Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Hey, @JoyBed, earlier in the thread, you mentioned the L3 cache chips in the MDD are clocked at 300MHz, and that you found faster, 370MHz chips that are 100% compatible instead. Do those chips have a name or product ID, so I can search for them? Do they come in sizes of both 2MB and 4MB?





My friend doing all the hard work for me somehow managed to pull this off with the SC7448 chips today:

View attachment 1747195

The beast is out!
Very Awesome stuff! would love to see some benchmark figures :) bet its very quick!, would be cool to see a picture of the CPU card itself too

I see the CPU Type: name in system profiler is "PowerPC G4" given that this is generally taken from OpenFirmware, im guessing you used the Sonnet Firmware patch in the end?

(which corrects it from the PowerPC 60? you get with 7447-7448's etc)
 
Hey, @JoyBed, earlier in the thread, you mentioned the L3 cache chips in the MDD are clocked at 300MHz, and that you found faster, 370MHz chips that are 100% compatible instead. Do those chips have a name or product ID, so I can search for them? Do they come in sizes of both 2MB and 4MB?





My friend doing all the hard work for me somehow managed to pull this off with the SC7448 chips today:

View attachment 1747195

The beast is out!

Question for you all who know the nitty-gritty:

How significant a performance penalty does a system experience when moving from a maxed-out/overclocked dual-745x set-up to, with the loss of L3-caching, a dual-7448 set-up? Or, is the loss of L3 relatively unremarkable?
 
Question for you all who know the nitty-gritty:

How significant a performance penalty does a system experience when moving from a maxed-out/overclocked dual-745x set-up to, with the loss of L3-caching, a dual-7448 set-up? Or, is the loss of L3 relatively unremarkable?
It depends on the application you are running, the largest limiting factor of the G4 is bus and memory bandwidth in some situations, like 3d gaming or other complex render models that move massive amounts of data across the bus or into out of memory.

Something like encoding DV footage to MPEG2 would benefit from the higher clock speeds of the 7448, because that's just not a massive about of data, but encoding from ProRes RAW to h.26x, if you had the software to do it on the G4, would be a lot more data moving, and would tend to benefit more from the L3 Cache, but just not a great deal more.

So there is a trade off between more raw clock speed and more cache, given the fact that the G4 didn't support DDR bus or Memory.
 
Question for you all who know the nitty-gritty:

How significant a performance penalty does a system experience when moving from a maxed-out/overclocked dual-745x set-up to, with the loss of L3-caching, a dual-7448 set-up? Or, is the loss of L3 relatively unremarkable?
Well yes as @DearthnVader said it depends on application you want to run. Some will benefit from the L3 some will not. But the L3 is generally better to have as it compensates the slow G4 bus. The 7448 has the ability to go to 2,2GHz but the MDD cant benefit from the 7448 that much, the 7448 can shine in the DLSD for example where you can bump up even the bus to the stock 200MHz the 7448 supports and the lower power draw of it while maintaining higher clockspeeds and having bigger caches and improved core. The MDD on the other hand cant go on the bus higher than 167MHz so it cant benefit from the faster bus of the 7448 whereas it can take advantage of the 7457 having the ability to have double the L3 at higher speed and also higher clockrates of the core and the cache clocks and also higher caches compared to 7455. The thing i was constantly thinking about in last days was not just making a whole new MDD motherboard but also making a new DLSD motherboard which would be easier done as I have the schematics for it. Interesting thing is that the northbridge of the DLSD is dual CPU capable and there also was a prototype dual DLSD. Imagine the creation of a whole new motherboard with dual 7448 for the DLSD with bus at 220MHz and core clock at 2,2GHz.
 
  • Like
Reactions: B S Magnet
Imagine the creation of a whole new motherboard with dual 7448 for the DLSD with bus at 220MHz and core clock at 2,2GHz.
That would be even more interesting than that PowerPC "modern" laptop project that is going on. If that happens, I'm jumping in. I just wish the DLSD was a bit more OS-9-friendly in regards to the GPU and the trackpad. (I.e. Radeon 8500 and/or NVidia Ti4600, and an ADB trackpad.)

The 167MHz MDD bus means the 7448 works like a 7447A (including 7447B) that runs faster, cooler and with twice the L2 cache. And that's also all I could ask for! But indeed, I would love it if the day came where we could equip our MDDs with a 200MHz (220MHz?) bus. It's almost inspiring me to go and learn hardware electronics to try to slowly reverse-engineer the whole thing...

I like the idea of having both overclocked 7448 cards and overclocked/maxed out 7457 cards, and have both do real world tests. Games, Photoshop, virtual machines, compression/decompression tools etc.. This comparison will be done, as long as I'm breathing. I know both cards will show their strengths, and I encourage every PPC fan to secure their own pair.

