Just tried TM and no improvement over iStat. Still not detecting the SATA drive on the ATA100 bus. The other drives attached to the ATA66 bus are there. Must be the adapter. Not sure what's up with the RAID. I get slightly better Xbench results from my single SATA on the ATA100 bus than you did with your RAID setup, so RAID0 doesn't really bring that much to the table, sadly.
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.
My PowerMac G4 Lucky Dip (Twins!)
- Thread starter AphoticD
- Start date
- Sort by reaction score
Just tried TM and no improvement over iStat. Still not detecting the SATA drive on the ATA100 bus. The other drives attached to the ATA66 bus are there. Must be the adapter. Not sure what's up with the RAID. I get slightly better Xbench results from my single SATA on the ATA100 bus than you did with your RAID setup, so RAID0 doesn't really bring that much to the table, sadly.
I was quick to blame Tiger and the RAID0 setup... I just spent a couple of hours trying different configurations and reinstalls. It turns out that I had another dud (or at least intermittent) hard drive!
I put the third, unused 160GB, drive C in place of drive A, setup the RAID0 again in Tiger, restored the existing system across again from the original 20GB rattler and rebooted to a happy, zippy Mac.
Knowing that my first attempt resulted in that same error, I went and reformatted drive A and ran it through some tests. On the same bus, drive A benchmarked a total score of 34.98, whereas drive B (the same brand and model number with only a slight change in the revision + about 6 months newer) benchmarked a solo score of 57.25. With this bizarre variation, I decided to install it in a JBOD eSATA box connected to my Mac Pro. I run through a few things for a minute or so and my Mac Pro crashed and forced itself to reboot! (I've never seen that before).
I went back and repeated the same thing and the cMP crashed and rebooted again.. weird. I put the drive in a G5 and it seems to be running fine. I installed Panther on it and the G5 is zipping along booted from this strange drive. It tested 65.70 on it's own in xbench while on the G5's SATA bus.
So back to the G4, drives B + C as a RAID0 set results in a much improved score of 68.70! And it feels fast. I'll leave it like this (again) and see how it goes throughout the week. I have a feeling that swapping these drives around will result in a better (and more reliable) experience.
While I was at it, I ran the Sawtooth through Geekbench 2
I ran the test twice to be sure and yep, bang on 200 points.
Last edited:
The DA 533 CPU arrived today. I cleaned it up, pushed some bent pins back into shape and did some reading on which resistors to add/remove to increase the multiplier. I went the safest, easiest option and just popped off R13 with my fingernail to go from 4x to 5x multiplier. I installed the CPU card with a fresh coat of silver thermal grease and powered it on. After a short OF-stage delay, the G4 booted up into Tiger and shows a 500Mhz 7410 CPU.
Everything is working well. I ran through Geekbench again and received a score of 285 (up from 200), which is in almost direct correlation with the increase from 350 to 500Mhz (~42.8% increase). The machine feels slightly faster, but it's hard to gauge with something like Webkit. There is a noticeable difference, yes, but it's still uncomfortably slow to browse without disabling Javascript.
The RAID drives are kicking along okay still (no more crashes --knock on wood). Xbench showed a 1 point increase with the faster CPU. Now 69.19.
I'll leave it at 500Mhz for now, until I am certain it is stable and then attempt to go to 6x by removing the R9 resistor and bridging R11.
I am happy with my cheap CPU upgrade, which required no additional modifications and appears to stay cool even with the original chopped-in-half L-shaped Heatsink. I guess this is due to the lower power requirement of the 7410.
(I am typing this from the G4 now.)
Everything is working well. I ran through Geekbench again and received a score of 285 (up from 200), which is in almost direct correlation with the increase from 350 to 500Mhz (~42.8% increase). The machine feels slightly faster, but it's hard to gauge with something like Webkit. There is a noticeable difference, yes, but it's still uncomfortably slow to browse without disabling Javascript.
The RAID drives are kicking along okay still (no more crashes --knock on wood). Xbench showed a 1 point increase with the faster CPU. Now 69.19.
I'll leave it at 500Mhz for now, until I am certain it is stable and then attempt to go to 6x by removing the R9 resistor and bridging R11.
I am happy with my cheap CPU upgrade, which required no additional modifications and appears to stay cool even with the original chopped-in-half L-shaped Heatsink. I guess this is due to the lower power requirement of the 7410.
(I am typing this from the G4 now.)
In preparation for my clamshell iBook DC-In board repair, I thought I would do some soldering jobs I had put on the backburner.
First up was the overclocking of my Sawtooth G4 which has the DA 533Mhz CPU card installed (these G4s were originally 350Mhz). As many of you may know, the DA 533 CPU card installed in a Sawtooth with a 100Mhz system bus clocks itself down to 400Mhz due to the PLL multiplier which is at a default of 4x. By following the information found earlier in this thread, I was able to increase to a 5x multiplier (500Mhz) by removing one resistor on the CPU card (R13). I then removed another resistor (R9) and carefully drew in a very fine solder blob to bridge another (R11).
I now have a 6x Multiplier for a running speed of 600Mhz! Coupled with the flashed Radeon 9700 Pro (128MB) and an assortment of dirt cheap PC133 RAM for a total of 768MB installed, I decided to give LeopardAssist a go and install 10.5.8 on this unsupported G4.
Installation was quick and easy (by faking a 933Mhz CPU speed) and then a quick reboot and I was greeted with a very smooth welcome video. Everything is fluidly smooth, which I really wasn't expecting.
Leopard runs great on this Mac! I understand Apple like to restrict machines to sell new ones for new OS releases, but I believe this G4 is more than capable of operating Leopard. Time Machine, Exposé, Spaces, Dashboard, etc are all silky smooth (Widget ripple effect included). It even natively supports the transparent menu bar.
