*Almost* Fastest PowerBook ever, and more to go.

[asaggynoodle]

Armed with the Engineering Schematics, PCB files, and a soldering iron we have achieved PowerBook glory!

http://imgur.com/a/Yviy6

GeekBench:

https://browser.primatelabs.com/geekbench2/2603039

With a score of 1136 we've claimed the seat of the highest Geekbench Powerbook Benchmark ever!

Thanks to the 90nm 7448A chip, it has some pretty controllable thermals. Once I get some more AS5 I'll be going further as there is plenty of headroom here both thermally and Voltage wise.

I'm going to expect somewhere in the ballpark of 2.3Ghz - 2.5Ghz, being limited by the thermal capacity of the cooling solution when it's all said and done.

 

[TheWhiteFalcon]

Close, so close, but it's not the fastest PowerBook ever, not yet (at least not from a benchmark standpoint).

https://68kmla.org/forums/index.php?/topic/21144-haplains-never-ending-quest/?p=263061

 

[asaggynoodle]

I'll beat that by the end of next week, I promise. I'm expecting another 20% out of this baby as I'm still on stock voltage. 

Never seen that before, not on the GB database either. Hmm.

 

[TheWhiteFalcon]

It's a one of a kind prototype dual G4 PowerBook, and I believe it had hardware issues that prevented a net connection.

 

[Elfen]

Now this I like!

Excellent work AsaggyNoodle!

 

[asaggynoodle]

It's a one of a kind prototype dual G4 PowerBook, and I believe it had hardware issues that prevented a net connection.

Very strange machine indeed, no doubt a proper dual would have scored much higher than that. 

@Elfen Thanks Elfen, stay tuned there will be much more to come as this is just the beginning. I think I've figured out the Voltage Tables, so I should be good to go as high as the Cooling system allows me to go. Even at that, I plan to lower the Die temperature ~10-15c by applying some actually decent performing thermal compound rather than the "Toothpaste" that's on this thing. Personally I'd just use some Liquid Gallium, but my only concern is if any sudden jerks while under load occurred some could get on the Aluminum on the heatsink... Gallium + Aluminum = No more Powerbook. 

Dual SSD's in RAID 0 (Currently 5400RPM 320GB) will soon be making an addition in the coming weeks. I fully intend to make this the fastest "Consumer" made Powerbook ever made in every way. I'm going to be losing the Airport card in this thing, as it's litterally slowing my entire network down, so I'm going to be getting an N+ solution in the works. 

I already attempted to see if the Memory Controller would detect 3/4GB of ram, but no dice. A single 2GB stick works fine, but anything else just doesn't show up (boots fine just not acknowledged).

Practically speaking, 2Ghz is just enough to allow for a realistically usable modern web surfing experience on a modern browser like TenFourFox. Only reason it performs like a turd is because there is no Hardware Acceleration. If OpenGL was supported like it is in Morphos you'd probably see a double in Web Performance. But for the time being, 2Ghz + N is a VERY nice jump in usability.

 

[Cory5412]

This is really great and interesting work!

If you have time, I would be really interested in seeing the CPU results for the CPU benchmark test in Cinebench R11.5 on this thing.

 

[asaggynoodle]

Give me an hour or two and I'll get that posted for you!

BTW: If anyone knows any more good comparable benchmarks (As mentioned above) let me know. 

As for stability, this thing is a rock. I mean that as in VLC (480p), Virtual PC w/ Windows 98SE Running Landmark Benchmark, Geekbench 2.2.7, and Halo running at Extreme settings; at the same time of course. I ran this setup ALL day yesterday, not a hiccup or change seen. Personally I'm blown away to think that this chip had 20%+ performance headroom without ever screwing with the core voltage yet. I suppose lowering the fab size from 130nm to 90nm will do that. For everyone that is thinking "Melting laptop," I can assure you I've measured an almost indifferent thermal specification from the chassis perspective. The Laser Thermometer tops out at about 109*f (which is what stock is), which leads me to believe there is a bottleneck caused by the poor thermal transfer characteristics of the "Toothpaste" on the heatsink.

