc would drown out the noise from the fans in the computer.

Apple will never let you overclock from the energy saver control panel. That wouldn't be good for their business model :D

[Rincewind42]

Originally posted by hvfsl
I wonder if that means there is a way to over-clock the new G5s in the OS, by flashing with a modified bios or something. The G5 can handle upto 1Ghz system bus, so it would be good to overclock the 1.6 and 1.8 versions to 1Ghz bus.

No it wouldn't. The 1.6 & 1.8 CPUs can't handle a 1Ghz bus - it is too high for their timings. (the best timing on the G5 is 2:1, so the highest bus is half the clock). And if you do try to overclock the CPU too, your talking about 2 chips that could fail from this experiment. I wouldn't risk it on a brand new $2K+ machine...

[Dave K]

Originally posted by Sol
I do not see the point of this feature on a desktop system. The G5 PowerMacs are supposed to be quieter than the G4 PowerMacs, even with nine fans spinning at full speed. How much difference can Slewing make? This feature has been known about since WWDC:

http://arstechnica.infopop.net/OpenTopic/page?a=tpc&s=50009562&f=8300945231&m=9080959175&r=9000981275#9000981275

Relevent bit:Normally, the machines are running at about 2/3 their total clock speed (for 2GHz machines, this is 1.4GHz), this jumps up to the full speed whenever it's required. The ramp time up or down is ~1ms, but the CPU is running normally during this time, so there is no performance "hiccup". This results in about 60% power/heat savings, which jumps up to about 85% savings if the machine is idle and they "turn on other power saving features". When idle, the CPU fans are barely turning.Being able to run the box with nearly fanless is one of the primary ways they keep it so quiet.

But it also means that if you're testing, you may as well throw the first test out if you're testing from Idle.

[Photorun]

Another thing to consider is, though it may be pretty miniscule, having the computer cycle down it's power would mean a slightly lower energy bill.

[Snowy_River]

Originally posted by moki
...For testing to be meaningful, the RAM and running applications/system version of all of the machines need to be identical, and the tests need to be done at least a dozen times, taking the mean of the results....

I have to beg an exception to this. It seems that you're saying that comparing two computers that are running different operating systems (i.e. WinXP and OS X) is not a meaningful test. While there are, of course, other factors to consider when comparing two such systems, it's still a very meaningful comparison to see the relative performance of a piece of software (i.e. Photoshop) on these two different systems...

[MacBandit]

Sounds to me we have found the cause of the inconsistent benchmark results. We all know that even if something is supposed to happen instantaneously there is always a microscopic amount of delay. Also with benchmarking apps such as XBench the faster the computer is the faster it will perform the overall test and therefore the shorter amount of time the test is being performed.

What this means is if you have a computer that is capable of running a test in 1/10 of a second the results will be much less accurate then a test that takes a minute or an hour for example because there is less time to take fluctuations into account and average the result.

So with the G5s we have a double problem. Not only is it performing many of the tasks very quickly but it also has to ramp up the processor speed in the process. The two put together are probably bringing the overall results down quite a bit.

Just for comparison most of us know that even if a process is made nice +20 it can have an impact of overall processing performance on another program that should be able to take all the processing time. This is because the process being ran isn't always consistently taking all processing time and the nice +20 process is trying to take some of the wasted processing time for itself. In the process the system can't possibly compensate fast enough to keep the nice +20 process from having an affect on the normal process. Also at the same time the system itself is having to take some more processing time because it is trying to compensate for the two apps and there processing time. This is the same problem with the CPU slewing and will of course have a much greater affect on a process that only takes a minute amount of time to process.

[zach]

Hunh. That seems pretty cool.

I have an iBook, and honestly, when I turn down the processer performance, the only thing it seems to cut down on is noise and heat. The computer still seems to perform just as well, so if slewing can cut down on noise, well, that is just another reason to buy a G5.

[hvfsl]

Originally posted by Rincewind42
No it wouldn't. The 1.6 & 1.8 CPUs can't handle a 1Ghz bus - it is too high for their timings. (the best timing on the G5 is 2:1, so the highest bus is half the clock). And if you do try to overclock the CPU too, your talking about 2 chips that could fail from this experiment. I wouldn't risk it on a brand new $2K+ machine...