The only time I have noticed that the Sawtooth is in any way underpowered is when installing the software updates as it seems to take much longer than Tiger does.
Regardless, I am very impressed with the Sawtooth and Leopard in this setup.
I've found a stack of 256MB RAM for 99c, so I will bump it up to 1GB and call it enough. I also ended up scrapping the RAID setup as it was intermittently refusing to boot, but it's really not much different with a single adapted SATA HDD on the ATA/66 bus.
(Posted from the G4)
First up was the overclocking of my Sawtooth G4 which has the DA 533Mhz CPU card installed (these G4s were originally 350Mhz). As many of you may know, the DA 533 CPU card installed in a Sawtooth with a 100Mhz system bus clocks itself down to 400Mhz due to the PLL multiplier which is at a default of 4x. By following the information found earlier in this thread, I was able to increase to a 5x multiplier (500Mhz) by removing one resistor on the CPU card (R13). I then removed another resistor (R9) and carefully drew in a very fine solder blob to bridge another (R11).
I now have a 6x Multiplier for a running speed of 600Mhz! Coupled with the flashed Radeon 9700 Pro (128MB) and an assortment of dirt cheap PC133 RAM for a total of 768MB installed, I decided to give LeopardAssist a go and install 10.5.8 on this unsupported G4.
Installation was quick and easy (by faking a 933Mhz CPU speed) and then a quick reboot and I was greeted with a very smooth welcome video. Everything is fluidly smooth, which I really wasn't expecting.
Leopard runs great on this Mac! I understand Apple like to restrict machines to sell new ones for new OS releases, but I believe this G4 is more than capable of operating Leopard. Time Machine, Exposé, Spaces, Dashboard, etc are all silky smooth (Widget ripple effect included). It even natively supports the transparent menu bar.
The only time I have noticed that the Sawtooth is in any way underpowered is when installing the software updates as it seems to take much longer than Tiger does.
Regardless, I am very impressed with the Sawtooth and Leopard in this setup.
I've found a stack of 256MB RAM for 99c, so I will bump it up to 1GB and call it enough. I also ended up scrapping the RAID setup as it was intermittently refusing to boot, but it's really not much different with a single adapted SATA HDD on the ATA/66 bus.
(Posted from the G4)
Attachments
Last edited:
In preparation for my clamshell iBook DC-In board repair, I thought I would do some soldering jobs I had put on the backburner.
First up was the overclocking of my Sawtooth G4 which has the DA 533Mhz CPU card installed (these G4s were originally 350Mhz). As many of you may know, the DA 533 CPU card installed in a Sawtooth with a 100Mhz system bus clocks itself down to 400Mhz due to the PLL multiplier which is at a default of 4x. By following the information found earlier in this thread, I was able to increase to a 5x multiplier (500Mhz) by removing one resistor on the CPU card (R13). I then removed another resistor (R9) and carefully drew in a very fine solder blob to bridge another (R11).
I now have a 6x Multiplier for a running speed of 600Mhz! Coupled with the flashed Radeon 9700 Pro (128MB) and an assortment of dirt cheap PC133 RAM for a total of 768MB installed, I decided to give LeopardAssist a go and install 10.5.8 on this unsupported G4.
Installation was quick and easy (by faking a 933Mhz CPU speed) and then a quick reboot and I was greeted with a very smooth welcome video. Everything is fluidly smooth, which I really wasn't expecting.
Leopard runs great on this Mac! I understand Apple like to restrict machines to sell new ones for new OS releases, but I believe this G4 is more than capable of operating Leopard. Time Machine, Exposé, Spaces, Dashboard, etc are all silky smooth (Widget ripple effect included). It even natively supports the transparent menu bar.
The only time I have noticed that the Sawtooth is in any way underpowered is when installing the software updates as it seems to take much longer than Tiger does.
Regardless, I am very impressed with the Sawtooth and Leopard in this setup.
I've found a stack of 256MB RAM for 99c, so I will bump it up to 1GB and call it enough. I also ended up scrapping the RAID setup as it was intermittently refusing to boot, but it's really not much different with a single adapted SATA HDD on the ATA/66 bus.
(Posted from the G4)
very Nice
im glad to see the 7410 clocked to 600Mhz, do you plan on pushing your luck and trying for 650Mhz?
I have always found it quite hard to make a solder jumper/blob between the 2 pads when doing overclocking like that, as the solder resist between the pads does its job very well LOL so your doing better then me
, in the end on the pismo for example I used some small copper strands I cut out of a bad ATA cable, and used those as very small jumper wires between the 2 pads I wished to jumper.the Radeon 9700 is probably whats helping out quite a lot with how well leopard runs, stick the stock Rage 128 back in and you might see how much of a slide show leopard can be at times on such Hardware
(this is probably one of the reasons apple cut off machines slower then 867Mhz with leopard as keep in mind slower clocked machines not only had slower CPUs but slower GPUs as well)its worth noting that id not be surprised if your 600Mhz setup bench marked faster then say a 700Mhz iMac G4/eMac since the 7410 clock for clock is faster then a 7450, due to the 7410 having a shorter pipeline among other things.
it would be also very interesting to see how well it compares (like in geekbench 2.2.7) to my 600Mhz 7410 Pismo that I made
as your 7410 is running in MPX bus mode, where as my Pismo due to originally starting out with a 750 is running in 60x bus mode im pretty sure. but every other factor apart from the CPU bus modes is the same, the Pismo uses the same Uninorth 1 and keylargo chipset as the sawtooth at the same clock speed. (the MPX bus of the G4 is about 20% faster IIRC, then the 60x bus) the 74xx CPUs supports both MPX and 60x buses but the 750XX CPUs only supports the 60x bus, so the benchmarks should be very interesting indeed.(PS I like the little system profiler thing bellow about this mac in that screen shot
)
very Nice
im glad to see the 7410 clocked to 600Mhz, do you plan on pushing your luck and trying for 650Mhz?
