Jump to content
ian128K

PowerPC 7457 upgrade for iMac G4

Recommended Posts

Back in the early 2005, Daystar was selling a PowerPC 7457 upgrade for the iMac G4 (PowerMac4,2), which replaced its original PowerPC 7450. Obviously they no longer offer this, but it does prove that you can put a 7457 in an original iMac G4 to bump it up to 1.35GHz. So, in short, I want to do this to my iMac.

 

I've been doing some research on the 7457, and one thing I've found was that there were several varieties of the 7457 made. For example, even though many sources claim that Apple never used the 7457 in their machines, the original iBook G4s did. But those were a low-power version, running around 800MHz on a 133MHz bus with a core power supply of 1.1V. (The original iMac G4, for comparison, ran on a 100MHz bus at 1.3V.) Which one? I have no clue, since I don't have an iBook G4 to crack open and look at. 

 

NXP, who used to sell them (and possibly might still?), has a listing of 9 different varieties. Unfortunately, I don't know which one Daystar used in their upgrade. I have found someone on eBay who's selling 4 of the MC7457RX1267LC chips, though, which has the right package type (CBGA483 - ceramic, ball grid array, 483 balls; 1.27 mm pitch; 29 mm x 29 mm x 3.2 mm body) to be pin-compatible with the 7450, and can run with a core power supply of 1.3V. So in theory it should be a drop-in replacement. And, in fact, the documentation I have on the 7457 mentions that "the MPC7457 is a footprint-compatible, drop-in replacement in a MPC7455 application if the core power supply is 1.3 V." (Looking in the documentation for the 7455, it mentions that "the MPC7455 is footprint-compatible with the MPC7450 and MPC7451".)

 

According to the documentation on the 7457, the 1267 in the part number means it's a 1267MHz chip. Page 50, Table 18 shows the PLL configurations for various bus-to-core multipliers to get different core frequencies. It does say to "[n]ote that these configurations were different in some earlier MPC7450-family devices and care should be taken when upgrading to the MPC7457 to verify the correct PLL settings for an application." I found a page that has several resistor configurations to change the bus multiplier on the original iMac G4's logic board. Since it's a 100MHz bus, it's easy to extrapolate the bus multipliers in the chart as 7x, 7.5x, 8x, 9x, and 10x. (Although he's got the configuration for 7.5x wrong, as I'll explain in the next paragraph.) 13.5x isn't on that chart, but we can infer that there is a configuration that sets it to a 13.5x multiplier, since that's presumably what Daystar used to get a clock frequency of 1350MHz in their upgrade. It then becomes a matter of finding what that configuration is—and I think it's moving the resistors to R252, R248, R249, R265, and R266.

 

In the documentation for the 7450, Table 16 on Page 36 shows the PLL configurations for the 7450 using a 600MHz chip as the example. The PLL_EXT value is given as a bit—either a 0 or a 1—and the PLL_CFG[0:3] value is given as a string of 4 bits. If you concatenate PLL_EXT and PLL_CFG[0:3], the bits for 7x, 8x, 9x, and 10x match what's in ThunderNerd's chart if you see the top row as 1 and the bottom row as 0. For example, with 8x, the concatenated PLL configuration is 01100; switching over to ThunderNerd's chart, he's got 800Mhz listed as bottom top top bottom bottom. The same pattern holds true for all the other whole number multipliers on his chart—they all match up with the 7450 documentation if you assume the top row is 1 and the bottom row is 0. This is why I think he got the configuration for 7.5x wrong; he's got no resistors on any of the pads, but the concatenated PLL configuration for 7.5x in the 7450 documentation is 00001.

 

So remember that note I mentioned two paragraphs ago that said the configurations were different in earlier 7450 devices? Well, if you look at Table 18 on the 7457 documentation, sure enough the configurations are different. If we assume, however, that the same pattern of top=1 and bottom=0 holds true, then the 13.5x PLL_CFG[0:4] value of 11100 should correspond with having resistors on R252, R248, R249, R265, and R266.

 

According to the PLL configurations of the 1267MHz chip in Table 18, however, a 13.5x multiplier on a 100MHz bus (to get the 1.35GHz of Daystar's upgrade) results in "core and/or VCO frequencies that do not comply with the 1-GHz column in Table 8." The part I'm confused about, though, is that there is no "1-GHz column" in Table 8. Unless they mean the 1000Mhz column? But that is for a 1000MHz chip, not a 1267MHz chip. I'm not quite sure what to make of that "warning". If anyone has any ideas, I'm all ears. But even doing, say, a 12.5x multiplier to get 12.5GHz wouldn't be that bad. That's still a 56% increase in core frequency over the original chip, which wouldn't be anything to look down on.