That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed. Ati used to do that on their video cards (have memory 2x clock speed) but it caused problems when they increased the gpu clock because they had to use faster (and more expensive) memory. So now Ati does the same as nvidia by allowing any ratio between gpu speed and mem. So if it is indeed true, Apple/IBM will have to redesign the motherboard/chip or they will have problems getting the cpu speed up.

I also would not risk modding a new machine but maybe in a few years someone will get a cheap G5 to see if they can over-clock it. There are already people that have done it to their PB G4s. I saw one person on www.xlr8yourmac.com who over-clocked their PB G4 550 to 667Mhz. Apparently a lot of the 550's were 667's with a bus speed of 100 instead of 133.

[Catfish_Man]

Originally posted by hvfsl
That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed. Ati used to do that on their video cards (have memory 2x clock speed) but it caused problems when they increased the gpu clock because they had to use faster (and more expensive) memory. So now Ati does the same as nvidia by allowing any ratio between gpu speed and mem. So if it is indeed true, Apple/IBM will have to redesign the motherboard/chip or they will have problems getting the cpu speed up.

I also would not risk modding a new machine but maybe in a few years someone will get a cheap G5 to see if they can over-clock it. There are already people that have done it to their PB G4s. I saw one person on www.xlr8yourmac.com who over-clocked their PB G4 550 to 667Mhz. Apparently a lot of the 550's were 667's with a bus speed of 100 instead of 133.

The G5 supports 2:1, 3:1, and 6:1 iirc. It may support 4:1, but I can't remember.

[WM.]

Originally posted by hvfsl
That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed.
The G5's CPU:bus ratio can be 2:1, 3:1, 4:1, or 6:1, IIRC. (See the ArsTechnica Q&A with the IBM guys about the 970 to make sure.)

So when we reach 3 GHz, we could still have a 1 GHz bus, no problem.

HTH
WM

P.S. Aaack, I see that Catfish_Man beat me to it! :)

[stingerman]

Originally posted by hvfsl
That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed. Ati used to do that on their video cards (have memory 2x clock speed) but it caused problems when they increased the gpu clock because they had to use faster (and more expensive) memory. So now Ati does the same as nvidia by allowing any ratio between gpu speed and mem. So if it is indeed true, Apple/IBM will have to redesign the motherboard/chip or they will have problems getting the cpu speed up.

I also would not risk modding a new machine but maybe in a few years someone will get a cheap G5 to see if they can over-clock it. There are already people that have done it to their PB G4s. I saw one person on www.xlr8yourmac.com who over-clocked their PB G4 550 to 667Mhz. Apparently a lot of the 550's were 667's with a bus speed of 100 instead of 133.

I believe the 970 allows a ratio all the way to 12:1, Apple decided to go for the max of 2:1 and they developed the controller to handle it (U3). Apple's firmware is in its software, so you can overclock, for example, the G4 AGP bus from 100Mhz to 133. But, since the processor basis its core clock as a multiple of the bus, you also need to be able to adjust the ratio. i believe on AGP 500 forward G4's, Apple hardwires the ratio pins on the processor daughter cards. You need to use a soldering iron to adjust them (not for the novice or feint of heart.) The Gigadesign 1 GHz upgrade (7455A processor) daughtercard gives you jumpers to adjust and easily overclock it (up to 1.3 GHz).

The question is whether Apple hardwires the 970's ratio or not. I am hoping not, but we will see. I have a feeling Apple wants to encourage the hobbyist community to adopt the G5 and with a little Forth knowledge overclocking should be a breeze.

[Mav451]

I was wondering how the G5 could run so quiet and this is the answer. It isn't running at 2ghz all the time, so why bother having the fans running at their highest. With any modern cpu, reducing clockspeed by 20-50% can have dramatic heat-reducing effects.

I know if you underclock the AMD thunderbirds they definitely run alot cooler--with the 1700+ if i was running @ 1ghz, i doubt it would even break 30C's...with a bad cooler. If it was running a "silent" SLK800 setup, it'd prolly hit even 27C's (80 F)--that's the territory of motherboard/case temperatures normally.

[Rocketman]

Bus Slewing seems to be introducing a feature that will be used with a variety of technologies including the Front Side Bus, the Interconnect, and possibly in the future memory, as the spaces merge.

[Phil Of Mac]

Originally posted by VicMacs
WHERE IS PANTHER!?

Avie Tevanian's hard drive.

[Abstract]

Originally posted by tduality
I take the bus slewing as a feature primarily for laptops. But why not use it for the desktops as well since it's there anyway.