Thanks man. You inspired me once again
I think I will try going to 650. I'll run it like this for a little and then bridge R7 for the 6.5 multiplier. I am blown away by the low temperature readings.
Here's Gauge Pro in Mac OS 9.2.2:
I was able to push it to reach 26°C after running through 10 x Memory Tests in GP.
I then rebooted into Mac OS 8.6 and checked again and it was sitting at 19°C. I don't know if I can rely on these readings. But without a doubt there is only very slightly warm air coming from the PSU exhaust and the stock (L-shaped) CPU heatsink is only warm to the touch (not hot as you'd expect).
In comparison, the Sawtooth with the stock 350Mhz CPU card shows a temperature reading of 34°C in GP in the same operating environment.
I have always found it quite hard to make a solder jumper/blob between the 2 pads when doing overclocking like that, as the solder resist between the pads does its job very well LOL so your doing better then me
That's a good idea. It took a bit of patience, but found I could heat/melt the solder and then drag away from one side to form a little bridge. It worked well. I used the same technique to restore my busted Radeon 9800 Mac Edition (256MB) and got it back up and running. I found a tiny resistor had popped and there was evidence of a short /arcing on the AGP connectors toward the ports/plate end. I bet if I were to pull out that G5's logic board and inspect inside the AGP port I would find evidence of this short and that finally explains how the logic board's AGP bus failed. (It would only boot WITHOUT a graphics card). I will possibly attempt a repair one day. At the very least, it won't require a trip into the oven as I had contemplated
It's not all good news for that 9800 though. Although it works, there is faint yellow vertical banding which appears on all GL surfaces (finder windows, games and anything Core Image related). So I suspect the VRAM could have overheated and resulted in irreparable damage. Either that or it's a capacitor related thing, but I couldn't spot any swelling on any.
the Radeon 9700 is probably whats helping out quite a lot with how well leopard runs, stick the stock Rage 128 back in and you might see how much of a slide show leopard can be at times on such Hardware
I agree with this. It's the GPU handling all the visual load. It runs so well considering the age of the Mac though. There's hardly a stutter and 9/10 animations run as smooth as my G5s. It certainly has an edge in the smooth visual department over my first-gen Dual 1.8 G5 which has the stock Nvidia FX 5200 GPU.
its worth noting that id not be surprised if your 600Mhz setup bench marked faster then say a 700Mhz iMac G4/eMac since the 7410 clock for clock is faster then a 7450, due to the 7410 having a shorter pipeline among other things.
it would be also very interesting to see how well it compares (like in geekbench 2.2.7) to my 600Mhz 7410 Pismo that I made
Geekbench is up from 285 (500Mhz), but that was in Tiger and not Leopard. I've always found the Leopard version to be slightly more generous for scores.
(PS I like the little system profiler thing bellow about this mac in that screen shot
Thanks. That's a little bit of GeekTool magic; Add a shell 'tool' with: system_profiler SPHardwareDataType | head -n 14 | tail -n 12 (or 'head -n 13' to trim off the serial number)
Here's a new Xbench score with the single SATA HDD:
As you can see, it's only slightly under the Striped RAID scores and the top throughput is actually slightly higher (but this could be due to the CPU speed bump). Either way, I think I wasted my time getting a striped RAID boot setup on the Sawtooth's stock ATA bus.
Last edited:
Thanks man. You inspired me once again
I think I will try going to 650. I'll run it like this for a little and then bridge R7 for the 6.5 multiplier. I am blown away by the low temperature readings.
Here's Gauge Pro in Mac OS 9.2.2:
View attachment 741943
I was able to push it to reach 26°C after running through 10 x Memory Tests in GP.
I then rebooted into Mac OS 8.6 and checked again and it was sitting at 19°C. I don't know if I can rely on these readings. But without a doubt there is only very slightly warm air coming from the PSU exhaust and the stock (L-shaped) CPU heatsink is only warm to the touch (not hot as you'd expect).
In comparison, the Sawtooth with the stock 350Mhz CPU card shows a temperature reading of 34°C in GP in the same operating environment.
That's a good idea. It took a bit of patience, but found I could heat/melt the solder and then drag away from one side to form a little bridge. It worked well. I used the same technique to restore my busted Radeon 9800 Mac Edition (256MB) and got it back up and running. I found a tiny resistor had popped and there was evidence of a short /arcing on the AGP connectors toward the ports/plate end. I bet if I were to pull out that G5's logic board and inspect inside the AGP port I would find evidence of this short and that finally explains how the logic board's AGP bus failed. (It would only boot WITHOUT a graphics card). I will possibly attempt a repair one day. At the very least, it won't require a trip into the oven as I had contemplated
It's not all good news for that 9800 though. Although it works, there is faint yellow vertical banding which appears on all GL surfaces (finder windows, games and anything Core Image related). So I suspect the VRAM could have overheated and resulted in irreparable damage. Either that or it's a capacitor related thing, but I couldn't spot any swelling on any.
I agree with this. It's the GPU handling all the visual load. It runs so well considering the age of the Mac though. There's hardly a stutter and 9/10 animations run as smooth as my G5s. It certainly has an edge in the smooth visual department over my first-gen Dual 1.8 G5 which has the stock Nvidia FX 5200 GPU.
View attachment 741944
Geekbench is up from 285 (500Mhz), but that was in Tiger and not Leopard. I've always found the Leopard version to be slightly more generous for scores.