 

So I think I'm going to pull the trigger and buy one of those MC7457RX1267LC chips on eBay. It's $360, but I don't really mind; I love my iMac G4, and to see it get such a huge upgrade would warm my heart. I'll have to wait until I get home in September to put it in, but that just means more time for research. If anyone has any helpful input, I'd love to hear it.

 

One note: Everymac.com says that the original iMac G4 has a 7445 processor; but if that were the case, then Daystar wouldn't have been able to install a 7457 in them, since they're not pin-compatible. Mactracker.app lists it as having the 7450, which makes a lot more sense.

Edited by ian128K
Added missing link, removed strikethrough.

Share this post


Link to post
Share on other sites

I just sent off an email to XLR8 (which I believe is what Daystar became) requesting any technical info they might still have lying around on their 7457 upgrade, including 7457 part numbers, confirmation of the PLL configuration, and information on what their "21-point checkup" was. Can't hurt, right?

 

My next step is to do some research on thermal characteristics of the 7457 to see what kind of cooling I'll need to use. Obviously I'll need something that can fit the profile of the chip and inside the limited space of the iMac G4, but I figure I should try to get the most cooling I can. If anyone has any ideas, let me know! Another goal, though, is to not increase the noise of the cooling beyond what it was stock. (And if I can even decrease the noise, that would be great!)

Share this post


Link to post
Share on other sites

I would hold off on that eBay 7457 chip for now. One of the problems you are likely to encounter is that the chip won't work without a firmware patch. Apple never used the 7457, so it's not as drop-in as a 7410 in place of a 750. The same goes for 7448 upgrades in 7447 Macs, I know @dosdude1 currently has a G4 Mac Mini with a 7448 in it that won't POST anymore due to the firmware not supporting it. For the iMac, you'd have to see if Gary from Daystar still has the firmware patch and is willing to give it to you.

 

That, and the fact eBay sellers are price gouging these old chips, are good reasons to not buy one straight away. If you do eventually buy one, do try to low ball the seller with a Best Offer.

Share this post


Link to post
Share on other sites

That's a good point about the firmware, and one I hadn't considered. I haven't received a response from Gary yet, but I did find an archived post on xlr8yourmac.com that had an email from him regarding the Daystar iMac G4 upgrade, which says:

Quote

For OS9, we provide a bootx / nvram patch similar to what Giga has for the 7447. (I guess they revised the patch after PL sued them over their first 7457 patch code.)

No mention of any patch for OS X, but that doesn't necessarily mean that there wasn't one.

 

I'm starting down the rabbit hole of learning about the New World ROM boot sequence, BootX, Mac OS ROM, and OpenFirmware, as well as how to write to NVRAM, starting with this document: BootX: The Mac OS X Bootloader. There's a thread on MacOS9Lives.com about hacking the Mac OS ROM that has lots of useful information. I'm going to grab a copy of Mac OS X Internals: A Systems Approach and read through that. I've also found a post from ThinkClassic.org regarding someone's efforts to get OS 9 booting on an iBook G4, which includes some ROM hacking. Could come in handy at some point. I mean, this stuff has been done before to get a 7457 running on an iMac G4, just never out in the open.

Share this post


Link to post
Share on other sites
5 hours ago, Byrd said:

Don't forget that the 7445 G4 CPU used on some G4 iMacs and eMacs were highly overclockable, 700/800Mhz parts attaining 1.2-1.3Ghz without much issue

 

Not quite the same—I'm talking about a 7457 replacing the 7450 used in the 1st-gen iMac G4 15", as done in the old Daystar XLR8 1.35GHz MAChSpeed G4 iMac FP upgrade. The 7445 wasn't used in the 1st-gen iMac G4 15", and isn't pin-compatible. The 7455 is, but as Gary's email mentioned, there's not much of a performance increase with that chip over the 7450 because it has the same 256K L2 cache as the 7450, whereas the 7457 has that sweet, sweet 512K L2 cache. ('Just too bad the iMac's logic board doesn't have space for an external L3 cache, which the 7457 supports—although I must admit it is a little tempting to design an upgrade card with an L3 cache, but that's something maybe for a few years down the line; for now, replicating the MAChSpeed upgrade is enough.) Also, I mentioned in the OP about overclocking the 7450, and even figured out the PLL configurations to do so. But either way I don't want to overclock the 7450, I want to replace it with a 7457.