Will this be a case of 'premature powerbook rumor of the week'?

Yup, that's what I think too. Its a feature built-in already, so why would they remove it for desktops? They wouldn't/didn't, and so its there for all of us to see. However, this is more of a laptop thang than anything else.

[MacBandit]

Originally posted by hvfsl
That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed. Ati used to do that on their video cards (have memory 2x clock speed) but it caused problems when they increased the gpu clock because they had to use faster (and more expensive) memory. So now Ati does the same as nvidia by allowing any ratio between gpu speed and mem. So if it is indeed true, Apple/IBM will have to redesign the motherboard/chip or they will have problems getting the cpu speed up.

I also would not risk modding a new machine but maybe in a few years someone will get a cheap G5 to see if they can over-clock it. There are already people that have done it to their PB G4s. I saw one person on www.xlr8yourmac.com who over-clocked their PB G4 550 to 667Mhz. Apparently a lot of the 550's were 667's with a bus speed of 100 instead of 133.

As stated by other the FSB speed isn't stuck at 2:1 though it really wouldn't matter a lot if it was. I don't think they would have a lot of problems getting the FSB to run at higher speeds. Remember we aren't talking about the system bus here and the current system bus already runs slower then the FSB.

[adamfilip]

doesnt the power supply still convert.. 350 watts (of whatever it is) regardless of how much power the system actually uses.

so even if the chips slow down.. there isnt any power savings.. from the wall?

[Phil Of Mac]

Originally posted by Abstract
Yup, that's what I think too. Its a feature built-in already, so why would they remove it for desktops? They wouldn't/didn't, and so its there for all of us to see. However, this is more of a laptop thang than anything else.

This indicates to me that there may be G5 PowerBooks soon enough to implement this feature now, and not later.

[daveL]

Originally posted by adamfilip
doesnt the power supply still convert.. 350 watts (of whatever it is) regardless of how much power the system actually uses.

so even if the chips slow down.. there isnt any power savings.. from the wall?
No, the power supply doesn't run at it's maximum rating all the time. The system draws whatever amount of power it needs at the moment, up to the maximum rating of the supply.

[Fukui]

Originally posted by hvfsl
That can't be true or the 3Ghz G5s that come next year will have a 1.5Ghz bus. If it is true then it looks like we will have even more problems getting the G5 Mhz than the G4. This is because it is harder to increase the bus speed than the cpu speed. Ati used to do that on their video cards (have memory 2x clock speed) but it caused problems when they increased the gpu clock because they had to use faster (and more expensive) memory.

Unless they go for DDR2 (500MHZ) 3-way interleaving:D

[moki]

Originally posted by Snowy_River
I have to beg an exception to this. It seems that you're saying that comparing two computers that are running different operating systems (i.e. WinXP and OS X) is not a meaningful test. While there are, of course, other factors to consider when comparing two such systems, it's still a very meaningful comparison to see the relative performance of a piece of software (i.e. Photoshop) on these two different systems...

No, I agree with you that this is useful and meaningful. But it is more meaningful when the conditions are controlled.

My comments were directed at the Xbench scores.

[MacBandit]

Originally posted by Fukui
Unless they go for DDR2 (500MHZ) 3-way interleaving:D

As I stated previously the current memory/system bus architecture doesn't match the FSB speed. The system bus is the one that is hard to scale to faster speeds as far as I know it's much easier to do with the FSB if the CPU was designed in that way to start with. The current system bus is only 400MHz but it is in affect what the PC people called double pumped as it runs at Double Data Rate speeds reading on the rise and fall of the clock cycle. This only gets it to 800MHz so it is still 200MHz short of the Dual/2GHz's 1GHz FSB. Also since there is only one system bus and 2 cpus the system bus is in affect 1.2GHz slower then the FSB total capacity on paper. In the real world this makes little difference as the CPUs never stream at there maximum potential only utilizing the maximum throughput in short bursts. Also there is no reason that a 1.5GHz FSB on a 3GHz CPU couldn't use the same system bus at 400MHz DDR that the current system uses. I don't think it would show up as much of a bottle neck especially if the next generation of PPC970s has a larger cache.

[tduality]

Originally posted by VicMacs
WHERE IS PANTHER!?

Originally posted by Phil Of Mac
Avie Tevanian's hard drive.