Thanks. That's a little bit of GeekTool magic; Add a shell 'tool' with: system_profiler SPHardwareDataType | head -n 14 | tail -n 12 (or 'head -n 13' to trim off the serial number)
Here's a new Xbench score with the single SATA HDD:
View attachment 741945
As you can see, it's only slightly under the Striped RAID scores and the top throughput is actually slightly higher (but this could be due to the CPU speed bump). Either way, I think I wasted my time getting a striped RAID boot setup on the Sawtooth's stock ATA bus.
Very Cool (literally!)
the 7410 is a cooler running and more efficient chip then the 7400 (enabling Apple to build the G4 PowerBook) but I doubt its running that cold, the Thermal sensors in the 750-7410s are most of the time not calibrated properly, so they report erroneous figures (both of the 7410s in my G4 iMac G3 and my G4 Pismo report 2c! LOL)
I once fixed my G5s Radeon 9600 XT using a graphite pencil to jumper a resistor too
I had knocked off a resistor at one point and so i used the graphite as a sort of jumper as it its conductive. I also used the same technique to overclock my Mac Mini G4 to 1.67Ghz (edit: speaking of G5s this is my 970th post )as I expected your Sawtooth geekbenched just a bit higher then my Pismo: https://browser.geekbench.com/geekbench2/2631615
it would be interesting to see what the memory scores of your geekbench test are, those specify is where you sawtooth should be faster thanks due to the MPX bus and the fact your L2 cache is running at 300Mhz vs 240Mhz in my Pismo.
Last edited:
Well, I tried bumping it up to 650Mhz and it just resulted in either going straight into Open Firmware with a "DEFAULT CATCH" or instant Kernel Panics on boot (CPU invalid pmap). Removed resistor bridge and tried again, same result. Reverted back to 600Mhz OK. It looks like this is the ceiling for the 533Mhz D.A. CPU.
would be quite interesting if bumping the Core voltage would get it stable/booting at 650Mhz?
7410s have a recommended Core voltage of 1.8, with 2.1V being stated as the Max. in my Pismo due to it running a PPC750L before, the core voltage is set to 2V where my 7410 runs just fine but im wondering if its why I have been able to overclock my 7410 so far, (getting a 100Mhz overclock on a 7410 is normally very rare)
A few notes and updates on the Twin G4s...
SawtoothAlpha: Running well at the 600Mhz overclock with the DA 533 7410 (1MB L2 backside cache). Bumped RAM up to 1.5GB (2x 512MB + 2x 256MB). Runs Leopard incredibly well with the Radeon 9700 Pro.
In a gamble I snapped up a cheap offer for 2x 512MB PC133 *ECC* SDRAM DIMMs after researching about whether ECC modules would work. There was no definitive answer of yes or no, so I took a chance and it turns out they don’t work in these Sawtoothens (even standalone, without other RAM). So I shelved the modules and bought 2x non-ECC for a little bit more, but they are working perfectly.
SawtoothBeta: Solder bridged the PLL resistors to bump up the stock 7400 350Mhz CPU to 400Mhz. Increased RAM to 640MB (1x 256MB, 3x 128MB) all PC133 speeds. Runs well at 400Mhz. I will try bumping up to 450Mhz next.
Now, I spotted a few cheap upgrade options and wanted to run these by the PowerPC knowledgebase (a.k.a @LightBulbFun and crew).
1. I found a cheap Gigabit Ethernet logic board complete with CPU;
https://www.ebay.com.au/i/142674641896
This looks like it might be the single 400Mhz CPU as the heatsink appears to be the same size as my sawtooth’s. Does the dual CPU heatsink on a GigE look different? Is there a CPU cooling fan like the DA/QS?
Also, as the GigE has support for ADC Power, does it require a power supply with additional output or can I just wire in from the 12v lines for the additional two logic board power pins? I understand the power plug has 22pins on the GigE over 20pins on the Sawtooth.
If I could put the GigE logic board in then it should open up availability for dual CPUs, and I could possibly install my 9800 Pro Mac Edition from my G5 and run a Cinema Display from it.
2. Back to the original Sawtooth logic board. I have spotted an 800Mhz Quicksilver CPU Card for very cheap. Firstly it would drop to 600Mhz (6x multiplier), but it can be bumped back up to 8x (max) with a single solder bridge (R13 iirc). It would require additional 12v on the 4th post hole and some slimming of the IDE ODD connector, but this should work fine with the stock cooling, right?
My question is; would a 7455 @ 800mhz with 256k L2 (1:1) show much of an improvement over the 7410 @ 600Mhz with 1MB L2 (1:2)? The cache is a quarter of the size, but would run at more than twice the current speed (800mhz over 300Mhz). Would there be much of a difference or should I just hold out for a DA 733Mhz Card with 1M L2 (1:2) + 1M L3 and bump that up to 8x PLL multiplier? Would that be the best Apple CPU for the Sawtooth? I tested my boards and their UniNorth’s don’t support dual proc.
There was also the QS 2002 single 933Mhz 7455 with 256K L2 (1:1) + 2M L3. This might be the best single CPU for these machines. Has anyone tried these?
I haven’t seen any Sonnet or maxpower upgrade cards on eBay, but the various scrapped stock G4 CPUs are (occasionally) cheap and readily available, so I’m trying to make the most of these Macs without spending much.
SawtoothAlpha: Running well at the 600Mhz overclock with the DA 533 7410 (1MB L2 backside cache). Bumped RAM up to 1.5GB (2x 512MB + 2x 256MB). Runs Leopard incredibly well with the Radeon 9700 Pro.
In a gamble I snapped up a cheap offer for 2x 512MB PC133 *ECC* SDRAM DIMMs after researching about whether ECC modules would work. There was no definitive answer of yes or no, so I took a chance and it turns out they don’t work in these Sawtoothens (even standalone, without other RAM). So I shelved the modules and bought 2x non-ECC for a little bit more, but they are working perfectly.
