@sailor

Thats a single G4 Sonnet card that came with the PMac I bought when I thought I had burned the dual 1.8GHz Sonnet that had used before. Turned out if was just the motherboard so now the dual one is collecting dust.....



This should be the one.

@ Kronos
thank you. So it should be some version of 7457 ( probably the last, the same I have, 1267 MHz ) and modded VCore, which is unknown for me.
Because this Sonnets are rare, I thinked you did CPU swap and play with overclocking and VCore. My mistake.

Fortunatelly I have nearly all necessary info from macrumors, last piece I missing is exact increased VCore for 1.8 GHz. So I will try 1.75 GHz + some calculated VCore value.

>> This should be the one.

> So it should be some version of 7457

Sonnet Encore/ST G4 1.8 GHz (SG4-1800, SG4-1800L) use slightly overclocked MPC7447A (or more likely MPC7447B/C).

Quote:

According this page Sonnet Encore 1.8 GHz used also 7457, probably the last fastest type MC7457 RX1267LC. This CPUs can run at frequencies 1.5-1.8 GHz. Also in manual from Kronos link is L3 cache visible ( but there lower freq ).
If Kronos has 7457 or 7447 can be easily checked by L3 presence (7457 has, 7447 hasn't).

But all this is irelevant details, point of message was, that I ask for real experience with VCore + overclocked 7457. And if Kronos made no modification, he not touched anything.

>> Sonnet Encore/ST G4 1.8 GHz (SG4-1800, SG4-1800L) use slightly
>> overclocked MPC7447A (or more likely MPC7447B/C).

> According this page Sonnet Encore 1.8 GHz used also 7457

Thanks for the reference, but I believe it to be incorrect. Up to 1.4 GHz is definitely 7457 yes, maybe even 1.6 GHz is (doubtful), but 1.7 and 1.8 GHz are definitely later CPUs (7447A/B/C or maybe even 7448). I doubt they sold 42% overclocked CPUs.
See there for another reference that even says it has 7448: https://beta.ivc.no/wiki/index.php/PowerMac_G4_Upgrades#Upgrade_candidates

There're also some hints in the original Sonnet press releases for the 1.7 and 1.8 GHz versions in 12/2004 and 08/2005, respectively:

"These cards are based on the latest and fastest PowerPC G4 processor (used in Apple’s current PowerBook line) [...], and feature 512K of on-chip L2 cache."
http://web.archive.org/web/20060821124418/http://www.sonnettech.com/news/press/pr2004/pr120704_sg4_prcdrop.html

"These cards, based on the latest and fastest PowerPC G4 processor (used in Apple’s current PowerBook line) [...] feature 512K of SRAM on-chip L2 cache"
http://web.archive.org/web/20061111032204/http://www.sonnettech.com/news/press/pr2005/pr080105_prcdrp.html

Apple switched their PowerBooks from 7447 to 7447A in 04/2004 and to 7447B in 01/2005, so this cleary hints at those G4 iterations being used by Sonnet in their 1.7 GHz upgrade in 12/2004 and their 1.8 GHz upgrade in 08/2005.

I have a hunch that the SG4-1800 came with 7447B and the SG4-1800L with 7448, but I have no proof on that. It would explain why for the 1.8 GHz model, the official source hints at the 7447B but another source mentions the 7448.

> in manual from Kronos link is L3 cache visible ( but there lower freq ).

Yes, I don't expect them to have changed the product pictures in the instruction manuals for later L3-cache-less versions just because the chip arrangement on the PCB is different, as it has no bearing on the installation process.

> If Kronos has 7457 or 7447 can be easily checked by L3 presence
> (7457 has, 7447 hasn't).

745x processors can be deployed without connecting actual L3 cache to the corresponding pins. The PCB doesn't even have to have any footprint for L3 cache chip. In this case, any 745x CPU acts exactly like the corresponding 744x CPU. But I'm with you that in the case of the Sonnet Encore/ST G4 cards, 745x comes with L3 cache deployed.

I am not insist on 7457 with Sonnet 1.8 GHz single. I just looking for real info about overclocked 7457 VCore. Information if Sonnet had 7447 or 7457 is redundant for this question. Of course can be useful for somebody who search for exact Sonnet type.

Quote:

Yes, 7457 can be connected without L3 cache. For example some AmigaOne XE CPU modules has 745x without cache chips. But it makes no big sense in case of Sonnet - 7447 had these times higher default frequency and probably was cheaper. And L3 cache gives great benefit with certain tasks ( like compiling ) see here. There several posts, graph is #171.
And if slower Sonnets with single CPU has L3 cache for sure (picture in manual) and their design was made for L3, very probably L3 was also in 1.8 GHz version, if there was 7457 CPU.

> I am not insist on 7457 with Sonnet 1.8 GHz single.

I know :-) I just want to explain myself why I'm confident that the lowendmac.com list is incorrect in this regard. Feel free to ignore my ramblings if you don't find them interesting or useful.

