bus speed
- logancircle
- tinnitus
- Posts: 1107
- Joined: Fri May 09, 2003 8:45 am
- Location: Brooklyn, NY
bus speed
"Dude, you can upgrade your CPU all you want, but what really matters is bus speed!"
I am considering upgrading a G4 (sawtooth, bus speed 100MHz, original CPU was like a 4xxMHz). It was upgraded to 1Ghz single processor a few years ago, now I have a 1.8GHz to put in it, but will I notice any difference at this point? Some people insist that my bus speed is fast enough to see a difference in processing with the 1.8, but I want some other input. ? Thanks.
c in DC
PS. I have 1.5GB RAM
I am considering upgrading a G4 (sawtooth, bus speed 100MHz, original CPU was like a 4xxMHz). It was upgraded to 1Ghz single processor a few years ago, now I have a 1.8GHz to put in it, but will I notice any difference at this point? Some people insist that my bus speed is fast enough to see a difference in processing with the 1.8, but I want some other input. ? Thanks.
c in DC
PS. I have 1.5GB RAM
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
- @?,*???&?
- on a wing and a prayer
- Posts: 5804
- Joined: Wed May 07, 2003 4:36 pm
- Location: Just left on the FM dial
- Contact:
You will not notice anything with regard to buss speed, but your processor with kick butt. Faster processor = more plugins. Also, many functions will work more smoothly.
I have a sawtooth G4 AGP Mac running a 1.5 GHz processor and it's been great. I experimented with the jumper settings too as you can change the speed of the processor by doing so. The unrecommended change to 1.6GHz made the computer extremely unstable with that particular OWC processor. A 1.8 would likely have such options available too. There is a 2.0 GHz processor out for these computers now and I may be upgrading to that soon.
They would not keep making faster processors if we were dangerously near the threshold of buss speed for these units.
I have a sawtooth G4 AGP Mac running a 1.5 GHz processor and it's been great. I experimented with the jumper settings too as you can change the speed of the processor by doing so. The unrecommended change to 1.6GHz made the computer extremely unstable with that particular OWC processor. A 1.8 would likely have such options available too. There is a 2.0 GHz processor out for these computers now and I may be upgrading to that soon.
They would not keep making faster processors if we were dangerously near the threshold of buss speed for these units.
- logancircle
- tinnitus
- Posts: 1107
- Joined: Fri May 09, 2003 8:45 am
- Location: Brooklyn, NY
I thought about that, but then I thought they sell us all kinds of shite we don't need, all they have to do is not tell us we don't need it. Hoping you're right tho, I'll post the results. Thanks for the info, whatever-your-name-is.@?,*???&? wrote:They would not keep making faster processors if we were dangerously near the threshold of buss speed for these units.
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
- @?,*???&?
- on a wing and a prayer
- Posts: 5804
- Joined: Wed May 07, 2003 4:36 pm
- Location: Just left on the FM dial
- Contact:
btw:logancircle wrote:I thought about that, but then I thought they sell us all kinds of shite we don't need, all they have to do is not tell us we don't need it. Hoping you're right tho, I'll post the results. Thanks for the info, whatever-your-name-is.@?,*???&? wrote:They would not keep making faster processors if we were dangerously near the threshold of buss speed for these units.
http://eshop.macsales.com/item/Newer%20 ... XG47S2000/
-
- steve albini likes it
- Posts: 339
- Joined: Sat Nov 22, 2003 10:51 am
- Location: Iowa
- Contact:
- logancircle
- tinnitus
- Posts: 1107
- Joined: Fri May 09, 2003 8:45 am
- Location: Brooklyn, NY
Do it, you'll freak out at how powerful your computer will be. It sounds like we have the same system, both using 001 systems. Go here http://eshop.macsales.com/shop/accelerators/ and specify your computer model. If you do have the same sawtooth version as mine, get the Newertech 1.8GHz single processor. My 32-track sessions are enjoying some leg room.stereopathetic_banjo wrote:And you're still running a 001 system with the processor upgrade, aren't you? I have an old 466 single processor that i'm looking to do this to. Any problems getting things to work as normal? Thanks.