SawtoothBeta: Solder bridged the PLL resistors to bump up the stock 7400 350Mhz CPU to 400Mhz. Increased RAM to 640MB (1x 256MB, 3x 128MB) all PC133 speeds. Runs well at 400Mhz. I will try bumping up to 450Mhz next.
Now, I spotted a few cheap upgrade options and wanted to run these by the PowerPC knowledgebase (a.k.a @LightBulbFun and crew).
1. I found a cheap Gigabit Ethernet logic board complete with CPU;
https://www.ebay.com.au/i/142674641896
This looks like it might be the single 400Mhz CPU as the heatsink appears to be the same size as my sawtooth’s. Does the dual CPU heatsink on a GigE look different? Is there a CPU cooling fan like the DA/QS?
Also, as the GigE has support for ADC Power, does it require a power supply with additional output or can I just wire in from the 12v lines for the additional two logic board power pins? I understand the power plug has 22pins on the GigE over 20pins on the Sawtooth.
If I could put the GigE logic board in then it should open up availability for dual CPUs, and I could possibly install my 9800 Pro Mac Edition from my G5 and run a Cinema Display from it.
2. Back to the original Sawtooth logic board. I have spotted an 800Mhz Quicksilver CPU Card for very cheap. Firstly it would drop to 600Mhz (6x multiplier), but it can be bumped back up to 8x (max) with a single solder bridge (R13 iirc). It would require additional 12v on the 4th post hole and some slimming of the IDE ODD connector, but this should work fine with the stock cooling, right?
My question is; would a 7455 @ 800mhz with 256k L2 (1:1) show much of an improvement over the 7410 @ 600Mhz with 1MB L2 (1:2)? The cache is a quarter of the size, but would run at more than twice the current speed (800mhz over 300Mhz). Would there be much of a difference or should I just hold out for a DA 733Mhz Card with 1M L2 (1:2) + 1M L3 and bump that up to 8x PLL multiplier? Would that be the best Apple CPU for the Sawtooth? I tested my boards and their UniNorth’s don’t support dual proc.
There was also the QS 2002 single 933Mhz 7455 with 256K L2 (1:1) + 2M L3. This might be the best single CPU for these machines. Has anyone tried these?
I haven’t seen any Sonnet or maxpower upgrade cards on eBay, but the various scrapped stock G4 CPUs are (occasionally) cheap and readily available, so I’m trying to make the most of these Macs without spending much.
EDIT: After much deliberation, I took a stab and made a half price offer on a DA 733Mhz CPU Card out of the UK (total £28 inc shipping to Australia), which the seller accepted. So within a few weeks, SawtoothAlpha will be rocking an OC’d 800mhz 7450 with a 1M L2 + 1M L3 caches.
Just looking over the specs on Mactracker, I think Apple could have felt obligated to add the L3 to this CPU because as @LightBulbFun states, the 7410 was faster per MHz than the 7450. So maybe the L3 was added on to distinguish the 733 (7450) as a significant upgrade to the 7410 based 533Mhz model.
I’ll move the 600Mhz 7410 into SawtoothBeta and call them both significantly “enhanced” enough for my liking.
Would it be advised to add a cooling fan for the 733 CPU? The CPU is shipping sans-heatsink, so i’ll reuse the sawtooth’s original, which has been doing a great job keeping the 600Mhz 7410 cool. Any advice or recommendations?
Just looking over the specs on Mactracker, I think Apple could have felt obligated to add the L3 to this CPU because as @LightBulbFun states, the 7410 was faster per MHz than the 7450. So maybe the L3 was added on to distinguish the 733 (7450) as a significant upgrade to the 7410 based 533Mhz model.
I’ll move the 600Mhz 7410 into SawtoothBeta and call them both significantly “enhanced” enough for my liking.
Would it be advised to add a cooling fan for the 733 CPU? The CPU is shipping sans-heatsink, so i’ll reuse the sawtooth’s original, which has been doing a great job keeping the 600Mhz 7410 cool. Any advice or recommendations?
who said ECC RAM does not work in Sawtooths? (sawteeth?) regular ECC SDRAM sticks will work, I dont THINK the Uninorth memory controller actually actually uses the ECC functions however (this is like installing ECC RAM in a consumer Motherboard if it works its just going to ignore the ECC part). but if your going to use ECC RAM it has to be ECC un-registered, registered ECC RAM wont work. (Registered RAM is easy to spot, they have extra non RAM ICs on them. ECC un-registered sticks just have the 9th RAM and no extra ICs apart from the SPD data IC that all PCxxx RAM has of course)
the GigE G4 does use a diffrent PSU setup compared to the sawtooth (the G3 BW Yikes! and sawtooth all share the same PSU) this site http://atxg4.com/ has some useful info on the various diffrent PSUs including nice pinouts. (I dont think the guy is actually selling ATX adapter any more sadly)
im guessing you mean 256KB of L2 + 1MB of L3 the only G4 with 1MB of on die L2 was the legendary 7448
I think some of the 800Mhz ish sonnet CPUs which used the same heatsink design the sawtooth had, did get away without using a fan but if your planning to clock things up It wouldn't go amiss to stick one on the end. but if you cant I think you can go fan-less at 800Mhz but make sure to monitor temps
in regards to the Best CPU for a Single CPU only Sawtooth
well that would be one of the very rare Newertech Single 2Ghz 7448 G4 upgrades I only know of one person here who owns one. (and ironically its in a Digital audio ) but in terms of Apple CPUs yeah the 933Mhz CPU from a QS would be the fastest you could shoe horn into a Sawtooth.