 

As an aside, it's also worth noting that Gary's email mentioned "a new CPU power supply, which we mount on the logic board as a daughtercard". So I'll have to figure that one out, too. There's so much to do! This is going to be fun. :)

Share this post


Link to post
Share on other sites

Update: I was able to find a copy of the NewerTech 7457/7447A/7448 Enabler 3.1b5. It's a bootable CD, and (according to the manual) when you boot it up, you'll see this menu:

 

---------------------------------------------------------------------------------
NewerTech 7457/7447A/7448 Enabler 3.1b5
Copyright (c) 2003-06 by Newer Technology, Inc. All rights Reserved.
---------------------------------------------------------------------------------
Mapping flash and mirroring in RAM . . .
Identifying flash. . .
Identified Micron B1 Bottom flash.
Validating flash images. . .
Identifying which areas of ROM need to be patched.
Identifying which patches have been installed.
Please select an option from the following menu:
1. Apply 7457/7447A/7448 enabler patch (7457/7447A/7448 CPUs will not boot).
2. Apply PLL patch (Apple System Profiler should report correct CPU speeds).
3. Apply L3CR extended ratios patch (L3 cache ratios above 6:1 will work).
4. Apply thermal fan control patch (Fan will turn on only when necessary).
5. Apply all patches
6. Set up NVRAM for booting into Mac OS 9.
7. Set up NVRAM for booting into Mac OS X.
8. Shut down.
>

 

I'm going to poke around in the files a bit, but I think this might negate the need to make my own 7457 enabler.

Share this post


Link to post
Share on other sites

ill mention a few things

 

1 your going to need a Firmware Patch for the 7457 to boot

 

2 I have schematics for all iMac G4s, and on the first gen ones which use 7450 CPUs, there are un-used pads for some L3 cache however I dont know exactly what needs to be populated for those to work, they do include a nifty PLL chart tho (all the PPC schematics do)

 

3 $360 is waay too much for a 7457 send the guy a best offer :)(I know the listing your referencing, im tempted to also send the guy a Best offer for 1 or 2, id like to replicate sonnets 1.2Ghz VGA TiBook upgrade, as they do provide the FW patcher for that, however I dont own a 550/667 VGA TiBook LOL)

 

4 that newer tech FW patcher, in my experience only works on PMG4s, however I have not tested it on a 4.5.x series BootROM machine thats not a PMG4, I have only tested it on 4.7+ machines where it does not work sadly.

 

hopefully this helps and let me know if theres anything else you need, I have experience doing stuff like this :)

Edited by LightBulbFun

Share this post


Link to post
Share on other sites
1 hour ago, LightBulbFun said:

4 that newer tech FW patcher, in my experience only works on PMG4s, however I have not tested it on a 4.5.x series BootROM machine thats not a PMG4, I have only tested it on 4.7+ machines where it does not work sadly.

 

hopefully this helps and let me know if theres anything else you need, I have experience doing stuff like this :)

Thanks! Do you know of a list of which Mac models have which firmware versions? I have a day off coming up on Saturday, so I'll probably spend that poking around the Newertech enabler to see how it works.

Share this post


Link to post
Share on other sites

I had a thought for the power supply. In Gary's email from that xlr8yourmac.com page, he mentioned that their 7457 upgrade required "a new CPU power supply, which we mount on the logic board as a daughtercard". Instead doing a custom job like they did, it should be possible to replace the original power supply with one from a USB 2.0 iMac G4 15" or 17". That would bump it up from 130W to 160W. According to the documentation for the 7450, the 600MHz chip (which I believe is the version used in the 1st-gen iMac G4, clocked up to either 700MHz or 800MHz, depending on the model) has a maximum power consumption of 17.5W; while according to the documentation for the 7457, the 1267MHz chip has a maximum power consumption of 25.6W. That gives us Δ8.1W, which would be more than covered by the addition of 30W from the new PSU, giving a remaining power budget of 21.9W. The 7457 documentation has a note on the power consumption values:

 

Quote
  1. These values apply for all valid processor bus and L3 bus ratios. The values do not include I/O supply power (OVDD and GVDD) or PLL supply power (AVDD). OVDD and GVDD power is system dependent, but is typically <5% of VDD power. Worst case power consumption for AVDD < 3 mW.