Hope there's also a backup somewhere. :D :p

[grahamtriggs]

Originally posted by Rincewind42
No it wouldn't. The 1.6 & 1.8 CPUs can't handle a 1Ghz bus - it is too high for their timings. (the best timing on the G5 is 2:1, so the highest bus is half the clock). And if you do try to overclock the CPU too, your talking about 2 chips that could fail from this experiment. I wouldn't risk it on a brand new $2K+ machine...

Eh? If you meant to say that you couldn't run a 1.6 or 1.8 at their rated speeds with a 1Ghz bus, then fine.

But the 1.6, 1.8 and 2Ghz chips are all identical in manufacture, and are all being run on a 2:1 ratio. The only barrier to running a 1.6 or 1.8 with a 1Ghz FSB (and hence at 2Ghz core), is the quality of the chip itself.

As I said, they will all be manufactured to theoretically be the same... but some chips will fail at 2Ghz and get marked down to 1.6 or 1.8... if yields are especially good, then numbers of 2Ghz chips will simple be marked down to 1.6 or 1.8 to fulfil demand (there have been some very notable cases of this in Intel land, for example, the 300Mhz Celeron was shipped for running on a 66Mhz FSB with a 4.5:1 ratio - you would have been *very* unlucky though to get a chip that wasn't able to run perfectly with a 100Mhz FSB, and hence 450Mhz core).

[zac4mac]

Simple bus speed dinking will usually get you a little speed boost with no trouble. I've been running my DP500 at 550 since it was a month old; for three years, practically 24/7. I resoldered the jumpers on the proc daghter card from 5X to 5.5X. Only downtime has been a P/S fan(bearings went out in 2001) and the whole P/S, this past spring.
I trust Michiro Isobe will find a way to squeeze the G5 for a little extra.

[Rincewind42]

Originally posted by grahamtriggs
Eh? If you meant to say that you couldn't run a 1.6 or 1.8 at their rated speeds with a 1Ghz bus, then fine.

But the 1.6, 1.8 and 2Ghz chips are all identical in manufacture, and are all being run on a 2:1 ratio. The only barrier to running a 1.6 or 1.8 with a 1Ghz FSB (and hence at 2Ghz core), is the quality of the chip itself.

1) Yes, you can't run a 1.6 or 1.8 (at those speeds) on a 1Ghz FSB. Those are both unsupported ratios.

2) There are two barriers to running the FSB at a rating greater than the one it shipped at - the CPU (as you say, not all 1.6/1.8 parts can run at 2.0Ghz) and the U3 controller. You can bet that not all U3s can run a 1Ghz FSB, as it's just as complicated a hunk of silicon as the 970 is. There are a lot of technological hurdles to running a bus at 1Ghz!

You can bet that the U3 is on the edge of system controller manufacture - remember that Intel's controllers actually run 200Mhz quad-pumped while the U3 runs at up to 500Mhz double-pumped - there is a big engineering between 200 & 500Mhz, but not much of one between double & quad pumped. All that says that the U3 has to be verified just as much as the 970 does for the speeds it is to run at.

[WM.]

Originally posted by MacBandit
As I stated previously the current memory/system bus architecture doesn't match the FSB speed. The system bus is the one that is hard to scale to faster speeds as far as I know it's much easier to do with the FSB if the CPU was designed in that way to start with. The current system bus is only 400MHz but it is in affect what the PC people called double pumped as it runs at Double Data Rate speeds reading on the rise and fall of the clock cycle. This only gets it to 800MHz so it is still 200MHz short of the Dual/2GHz's 1GHz FSB.
"System bus" really isn't the correct term here, since in the case of the G5 we're only talking about the bus between the U3 controller and RAM. I'd probably refer to it as the memory bus or the RAM bus instead.

Also, to be nit-picky, it doesn't run at 400 MHz DDR. Rather, it's like this: 200 MHz x 2 (DDR) x 2 (two banks of RAM) = 800 MHz.

Finally, when you say that it's hard to scale the "system bus" (meaning the RAM bus), I would add that it's basically impossible--because the RAM will never run faster than 200 MHz x 2 (DDR) = 400 MHz, or at least it really shouldn't. I can't say I've ever heard of someone safely and successfully overclocking their RAM...

HTH
WM

[MacBandit]

Originally posted by WM.
"System bus" really isn't the correct term here, since in the case of the G5 we're only talking about the bus between the U3 controller and RAM. I'd probably refer to it as the memory bus or the RAM bus instead.