regular ECC SDRAM sticks will work, I dont THINK the Uninorth memory controller actually actually uses the ECC functions however (this is like installing ECC RAM in a consumer Motherboard if it works its just going to ignore the ECC part). but if your going to use ECC RAM it has to be ECC un-registered, registered ECC RAM wont work. (Registered RAM is easy to spot, they have extra non RAM ICs on them. ECC un-registered sticks just have the 9th RAM and no extra ICs apart from the SPD data IC that all PCxxx RAM has of course)the GigE G4 does use a diffrent PSU setup compared to the sawtooth (the G3 BW Yikes! and sawtooth all share the same PSU) this site http://atxg4.com/ has some useful info on the various diffrent PSUs including nice pinouts. (I dont think the guy is actually selling ATX adapter any more sadly)
im guessing you mean 256KB of L2 + 1MB of L3
the only G4 with 1MB of on die L2 was the legendary 7448 I think some of the 800Mhz ish sonnet CPUs which used the same heatsink design the sawtooth had, did get away without using a fan but if your planning to clock things up It wouldn't go amiss to stick one on the end.
but if you cant I think you can go fan-less at 800Mhz but make sure to monitor tempsin regards to the Best CPU for a Single CPU only Sawtooth
well that would be one of the very rare Newertech Single 2Ghz 7448 G4 upgrades
I only know of one person here who owns one. (and ironically its in a Digital audio ) but in terms of Apple CPUs yeah the 933Mhz CPU from a QS would be the fastest you could shoe horn into a Sawtooth.
Last edited:
who said ECC RAM does not work in Sawtooths? (sawteeth?)
Well that clears that up. The sticks I bought must be of the registered variety. They had plenty of added ICs and didn't work
Another great resource. Thanks. I've decided against the GigE board after investing in the DA 733 CPU.
im guessing you mean 256KB of L2 + 1MB of L3
Cool. I read the specs as the 733 having both a 1MB backside(1:2) L2 plus the 1MB L3 cache. The current 600Mhz 7410 CPU has 1MB backside (1:2) L2 cache only, so the additional 200Mhz + 1MB L3 should make a decent economic speed bump, despite not being quite as fast clock for clock as you mentioned earlier.
Is this right?
I think some of the 800Mhz ish sonnet CPUs which used the same heatsink design the sawtooth had, did get away without using a fan but if your planning to clock things up It wouldn't go amiss to stick one on the end.
in regards to the Best CPU for a Single CPU only Sawtooth
well that would be one of the very rare Newertech Single 2Ghz 7448 G4 upgrades
LOL, "shoe horn" indeed. Both the aftermarket and the 933 QS2002 CPUs have been very rare to spot, so I figure the 733 DA is probably the best bet and it appears it doesn't require added power or mods to the ODD IDE and quite possibly, it will run around the same temps as the 600Mhz 7410. Maybe not at a reported 22°C, but we'll find out
Well that clears that up. The sticks I bought must be of the registered variety. They had plenty of added ICs and didn't work
Another great resource. Thanks. I've decided against the GigE board after investing in the DA 733 CPU.
Cool. I read the specs as the 733 having both a 1MB backside(1:2) L2 plus the 1MB L3 cache. The current 600Mhz 7410 CPU has 1MB backside (1:2) L2 cache only, so the additional 200Mhz + 1MB L3 should make a decent economic speed bump, despite not being quite as fast clock for clock as you mentioned earlier.
Is this right?
LOL, "shoe horn" indeed. Both the aftermarket and the 933 QS2002 CPUs have been very rare to spot, so I figure the 733 DA is probably the best bet and it appears it doesn't require added power or mods to the ODD IDE and quite possibly, it will run around the same temps as the 600Mhz 7410. Maybe not at a reported 22°C, but we'll find out
Yeah Mac tracker aint the clearest there, but noticed the comma separating the two, the 1MB of L2 (1:2) is referring to the 466-533Mhz 7400-7410 CPU options, with the first part mentioning the L2 cache 256KB of on die full speed L2 used on the 667Mhz and 733Mhz BTO options. (those 2 options using the 7450 plus 1MB of L3 mentioned in the second half)
im pretty sure with the 733Mhz DA CPU your still going to have to slim down the ODD ATA connector, as the CPU card is mirrored if that make sense compared to say a Dual GigE Card (Ie the long part of a DA card juts out to the left into the ODD connector, vs say a Dual GigE card which juts out to the right) you can see this best by comparing a Dual CPU GigE card to a Dual CPU DA Card.
Mactracker did give me the wrong impression it seems. Either way, 256k on chip + 1M L3 should lift it's game a little.
I'd expect to hit maybe 450 on the GB scores after clocking up to 800Mhz (The 600Mhz 7410 hits 385). Anyway, I have the two machines, so the 400Mhz OC'ed CPU will come out and then the 600 7410 and the 800 7450 will be running side by side, so that should give me a chance to answer if the DA733 card is really much of an improvement over the DA533 in the Sawtooth.
I'll get the whittling knife ready
Mactracker did give me the wrong impression it seems. Either way, 256k on chip + 1M L3 should lift it's game a little.
I'd expect to hit maybe 450 on the GB scores after clocking up to 800Mhz (The 600Mhz 7410 hits 385). Anyway, I have the two machines, so the 400Mhz OC'ed CPU will come out and then the 600 7410 and the 800 7450 will be running side by side, so that should give me a chance to answer if the DA733 card is really much of an improvement over the DA533 in the Sawtooth.
I'll get the whittling knife ready
Yeah the 800Mhz CPU upgrade should give you a nice little bump
but it willl defo be an interesting comparison another interesting comparison would be to set the 733Mhz card to 600Mhz and do a direct clock for clock comparison of the 2 CPU cards (you could also use CHUD tools to disable the various L2 and L3 caches too)
The DA 733Mhz Card is in.. now, has anyone successfully modified the ODD IDE to accommodate the larger CPUs without simply removing the connection from the logic board?