 

The PLL power consumption worst case scenario is on the mW scale, so we'll say that brings our budget down to 21.6W. 5% of core supply voltage (1.3V) should be ≈0.87W—assuming a current of 13.333…A (which is derived from 160W/12V), 1.3V * I = 17.333…W; 5% of that gets the ≈0.87W value. That gives a remaining budget of ≈20.73W, which is still plenty to spare; even if clocking to 1350MHz, which should only increase power consumption by 1–3W.

 

Looking at the service manual for my machine and the one for the USB 2.0, it looks like the two power supplies are swappable. A cursory Googling shows I could acquire the new PSU for only $46+shipping.

Share this post


Link to post
Share on other sites
16 hours ago, LightBulbFun said:

2 I have schematics for all iMac G4s, and on the first gen ones which use 7450 CPUs, there are un-used pads for some L3 cache however I dont know exactly what needs to be populated for those to work, they do include a nifty PLL chart tho (all the PPC schematics do)

 

It look like the whole voltage regulator area for the cache chips is not populated.

If you want to add cache and the iMac logicboard works the same way as the G4 towers you will have to set various L3 parameters in a small serial ROM somewhere.

Open Firmware will read those parameters to correctly enable the Cache on startup.

 

13 minutes ago, ian128K said:

Looking at the service manual for my machine and the one for the USB 2.0, it looks like the two power supplies are swappable. A cursory Googling shows I could acquire the new PSU for only $46+shipping.

 

Gary is not talking about the iMac's main PSU but the voltage regulator for the CPU.

Depending on what buck controller Apple used for the original CPU they had to do this because the original circuit was not stable at the lower voltage the 7457 requires.

Share this post


Link to post
Share on other sites
12 minutes ago, Bolle said:

Gary is not talking about the iMac's main PSU but the voltage regulator for the CPU.

Depending on what buck controller Apple used for the original CPU they had to do this because the original circuit was not stable at the lower voltage the 7457 requires.

Ahh, I see: the 7450 ran at 1.6V, according to the documentation. The 7457 doc says it's a drop-in replacement only if the core power supply is 1.3V. Do you know where the voltage regulator is on the 820-1257-A? (I'm assuming that's the logic board I have. Sucks being 5000 miles away from home right now.) I'm trying to find it, but the photos I have are not that high-res.

Share this post


Link to post
Share on other sites
30 minutes ago, Bolle said:

The controller itself is probably on the bottom of the board but I can't find any high-enough-res images either to tell.

I think you may be right. According to this page (though take it with a grain of salt; their photos are super low-res) the voltage regulator is a Linear Technology LTC1628CG. The chip looks like this:

 

https://www.dropbox.com/s/8kdj6xv1wo02qfj/LTC1628CG-LTC1628CG-PG-PBF.jpg?dl=0

 

I see a chip in the service manual that looks like it could be it:

 

https://www.dropbox.com/s/qdmgwav2t7zphcf/VR.png?dl=0

 

I found the datasheet for that chip. With that, I may be able to rig something to get it to step down the core voltage to 1.3V.

 

 

10 minutes ago, LightBulbFun said:

heres some iMac G4 schematics hopefully they are of use :)

 

iMac-G4-schematics.zip

Thank you very much indeed, LBF! I think that will help tremendously! According to page 22 of the 820-1257 schematics, replacing the resistor at R262 should do the trick!

 

 

Edited by ian128K
Removed typo; I think it's time for sleep...

Share this post


Link to post
Share on other sites
2 hours ago, ian128K said:

I found the datasheet for that chip. With that, I may be able to rig something to get it to step down the core voltage to 1.3V.

 

The chip will do it but you want to make sure the surrounding components can keep up with the higher current. It’s possible that you will blow the FETs (and killing the CPU in the process just like all the Cubes are dying when the VRM goes to heaven :p)

I have been doing successful processor swaps to lower voltage but higher TDP CPUs using the stock VRMs but I can imagine that some VRMs won’t like the higher current in the long run.

Daystar offered CPU swap upgrades on other machines as well without touching the voltage regulator at all so it might be needed on the circuit inside the iMac to run reliably.