-Travis
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
Studio and Field Recorder in NYC.
I like dirt.
IG: stormydanielson
~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~
-
- steve albini likes it
- Posts: 339
- Joined: Sat Nov 22, 2003 10:51 am
- Location: Iowa
- Contact:
1 - buss speeds have been getting faster too.@?,*???&? wrote: They would not keep making faster processors if we were dangerously near the threshold of buss speed for these units.
2 - If buss speed or CPU speed is limiting is an application specific question. There exist applications that will hit the CPU wall on a modern machine and there are applications that will max out the buss on a modern machine.
A faster buss speed probably won't let you run more plugins. But it may increase your total track count.
Symbolic-name guy, don't mean to burst yr bubble, but computer companies, upgrade companies included, will definitely sell you stuff that you don't need or won't do you any good. In this case the upgrade probably does some good, but the OP has it right in that there is a bus speed limitation....
IIRC the potential of the early G4's was limited by memory bandwidth (directly proportional to bus speed) which throttled the performance of the G4 down. Good processors with decent software, but too many cycles spent waiting for data to come back from memory....
Here's a very simplified example of the throughput in a G4:
128-bits (16 bytes) per cycle IO to Altivec engine, 1GHz (for example)
for a peak of 16 x 1G = 16Gbyte/s
Lets say for the sake of argument you do memory IO 50% of the time. (50% of the time, the data is available in the onboard cache)
8 x 1G = 8Gbyte/s
Now think about the 64-bit memory bus (and we'll assume no additional limitation between the memory controller and Altivec)
8 bytes x 100MHz = 800MBytes/s
In this simplified example we're off by a factor of 10. That is, the CPU can process so much more data than the memory can deliver, you don't get the benefit of such a powerful processor at all. The later machines got better with DDR once it was implemented right, and people saw some amazing performance bumps.
I don't know how the OWC boards exactly work (do they have local RAM?), but there's no way to realize the full potential of processor upgrades in an application like ours (that hits memory so hard) if your bus bandwidth is that limited. Yes you will see performance bumps, but it's not linear...
IIRC the potential of the early G4's was limited by memory bandwidth (directly proportional to bus speed) which throttled the performance of the G4 down. Good processors with decent software, but too many cycles spent waiting for data to come back from memory....
Here's a very simplified example of the throughput in a G4:
128-bits (16 bytes) per cycle IO to Altivec engine, 1GHz (for example)
for a peak of 16 x 1G = 16Gbyte/s
Lets say for the sake of argument you do memory IO 50% of the time. (50% of the time, the data is available in the onboard cache)
8 x 1G = 8Gbyte/s
Now think about the 64-bit memory bus (and we'll assume no additional limitation between the memory controller and Altivec)
8 bytes x 100MHz = 800MBytes/s
In this simplified example we're off by a factor of 10. That is, the CPU can process so much more data than the memory can deliver, you don't get the benefit of such a powerful processor at all. The later machines got better with DDR once it was implemented right, and people saw some amazing performance bumps.
I don't know how the OWC boards exactly work (do they have local RAM?), but there's no way to realize the full potential of processor upgrades in an application like ours (that hits memory so hard) if your bus bandwidth is that limited. Yes you will see performance bumps, but it's not linear...
"Revolution is not a dinner party." -Sun Yat-Sen
- @?,*???&?
- on a wing and a prayer
- Posts: 5804
- Joined: Wed May 07, 2003 4:36 pm
- Location: Just left on the FM dial
- Contact:
Good thing about this board is that we can, at times, cross the threshold into really useful information- but next to none of it is shared by the hardware companies to the consumers buying their systems. Oobedob, what's your role in the industry? I am a hardware user and professional producer and engineer- do you work for a hardware company?oobedoob wrote:Symbolic-name guy, don't mean to burst yr bubble, but computer companies, upgrade companies included, will definitely sell you stuff that you don't need or won't do you any good. In this case the upgrade probably does some good, but the OP has it right in that there is a bus speed limitation....