> I just looking for real info about overclocked 7457 VCore. Information
> if Sonnet had 7447 or 7457 is redundant for this question.

Not quite, I think. If Sonnet really used 7457 CPUs overclocked up to 1.8 GHz, the info you're seeking may be derivable somehow from (preferably non-destructively) examining such Sonnet card. If they didn't, they are not.

>> I'm with you that in the case of the Sonnet Encore/ST G4 cards,
>> 745x comes with L3 cache deployed.

> 7457 can be connected without L3 cache. [...]
> But it makes no big sense in case of Sonnet

Yes, absolutely.

> 7447 had these times higher default frequency

I'm not sure what "default frequency" means here, but yes, in terms of maximum rated frequency, during the first year, the 7457 was 1267 MHz vs. 1300 MHz of 7447. After that, the 7447 got downrated to 1267 MHz as well. This happened around 03/2004.

> > I just looking for real info about overclocked 7457 VCore. Information
> > if Sonnet had 7447 or 7457 is redundant for this question.

> Not quite, I think. If Sonnet really used 7457 CPUs overclocked up to 1.8 GHz, the info
> you're seeking may be derivable somehow from (preferably non-destructively) examining such
> Sonnet card. If they didn't, they are not.

I never thought Kronos had a Sonnet - they are rare. At first I supposed to find somebody who made CPU swap and overclocked 7457 to 1.8 GHz. I found such guy, but he don't know exactly the VCore, because after several month with 1.8 GHz he lower frequency because of high fan noise ( i.e. not ideal cooling )
If somebody has Sonnet, it will be fine, but probably he don't want to remove cooler and investigate settings.

> I'm not sure what "default frequency" means here, but yes, in terms of maximum rated
> frequency, during the first year, the 7457 was 1267 MHz vs. 1300 MHz of 7447. After that,
> the 7447 got downrated to 1267 MHz as well. This happened around 03/2004.

and after that 7447B/C rated frequency rises upto 1700 MHz, and frequency of 7457 stayed on 1267 MHz all the time.

> 7447B/C rated frequency rises upto 1700 MHz

Already 7447A had 1.7 GHz maximum rated frequency, as discussed a year ago :-)

> frequency of 7457 stayed on 1267 MHz all the time.

Interestingly, there was a 1333 MHz 7457 listed in 2006 as "No Longer Manufactured" (MC7457RX1333PC*, RoHS-compliant replacement part was 1267 MHz), while 7447 was kept at 1267 MHz.
With the advent of the higher-clocking 7447A, Motorola/Freescale simply EOL'ed the 745x line. I don't remember I ever read an official explanation as to why no 7457A/B/C or 7458 was released.


* I also found mention of MC7457RX1333PB, but not sure if genuine or fake part number

Quote:

Not really, at least back in the day when MorphOS started PMac_G4 support you could pretty much always find something on ebay, often as part of a full system (the fact that I could replace my thought to be broken 2x1.8 at short notice is proof to that).

What was rare and is pure unobtainium today are the 2GHt Nevertech units or the 2x1.8 Sonnet for MDD/FW800.

Quote:

If I will find such card, and there will be 7448 inside ( it should be) it will be fine. But is still very interesting to look how the caches speeds all things. Short info from graphs in pre-previous post:

7448 ( L2: 1 MB ) - 2.0 GHz = 108 %
7457 ( L2: 512 kB + L3: 2 MB ) - 1.6 GHz = 100 %
7447 ( L2: 512 kB ) - 2.0 GHz = 86 %

For me will be enough to clock 7457 CPUs somewhere between 1.58 - 1.75 GHz for ideal performance/cooling/consumption.
Even if CPU can go to 1.83 GHz, in this case will be slightly over max frequency also my L3 cache chips and I am too lazy to modified this.


And now I return to topics:
Please, can you recommend me device for measuring of real computer consumption?
Ideally some tested, with corrected errors of blind currents and with connection to standard plug?

>> the 2GHt Nevertech units or the 2x1.8 Sonnet for MDD/FW800

> If [...] there will be 7448 inside ( it should be)

There've been two versions of the Newer Technology MAXPower G4 2.0 GHz (which is not for MDD/FW800), an older one with 7447A(/B/C) and a newer one with 7448. The Sonnet Encore/MDX G4 Duet 1.8 GHz (XG4D-1800) only has 7447A(/B/C).

Is the clock speed stated on Wikipedia (G4) the theoretical max cycles the cpus can reach?