Edited by Bolle

Share this post


Link to post
Share on other sites
On 22/06/2018 at 1:52 AM, Bolle said:

The chip will do it but you want to make sure the surrounding components can keep up with the higher current. It’s possible that you will blow the FETs (and killing the CPU in the process just like all the Cubes are dying when the VRM goes to heaven :p)

 

Going through the schematics, I've identified every MOSFET on the 820-1257 board that's not in an IC.

 

BOOTBANGER (Page 21)

- Q36, Q37: 2N7002

 

Voltage Regulators (Page 22)

- Q16: 2N7002

- Q30, Q33, Q35: IRF7822

- Q38: IRF7822 (NOSTUFF)

 

DC/DC CONVERTER 5V AND 3.3V POWER CONNECTOR (Page 27)

- Q11, Q12, Q14, Q20: 2N7002

- Q19, Q29, Q31, Q32, Q34: IRF7805

 

PMU SERIAL DL, LPWR CTRL, TMDS POWER (Page 28)

- Q7, Q9, Q13, Q23: Si3443DV

- Q6, Q8, Q10, Q15, Q25, Q27, Q28: 2N7002

 

TUMBLER AUDIO, HEADPHONE DRIVER (Page 30)

- Q5: 2N7002

 

TUMBLER AUDIO, POWER AMP. (Page 32)

- Q22, Q24, Q26: 2N7002

 

I think only the MOSFETs on Page 22 would be affected by the voltage change to VGER_VCORE, as none of the outputs from the CPU go to any components with MOSFETs. Something interesting related to VGER_VCORE that I still need to figure out: on Page 33, that signal is listed as being 2V, but on Page 22 it says pretty plainly that VGER_VCORE is set for 1.65V. (Which would make sense, as that's the core voltage of the 7450/7451.) All the other signals on that page that have their set-for voltages declared match with Page 33, so I need to figure out why there's a discrepancy for VGER_VCORE.

 

Also, the voltage regulator on Page 22 is identified as an LTC1735CS, not an LTC1628 like I originally thought. (Though there is a LTC1628 at U31 on Page 27.)

 

Some or all of the MOSFETs on Page 22 may need heatsinks, which can be calculated. I just need to figure out what their Vgs is going to be with the resistor change on R262 so that I can then find what the Rds will be according to the datasheet, but I'm having a hard time following the voltage through the LTC1735CS on the schematic. I think I may have to make a test model of the circuit once I get home in September, but I'll keep trying to work this out on paper in the meantime. Maybe some Thevenising might help.

Share this post


Link to post
Share on other sites

Looks like someone actually paid $360 for one of those 7457s on ebay :O

 

im a bit annoyed by that, not that I mind what someone spends their money on but I dont want these sellers thinking they get away with charging such obscene prices for such out dated and old CPUs... :/

Share this post


Link to post
Share on other sites

Sheesh!

 

I could probably get some new-ish Core i7 (i don't know, maybe Skylake?) for that price! Granted, probably a low-end one, but one nevertheless.

 

I'll never understand some people's compulsion to engage in price gouging.

 

c

Share this post


Link to post
Share on other sites

I have a dead Daystar that needs a PPC601 chip but for $100 and desoldering effort I could just buy another Daystar (I won't).

 

People need to make money, so you get price gouging. Some people think all this stuff should be free, until they give something valuable away and see the buyer flip it on Ebay 5 minutes later.

 

 

Share this post


Link to post
Share on other sites
1 hour ago, Unknown_K said:

People need to make money, so you get price gouging. Some people think all this stuff should be free, until they give something valuable away and see the buyer flip it on Ebay 5 minutes later.

True, but the effect of someone buying the way overpriced $360 7457 has already caused another eBay ad to pop up for the 7457... also at $360. That seller saw it sell for that utterly insane price, and now wants in on it. And so, the price will remain that high.

Share this post


Link to post
Share on other sites

You assume somebody else doesn't have that chip and will want a quick sale at maybe $300 and somebody else will end up undercutting that. Prices stay high when there is great demand or very little supply.

 

Some items are so cheap on Ebay I don't see how they make any money.

 

 

Share this post


Link to post
Share on other sites

I wouldn't mind so much if it weren't for the fact that ebay apparently represents "Best Offer" amounts accepted as the original asking price, not the price actually paid. That leads to rampant inflation!

 

At least if it were true that someone ACTUALLY paid xyz inflated price, there would be an argument to be made for "market will bear," etc.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×