IIRC the potential of the early G4's was limited by memory bandwidth (directly proportional to bus speed) which throttled the performance of the G4 down. Good processors with decent software, but too many cycles spent waiting for data to come back from memory....
Here's a very simplified example of the throughput in a G4:
128-bits (16 bytes) per cycle IO to Altivec engine, 1GHz (for example)
for a peak of 16 x 1G = 16Gbyte/s
Lets say for the sake of argument you do memory IO 50% of the time. (50% of the time, the data is available in the onboard cache)
8 x 1G = 8Gbyte/s
Now think about the 64-bit memory bus (and we'll assume no additional limitation between the memory controller and Altivec)
8 bytes x 100MHz = 800MBytes/s
In this simplified example we're off by a factor of 10. That is, the CPU can process so much more data than the memory can deliver, you don't get the benefit of such a powerful processor at all. The later machines got better with DDR once it was implemented right, and people saw some amazing performance bumps.
I don't know how the OWC boards exactly work (do they have local RAM?), but there's no way to realize the full potential of processor upgrades in an application like ours (that hits memory so hard) if your bus bandwidth is that limited. Yes you will see performance bumps, but it's not linear...
Would be great to know a way to actually test the limits of buss speed- the performance gained from upgrading a processor on an AGP mac is earth-shattering, so inspite of the previous post, rest assured, you will experience the same dramatic effect. Linear or not, it's real and very, very worth it.
well, the short answer for the end user is that there's usually no way to tell in advance. Benchmarks are often used....
Take your platform of choice and instantiate the max amount of your favorite reverb. Take the reverbs out and instantiate the max amount of tracks (unless you're in an LE type situation). Record, play back, automate until it crashes/stutters or whatever symptom you decide is unacceptable. This gets you a fair approximation of the limits of your disk and memory IO.
Then do the upgrade and see if the symptom/limitation goes away. If it does go away, is it worth what you paid? What's your time worth? If the upgrade doesn't pass sensible criteria, return it and get a refund.
But I don't think you'll be able to isolate the effects of bus speed unless you can over/underclock the bus, and I don't know if 2002-era mac firmware is hackable that way at all. On OEM PC motherboards this is generally easier, as you often see quite a bit of adjustment available in the BIOS
Take your platform of choice and instantiate the max amount of your favorite reverb. Take the reverbs out and instantiate the max amount of tracks (unless you're in an LE type situation). Record, play back, automate until it crashes/stutters or whatever symptom you decide is unacceptable. This gets you a fair approximation of the limits of your disk and memory IO.
Then do the upgrade and see if the symptom/limitation goes away. If it does go away, is it worth what you paid? What's your time worth? If the upgrade doesn't pass sensible criteria, return it and get a refund.
But I don't think you'll be able to isolate the effects of bus speed unless you can over/underclock the bus, and I don't know if 2002-era mac firmware is hackable that way at all. On OEM PC motherboards this is generally easier, as you often see quite a bit of adjustment available in the BIOS
"Revolution is not a dinner party." -Sun Yat-Sen
- @?,*???&?
- on a wing and a prayer
- Posts: 5804
- Joined: Wed May 07, 2003 4:36 pm
- Location: Just left on the FM dial
- Contact:
I just found this link on the OWC site:
http://eshop.macsales.com/Service/rebate-program/
I'm sitting on an old 400MHz processor and if I upgrade to a 2.0 GHz processor now, I can get $190 back on my 1.5GHz processor and $2.50 for the 400MHz processor as trade-ins! That means I can upgrade for half-price!
Sweet.
http://eshop.macsales.com/Service/rebate-program/
I'm sitting on an old 400MHz processor and if I upgrade to a 2.0 GHz processor now, I can get $190 back on my 1.5GHz processor and $2.50 for the 400MHz processor as trade-ins! That means I can upgrade for half-price!
Sweet.
Who is online
Users browsing this forum: No registered users and 59 guests