Also, to be nit-picky, it doesn't run at 400 MHz DDR. Rather, it's like this: 200 MHz x 2 (DDR) x 2 (two banks of RAM) = 800 MHz.

Finally, when you say that it's hard to scale the "system bus" (meaning the RAM bus), I would add that it's basically impossible--because the RAM will never run faster than 200 MHz x 2 (DDR) = 400 MHz, or at least it really shouldn't. I can't say I've ever heard of someone safely and successfully overclocking their RAM...

HTH
WM

Agreed. Just another way of saying the same thing.

[WM.]

Originally posted by MacBandit
Agreed. Just another way of saying the same thing.
*shrug* Mmmkay.

And a note: although I like The Simpsons, I haven't seen very many episodes (probably around 5% of the 500), so I'm curious: have they ever really specified which Springfield they live in? I always figured it was the capital of Illinois, or another more-Middle-America city.

TIA
WM

[Phil Of Mac]

Unfortunately, they took down the site explaining this, but according to the Flat Earth Society, there is in fact only one Springfield. Every city named Springfield on Earth is simply a hyperspace link to the one Springfield, whose realspace location is unknown.

[MacBandit]

Originally posted by WM.
*shrug* Mmmkay.

And a note: although I like The Simpsons, I haven't seen very many episodes (probably around 5% of the 500), so I'm curious: have they ever really specified which Springfield they live in? I always figured it was the capital of Illinois, or another more-Middle-America city.

TIA
WM

There was an interview with Groening and in it he says that it's a combination of a lot of towns throughout the US but he lived in Springfield, OR for a short while and it's the town for which it is named.

[WM.]

Originally posted by MacBandit
There was an interview with Groening and in it he says that it's a combination of a lot of towns throughout the US but he lived in Springfield, OR for a short while and it's the town for which it is named.
Ah, cool. No better source than the creator, I guess. :)

Thanks
WM

P.S. OTOH, I like Phil's answer better. :D

[Snowy_River]

Originally posted by Phil Of Mac
Unfortunately, they took down the site explaining this, but according to the Flat Earth Society, there is in fact only one Springfield. Every city named Springfield on Earth is simply a hyperspace link to the one Springfield, whose realspace location is unknown.

Ah, if only you could enter one Springfield and leave another. A couple of hours driving to get from Portland, OR to Chicago, IL. It'd take less time than flying there...

[MacBandit]

Originally posted by Snowy_River
Ah, if only you could enter one Springfield and leave another. A couple of hours driving to get from Portland, OR to Chicago, IL. It'd take less time than flying there...

That would be cool. Actually you could probably go anywhere in the US in 2-3 hours and many places in the world I would guess.

[picnicbob]

Ok, seeing as the G5 doesnt have a level 3 cache, isnt it unfair to test the base model G5 against a base model G4? You have to take into consideration that the G5 uses RAM as it's level 3 cache unlike the G4 which already has a built in 2MB of cache to work with. Although 2MB really isnt that much, the G5 has 0. Shouldnt the tests be done with more ram in both computers (either top of the line both or something like 2GB in both) as to even out the difference in architecture? The tests run by Mac Addict at http://www.macaddict.com/news/news_007.html support this in the last test they did when the put 2GB of RAM into a G5 and ran photoshop tests. They should have put 2GB in the other models as well though. That would have been a true test.

[Phil Of Mac]

Since the G5 wins, L3 cache or not, I don't think it's that big of an issue.

You're testing an architecture with an L3 against one without an L3. That's an inherent unevenness. Just like testing an architecture with a G5 processor against one with a G4 processor.

[panphage]

Also, I think the consensus is that the L3 cache is added to G4s because their memory bus is too slow. The L3 is a kludge to improve performance since Moto can't improve the bus architecture. The G5's memory bus is megafast, so it doesn't need the L3 cache.

What is the speed of the bus between the G4 and its L3 cache anyway?

[MacBandit]

Originally posted by panphage
Also, I think the consensus is that the L3 cache is added to G4s because their memory bus is too slow. The L3 is a kludge to improve performance since Moto can't improve the bus architecture. The G5's memory bus is megafast, so it doesn't need the L3 cache.

What is the speed of the bus between the G4 and its L3 cache anyway?

The L3 cache on OEM G4 modules is accessed at 1/2 the speed of the CPU. The G5 accesses system ram at 1/2 the speed of the cpu. So overall the throughput potential for the two is the same though there has to be increased latency in accessing system ram rather than an on board L3 cache.