7th week 2001
iv personally not had any experience with fitting oversized CPU cards to sawtooths (sawteeth?) but i was wondering much of the ODD connector gets in the way? can you remove just the top part of the connector shell and apply some tape to the CPU card to insulate it from the IDE connector pins?...
7th week 2001
iv personally not had any experience with fitting oversized CPU cards to sawtooths (sawteeth?) but i was wondering much of the ODD connector gets in the way? can you remove just the top part of the connector shell and apply some tape to the CPU card to insulate it from the IDE connector pins?...
It will still be about 3mm too proud. I am thinking of removing the beige plastic IDE shroud completely, then bending all 40 pins (39?) down and out a little and pushing the the IDE cable directly onto the pins. I will need to slice off the tab from the blue IDE plug and maybe even whittle down the the plug a little more to make it slimmer.
The other tidier option would be to remove the IDE socket from the board and resolder it on upside down from the underside, cutting out the tab slot and moving the center pin position.
Sounds like a lot of work for a DVD-ROM drive which hardly gets used though! I'll keep thinking it through.
yeah that could work. just depends even if pushing them down can you slide the ODD cable on far enough before fowling the CPU card?
I never thought about flipping it over like that, could work
but yeah thats a lot of Pins to desolder if you dont have a proper desoldering gun on hand...
So I decided to tackle the installation of the Digital Audio G4 733Mhz CPU card (PowerPC 7450, 256k L2, 1MB L3) in Sawtooth Alpha.
1. First of all, the DA 733 card is too long to fit in the tidy way of the original 7400@350Mhz and the DA's 7410@533Mhz cards. So I start by modifying the optical drive IDE connection to accommodate this. I begin by cutting away at the beige connector with a blade and here I have pried up the top row of pins from the connection.
2. I carefully bent each pin back and begun on the second row. This procedure took a _long_ time. I imagine someone might be able to achieve quicker results with a hot air gun or something to just melt it off.
3. More careful whittling away, I am half way through the bottom row of pins here. For support, I used a sticky note pad, as I was able to adjust the height of this to support the logic board underneath as I applied pressure while cutting from above.
4. Finally after 40 minutes of carefully whittling away while trying to not damage anything, all 39 pins are free. I've bent them back and tried to straighten them out to prepare for bending them forward next.
5. Using the "adjustable thickness" sticky notepad again, I carefully bent the first row down and out.
6. Repeated again for the top row. I've wedged a folded strip of electrical tape in to act as insulation between the rows.
7. In the meantime, I've begun to cut down the IDE connector to make it thinner.
8. So this is looking pretty tidy so far. Unfortunately, even after all that, it is still too thick on top to accommodate the larger CPU card...
9. So I set about whittling the connector down even more. I've shaved a few mm off the top and also from the face to allow the pins of varying lengths to make a more solid contact.
10. I've bent the pins down by a few degrees and insulated them from above with a strip of electrical tape.
This is now perfect to accommodate the CPU card. It sits tight, but completely flat and without putting any pressure or bend in either the CPU card or the logic board.
However..... (To be continued in just a moment!)
1. First of all, the DA 733 card is too long to fit in the tidy way of the original 7400@350Mhz and the DA's 7410@533Mhz cards. So I start by modifying the optical drive IDE connection to accommodate this. I begin by cutting away at the beige connector with a blade and here I have pried up the top row of pins from the connection.
2. I carefully bent each pin back and begun on the second row. This procedure took a _long_ time. I imagine someone might be able to achieve quicker results with a hot air gun or something to just melt it off.
3. More careful whittling away, I am half way through the bottom row of pins here. For support, I used a sticky note pad, as I was able to adjust the height of this to support the logic board underneath as I applied pressure while cutting from above.
4. Finally after 40 minutes of carefully whittling away while trying to not damage anything, all 39 pins are free. I've bent them back and tried to straighten them out to prepare for bending them forward next.
5. Using the "adjustable thickness" sticky notepad again, I carefully bent the first row down and out.
6. Repeated again for the top row. I've wedged a folded strip of electrical tape in to act as insulation between the rows.
7. In the meantime, I've begun to cut down the IDE connector to make it thinner.
8. So this is looking pretty tidy so far. Unfortunately, even after all that, it is still too thick on top to accommodate the larger CPU card...
9. So I set about whittling the connector down even more. I've shaved a few mm off the top and also from the face to allow the pins of varying lengths to make a more solid contact.
10. I've bent the pins down by a few degrees and insulated them from above with a strip of electrical tape.
This is now perfect to accommodate the CPU card. It sits tight, but completely flat and without putting any pressure or bend in either the CPU card or the logic board.
However..... (To be continued in just a moment!)
Last edited:
And the moment of truth...
11. Well whaddya know? The original heatsink is too wide! You would think this was something I would check before putting hours into modifying things... So the only obstruction is this toroidal inductor... The options are to either modify the CPU card or the heatsink. Let's take the donut off and see what we can do with it.
12. After a _long_ time desoldering this heavy copper wire, I finally have it free and ready to rewind to suit a new orientation.
13. Another hour passes me by and I have rewound the coil and soldered it back together again.
14. In order to accommodate the heatsink, I've had to create an almost 90° gap in the coil where the heatsink will reside. (There was a change of cameras here as the ankle biters arrived home from school to commandeer my iPhone).
15. Well that looks tidy enough...
16. All back together again... Ready for it's first run.
By this stage, I also modified the PLL multipliers to jump from 5.5x to 8x (133Mhz x 5.5 in the DA equals 733Mhz, which now on a 100Mhz bus at 8x equals 800Mhz). An interesting side note is that the 733Mhz 7450 DA card used the same PLL configuration as the 733Mhz 7450 in the Quicksilver - see http://power-mac-g4.com/g4quicksilverclockup.html ). I was a little suspicious at this point that it may actually be a QS card, except for the fact that it didn't require extra power.