Very Awesome stuff! would love to see some benchmark figures :) bet its very quick!, would be cool to see a picture of the CPU card itself too

I see the CPU Type: name in system profiler is "PowerPC G4" given that this is generally taken from OpenFirmware, im guessing you used the Sonnet Firmware patch in the end?

(which corrects it from the PowerPC 60? you get with 7447-7448's etc)
Once I receive it (it will still be a while, I think a month), I'll take pictures. We were at first using the Giga Designs patcher, but expanded the attempts to consider Sonnet's and Powerlogix/newertechnology's patches, as well. Ultimately, I'm not sure yet which of the 3 were used to generate the screenshot. But once I receive it, I will be trying all 4 patches (Sonnet offers 2 versions; I will try both), and see how it all behaves, and share the findings here.

There's also a single 7448 that is currently being worked on, to see if it can surpass 2.0GHz. Only time will tell.
 
  • Like
Reactions: LightBulbFun
The 7448 has the ability to go to 2,2GHz but the MDD cant benefit from the 7448 that much, the 7448 can shine in the DLSD for example where you can bump up even the bus to the stock 200MHz the 7448 supports

Thanks for answering my first question. My next question is what voltage setting is the bus running if a 7448 is to operate at 2.2GHz? While I was poring over the MPC7448 PDF from NXP, my impression was that the MC7448HX1700LD at 1.3V would operate at 1.7GHz.

1616569951506.png


What voltage setting is used for 167MHz bus clock speeds on the DLSD for OEM 1.67GHz MPC7447A chips, and what voltage setting would be required for a 1.7GHz 7448 to run on a 167MHz bus and on a 200MHz bus? I guess more key to the question here is, does anyone here have a chart showing the possible (over)clock speeds for a 1.7GHz 7448 on a DLSD board and what voltage settings correspond to bus speeds (and CPU speeds) on the DLSDs?
 
That would be even more interesting than that PowerPC "modern" laptop project that is going on. If that happens, I'm jumping in. I just wish the DLSD was a bit more OS-9-friendly in regards to the GPU and the trackpad. (I.e. Radeon 8500 and/or NVidia Ti4600, and an ADB trackpad.)

The 167MHz MDD bus means the 7448 works like a 7447A (including 7447B) that runs faster, cooler and with twice the L2 cache. And that's also all I could ask for! But indeed, I would love it if the day came where we could equip our MDDs with a 200MHz (220MHz?) bus. It's almost inspiring me to go and learn hardware electronics to try to slowly reverse-engineer the whole thing...

I like the idea of having both overclocked 7448 cards and overclocked/maxed out 7457 cards, and have both do real world tests. Games, Photoshop, virtual machines, compression/decompression tools etc.. This comparison will be done, as long as I'm breathing. I know both cards will show their strengths, and I encourage every PPC fan to secure their own pair.


Once I receive it (it will still be a while, I think a month), I'll take pictures. We were at first using the Giga Designs patcher, but expanded the attempts to consider Sonnet's and Powerlogix/newertechnology's patches, as well. Ultimately, I'm not sure yet which of the 3 were used to generate the screenshot. But once I receive it, I will be trying all 4 patches (Sonnet offers 2 versions; I will try both), and see how it all behaves, and share the findings here.

There's also a single 7448 that is currently being worked on, to see if it can surpass 2.0GHz. Only time will tell.
just wish the DLSD was a bit more OS-9-friendly in regards to the GPU and the trackpad. - we are working on that as we speak. OS 9 will be getting a makeover soon.. 9.2.3 - its a long process.
 
Thanks for answering my first question. My next question is what voltage setting is the bus running if a 7448 is to operate at 2.2GHz? While I was poring over the MPC7448 PDF from NXP, my impression was that the MC7448HX1700LD at 1.3V would operate at 1.7GHz.

View attachment 1748338

What voltage setting is used for 167MHz bus clock speeds on the DLSD for OEM 1.67GHz MPC7447A chips, and what voltage setting would be required for a 1.7GHz 7448 to run on a 167MHz bus and on a 200MHz bus? I guess more key to the question here is, does anyone here have a chart showing the possible (over)clock speeds for a 1.7GHz 7448 on a DLSD board and what voltage settings correspond to bus speeds (and CPU speeds) on the DLSDs?
With the voltage set by Apple for the 7447 the 7448 can already go beyond that 1,7GHz. Altho its a notebook so it has a crappy cooling so I wouldnt raise it too much, maybe 0,1V more but thats it... My overclocked 7447 is using stock voltage.
 
  • Like
Reactions: B S Magnet
I can get supplies of G4 and G5 cpus for swapping for cheap
Where are you getting all those 7457s from? I bought an MC7457RX1257LC back in 2018 and have been keeping an eye out for them over the last couple years, but I haven't seen any for sale since. (Today I actually spotted some MC7457RX1000LB and MC7457RX1000LC chips, but I'm really only interested in the 1257 versions and they're selling for quite a bit of money anyway.)
 