Because mine is listed as 500–867 MHz whereas if I do a cli c:cpu, I get the wonderful :

System: PowerMac3,1
Machine: 0x1
PageSize: 4096 Bytes
CPUCount: 1
CPU0: 7451 (G4) Version 32768 Revision 515
CPU0: CPUClock 1600000000 BusClock 99630669
CPU0: FPU Unit exists
CPU0: Altivec Unit exists
CPU0: Performance Monitor Unit exists
CPU0: DataStream Unit exists
CPU0: ReservationSize 32
CPU0: Ticks 4
CPU0: CacheL1Type 0x0
CPU0: CacheL1Flags 0x1000F
CPU0: ICacheL1Size 16384
CPU0: ICacheL1Lines 512
CPU0: ICacheL1LineSize 32
CPU0: DCacheL1Size 16384
CPU0: DCacheL1Lines 512
CPU0: DCacheL1LineSize 32
CPU0: CacheL2Type 0x0
CPU0: CacheL2Flags 0x1001D
CPU0: ICacheL2Size 524288
CPU0: ICacheL2Lines 0
CPU0: ICacheL2LineSize 0
CPU0: DCacheL2Size 524288
CPU0: DCacheL2Lines 0
CPU0: DCacheL2LineSize 0

^Which is basically double the clock compared to what wikipedia states.

> Is the clock speed stated on Wikipedia (G4)
> the theoretical max cycles the cpus can reach?
> Because mine is listed as 500–867 MHz whereas if I do a cli c:cpu,
> I get the wonderful :
> [...]
> System: PowerMac3,1 [...]
> CPU0: 7451 (G4) Version 32768 Revision 515
> CPU0: CPUClock 1600000000 BusClock 99630669
> [...]
> ^Which is basically double the clock compared to what wikipedia states.

Yes, 867 MHz is the maximum rated clock frequency of the 7451. It can certainly reach nowhere near the MorphOS-stated 1.6 GHz. That's just MorphOS using bogus base values for some reason and thus calculating wrong CPU clock. Nothing to worry about and has happened before. It may have to do with your CPU card not being the original 350...500 MHz 7400 one.

Hmm, well, I do have a cpu accelerator on board but I can't be sure it's a Sonnet or not. What I am sure about is that the system is no way near 800-900Mhz target value because I get easily 75 fps on Quake 3 at 1280x1024 all max settings with multiple enemies on screen. Also, I have a powerbook 1.67Ghz and apart from faster system buses, cpu intensive tasks are almost identical. Finally as stated in this LW benchmark, my system is double faster than the 800Mhz microA1, faster than the pbook 1.33Ghz and much closer to the X5000 2.2Ghz.

I did some spoofing with OF but got nowhere since the cpu gets reported as /PowerPC, 60?@0 (I think this was an issue with some Sonnet boards, which were not reported properly and needed some patching?), whereas clock-frequency too states: 05f03e4d which is the hex for 99630669 decimal (in hz) which in return is equal to ~99.63Mhz and thus it must be referring at cpu bus instead of actual clock speed (is my logic correct so far?).

I also tried to burn an Lubuntu 16.04 LTS but as as soon as I fire up the terminal, system hangs. Couldn't be bothered to disconnect my devices as this is my main home setup, it's MorphOS exclusive, and don't want to interfere with it. Is there any linux live ppc distro with cmd prompt to check systems specs /wo entering the desktop mode nor installing it at all?

To me, it looks more probable that MorphOS could be reporting wrongly cpu version instead of clock speed, unless I have a miraculous 7451 G4 and Wikipedia is just full of bs one more time.

Sonnets usually have a purple heatsink and 1.6GHz versions do exist. So quick look and a picture coul clear that up.

The plot thickens :D

Here's a photo from the top and another one from the side.

Furthermore, according to this site the only possible candidates for this type of PowerMac (Sawtooth) are 4 options.and the only relevant photo I could find on the net is this one (from some Japanese website, claiming to be indeed a FastMac).

If the above are true then this is a 7455 @ 1.53Ghz (?), hence there is a chance that MorphOS could be reporting both cpu and clock wrong, wtf?

From the benchmark posted above (and the overall experience), I am pretty sure that given its speed this must be a cpu with a massive L3 cache in order to keep and/or beat my 1.67Ghz 7447A pbook in some areas of cpu intensive tasks. Also, the system was used for professional DTP (Photoshop) around 13 yrs ago when I bought it, which makes the 7455 even more attractive as an upgrade back then.

I 'll try to locate and burn another version of Lubuntu to fire a terminal and check with lshw command to see what it states.

> the only relevant photo I could find on the net is this one

The archived FastMac website has more photos.

> claiming to be indeed a FastMac

I just found you reported on your "FastMac 1.6Ghz G4 7451" already two years ago :-)

> If the above are true then this is a 7455 @ 1.53Ghz (?), hence there is a
> chance that MorphOS could be reporting both cpu and clock wrong, wtf?

According to the DIP switch table on page 4 of the installation guide, the 7455 on the card can be clocked up to 1.50 GHz on 100 MHz bus machines and up to 1.53 GHz on 133 MHz bus machines, so 1.50 GHz should be maximum on your Sawtooth. Either way, both "7451" and "1.6 GHz" seem (slightly) wrong.

> this must be a cpu with a massive L3 cache

Both 7451 and 7455 can support up to 2 MiB L3 cache, and according to the above-linked website, the 7455 on the card is fully equipped.