17. It boots! And it feels fast. The L3 cache certainly makes a big difference here.
Ok, so I was happy with how it all went. The CPU is stable at 800Mhz and the ODD appears to be still working fine.
I started to pack everything up when I noticed just how damn HOT the heatsink was getting. It felt hot enough to fry an egg! Unfortunately I have no software option to monitor temps on this CPU, but it runs WAY hotter than the 7410 @ 600Mhz..
I ran the 800Mhz 7450 through the paces in Geekbench and the result was okay, not as much of a jump as I had hoped for (I was hoping it would make the 500 point mark).
The Processor prefs "Allow Nap" option (CHUD Tools) refused to work, so I figure it is running at full voltage constantly, which would explain the heat.
Concerned about the heat and looking for a comparison, I installed the 600Mhz DA 7410 into Sawtooth Beta (replacing the 400Mhz 7400), fired that one up and let it run through Geekbench. After a solid run through, I opened the side door and put my hand on the heatsink.. The 7410 produced NO heat, not even a little bit, it was still running cool to the touch.
With this gauge, I pulled out the 800Mhz 7450 and put the soldering iron to work on the PLL multipliers again. First, I dropped down to 750Mhz and ran through again. The heat was still blistering.. So I repeated the steps and dropped down to 700Mhz. Although, MUCH hotter than the 7410 at 600Mhz, I can live with the temps of the 7450 @ 700Mhz. It seems ok for now.
Running through Geekbench to compare the 600Mhz 7410 (1M L2) and the 700Mhz 7450 (256k L2 + 1M L3), they are pretty much on par, with the 7410 coming through with faster memory results and the 7450 producing a lot more heat.
I've ordered in some 12v cooling fans in various shapes and sizes (40mm and 60mm). I'll have a think about how to cool this machine with the fans, but if I can't I'll have to drop down to 650 or 600Mhz just to keep it running cool like it's older sibling processor.
(... more Geekbench results to follow)
11. Well whaddya know? The original heatsink is too wide! You would think this was something I would check before putting hours into modifying things... So the only obstruction is this toroidal inductor... The options are to either modify the CPU card or the heatsink. Let's take the donut off and see what we can do with it.
12. After a _long_ time desoldering this heavy copper wire, I finally have it free and ready to rewind to suit a new orientation.
13. Another hour passes me by and I have rewound the coil and soldered it back together again.
14. In order to accommodate the heatsink, I've had to create an almost 90° gap in the coil where the heatsink will reside. (There was a change of cameras here as the ankle biters arrived home from school to commandeer my iPhone).
15. Well that looks tidy enough...
16. All back together again... Ready for it's first run.
By this stage, I also modified the PLL multipliers to jump from 5.5x to 8x (133Mhz x 5.5 in the DA equals 733Mhz, which now on a 100Mhz bus at 8x equals 800Mhz). An interesting side note is that the 733Mhz 7450 DA card used the same PLL configuration as the 733Mhz 7450 in the Quicksilver - see http://power-mac-g4.com/g4quicksilverclockup.html ). I was a little suspicious at this point that it may actually be a QS card, except for the fact that it didn't require extra power.
17. It boots! And it feels fast. The L3 cache certainly makes a big difference here.
Ok, so I was happy with how it all went. The CPU is stable at 800Mhz and the ODD appears to be still working fine.
I started to pack everything up when I noticed just how damn HOT the heatsink was getting. It felt hot enough to fry an egg! Unfortunately I have no software option to monitor temps on this CPU, but it runs WAY hotter than the 7410 @ 600Mhz..
I ran the 800Mhz 7450 through the paces in Geekbench and the result was okay, not as much of a jump as I had hoped for (I was hoping it would make the 500 point mark).
The Processor prefs "Allow Nap" option (CHUD Tools) refused to work, so I figure it is running at full voltage constantly, which would explain the heat.
Concerned about the heat and looking for a comparison, I installed the 600Mhz DA 7410 into Sawtooth Beta (replacing the 400Mhz 7400), fired that one up and let it run through Geekbench. After a solid run through, I opened the side door and put my hand on the heatsink.. The 7410 produced NO heat, not even a little bit, it was still running cool to the touch.
With this gauge, I pulled out the 800Mhz 7450 and put the soldering iron to work on the PLL multipliers again. First, I dropped down to 750Mhz and ran through again. The heat was still blistering.. So I repeated the steps and dropped down to 700Mhz. Although, MUCH hotter than the 7410 at 600Mhz, I can live with the temps of the 7450 @ 700Mhz. It seems ok for now.
Running through Geekbench to compare the 600Mhz 7410 (1M L2) and the 700Mhz 7450 (256k L2 + 1M L3), they are pretty much on par, with the 7410 coming through with faster memory results and the 7450 producing a lot more heat.
I've ordered in some 12v cooling fans in various shapes and sizes (40mm and 60mm). I'll have a think about how to cool this machine with the fans, but if I can't I'll have to drop down to 650 or 600Mhz just to keep it running cool like it's older sibling processor.
(... more Geekbench results to follow)
Last edited:
Here is a summary of the comparison between the 7410 @ 600Mhz and the 7450 @ 700Mhz when running on 10.4.11 (GB v 2.2.0)
This is the 700Mhz 7450 results in Leopard (GB 2.2.7). The result is as a comparison to the initial score of 468 when first installed at 800Mhz.
When booted into OS9, Gauge Pro doesn't seem to understand the 7450 as it reported the speed as 300Mhz. Temperature readings were also not available. Apple System Profiler reported the correct speed though.