Thanks for answering my first question. My next question is what voltage setting is the bus running if a 7448 is to operate at 2.2GHz? While I was poring over the MPC7448 PDF from NXP, my impression was that the MC7448HX1700LD at 1.3V would operate at 1.7GHz.

View attachment 1748338

What voltage setting is used for 167MHz bus clock speeds on the DLSD for OEM 1.67GHz MPC7447A chips, and what voltage setting would be required for a 1.7GHz 7448 to run on a 167MHz bus and on a 200MHz bus? I guess more key to the question here is, does anyone here have a chart showing the possible (over)clock speeds for a 1.7GHz 7448 on a DLSD board and what voltage settings correspond to bus speeds (and CPU speeds) on the DLSDs?
With voltage, you want the lowest stable voltage per you clock speed. The 7448 max voltage is 1.32v, anything much over 1.3v may burn the CPU, so you have to see what clock speed the chip will do by binning it yourself.
 
There are some ebay sellers from china selling salvaged chips. Some of the 7447 and 7455 chips are new ("NOS"), but the 7448 and 7457 look like pulled and re-balled parts, but they seem to be worth the gamble when a "genuine" part is a few hundred bux.

For voltages and such, there are datasheets with all this info. e.g.:
 

Attachments

  • Picture 201.png
    Picture 201.png
    32.9 KB · Views: 137
Now, that I have a 15 Inch DLSD G4, I may have Joybed do this. Sounds kewl.. I did see dual G4 in this PowerBook somewhere on the forum. Again, radically kewl to have a dual G4 PowerBook. :)
 
There are some ebay sellers from china selling salvaged chips. Some of the 7447 and 7455 chips are new ("NOS"), but the 7448 and 7457 look like pulled and re-balled parts, but they seem to be worth the gamble when a "genuine" part is a few hundred bux.

For voltages and such, there are datasheets with all this info. e.g.:
I have quite a lot of those salvage 7448s and 7457s. The 7457s are great, and haven't had any issues so far as I've used them (on MDD cards). The 7448s, however, are a bit more finicky. A couple of the ones I have ended up having very small cracks in the ceramic interposer, making them completely useless. The ones that did work (most of them) were able to run at 2.0 GHz with some careful VCORE voltage tweaking. I also recently reverse engineered the system firmware of some later PPC Macs, and was able to successfully patch them to add 7448 support (see https://forums.macrumors.com/thread...g-a-pencil-to-overclock.1973903/post-29596069).
 
Question for you all who know the nitty-gritty:

Having a bigger/faster caches is always better than smaller/nonexistent/slower caches. However, when comparing these chips the later versions usually clock higher. So the question is more complicated if you're asking about something like a 1.6GHz 7457 with L3 cache compared to a 2GHz 7448 without an L3 cache but with a larger L2 cache.

Bus speed is also a factor, and obviously when comparing 100, vs 133, vs 167MHz--faster is better. I've noticed that the presence of a 2MB L3 makes a bigger difference on a slower bus. A lot of the 7455 and 7400 G4 computers have 1MB of backside cache (instead of 2MB).

I have a few different CPUs that I could run some tests on if there's something in particular that people want to see.
 
  • Like
Reactions: Macbookprodude
Having a bigger/faster caches is always better than smaller/nonexistent/slower caches. However, when comparing these chips the later versions usually clock higher. So the question is more complicated if you're asking about something like a 1.6GHz 7457 with L3 cache compared to a 2GHz 7448 without an L3 cache but with a larger L2 cache.

Bus speed is also a factor, and obviously when comparing 100, vs 133, vs 167MHz--faster is better. I've noticed that the presence of a 2MB L3 makes a bigger difference on a slower bus. A lot of the 7455 and 7400 G4 computers have 1MB of backside cache (instead of 2MB).

I have a few different CPUs that I could run some tests on if there's something in particular that people want to see.
Yes, I do have one request

I want to see if the G4 1Ghz Titanium PowerBook with 1MB L3 is faster than the 1.67 DLSD with no L3 and only 512K L2. Thanks
 
The DLSD has a higher clock, larger L2, and faster bus speed. In general I think it would be faster. You'd have to find a specific case where some CPU intensive code will fit in the 1MB L3 of the 7455 that will not fit in the 512k L2 of the 7447. Perhaps some particular photoshop filter or something like that would favor the Ti.
 
I found some cinebench scores for CPU render: 1GHz Ti 101 1.67GHz PB 163.

That is a pretty simple CPU task though, and it mostly scales with clock speed.
 
Ok, so that answers my question.. so, clock speed is also relative to the performance ad not just having 1MB L3 Cache. I must say I feel my DLSD is slower, but then again SSD it does not have while my TIBook 1ghz has a 512GB M2 SATA. Now, would an MDD G4 1.42 be much faster than the DLSD though it clocks at 1.67 ?
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.