 

[360alaska]

I think this has been done before: http://www.xlr8yourmac.com/G4CARDS/daystar_1.92GHz_PB_G4.html

Although, I think they used 130nm chips?

 

[asaggynoodle]

Here we are, Cinebench 11.5 CPU score. From my research we're along the lines of a 1.67-2.00Ghz Intel Atom n450. Which I guess is kind of a thumbs up, as the atom wasn't introduced for another 3-4 years after this guy was.

Spoiler

And for the sake of a few laughs I figured I'd compare it to my main laptop... Luckily it's only 2761.15384% faster

Spoiler

@360alaska, Yes I'm sure it's been done many of times... But not anywhere near where I'm going. That upgrade they installed should have never happened by the way, obviously unstable-- which they should have caught right away if they had tested it at any length. Clearly nowhere near enough vCore to run that stable (But already too much as noted by temperatures), but none the less that 7447A eats mV for breakfast and runs WAY too hot. That's how I've achieved an even higher clock speed without even a noticable temperature differential. This overclock would not be possible on a 7447A or any 130nm chip without some sort of custom cooling solution/liquid cooling. But that would defeat the purpose of this being in a real, usable laptop I can take with me wherever I go.

 

[360alaska]

Is there any chance you might eventually be willing to offer an upgrade service ?

 

[asaggynoodle]

Honestly If everything here goes to plan I don't see why not. After all this is said and done I might go back and make a Video Series on YouTube about the steps involved and how to do it yourself. What to look out for-- certain signs relating to certain configs that need to be adjusted etc. Maybe even talk about the chips themselves, I finally talked to the owner at EveryMac.com, and got the A1138/A1139 Specs changed to reflect the 7448A (As even apple screwed up on that-- LONG STORY). The hardest part about all of this, is the actual component work. The resistors themselves we're dealing with are 0603 size, a lot of people don't know how small that is:

Literally an exhale too close and they are gone.

 

[360alaska]

OH,

I thought you swapped the CPU to 7448A?

I should be able to swap those little guys, I have before:

http://www.fordtaurus.net/forums/viewtopic.php?t=17765

 

[asaggynoodle]

That's the general consensus. This is exactly why it took me so many months of constant looking for an A1139, it was the only one that had the 7448A. Funny enough, Apple's own Internal Engineering team made an "Oops" and screwed up the Design datasheets. Obviously this was later corrected before production, but it appears that the Marketing department didn't get the memo. Therefore the A1139 will always live in infamy (in my mind) as the machine that was always labeled as the 7447A w/ DDR2. Which technically isn't possible.

 

[360alaska]

So what about the 1.67 ghz 15 incher (A1106)?

 

[Unknown_K]

A1095 and A1106 were DDR1 models.

 

[Cory5412]

Thank you, asaggynoodle!

The 15-inch PowerBook G4 1.67GHz with DDR2 memory gets about 0.22 points, in stock configuration.

I've always thought that was an interesting result, because a 1.8GHz (a midrange CPU option) ThinkPad T30 from 2003 was getting the same score. I don't know what a 2.4 or 2.6GHz model would get.

Does the 2.0GHz system feel appreciably faster at other tasks, or is it really just web browsing situations with tenfourfox?

 

[Unknown_K]

7448A G4 chip 1MB of cache, 200FSB

7447 G4 chip 512K cache, 133? FSB

So more cache and faster FSB would mean it should feel faster unless the G4 design was maxed out so speed increases didn't make much difference.

 

[asaggynoodle]

Did a quick video of the machine for you, sorry about quality/vertical video etc. Don't have any camera equipment, but it's something I plan on getting over the next month or so.

I'll be doing a stress test next so please stay tuned. But it's noticably faster in all situations no doubt. Virtualization is totally transformed, and can do just about anything within reason. Games are only limited to the Radeon 9700 (Which will be overclocked soon). Anything within 10.5.8 itself is very snappy, probably to 10.4.11 levels of quickness now. I should try 1080P video playback in VLC to see how that works. Should be able to do 720p pretty smooth i'd guess. Any applications in particular you'd like to see?

https://www.youtube.com/watch?v=v4mkqdi53kQ

@360alaska only the A1139/A1138 (Late 2005 - Early 2006) models have the 7448A chip. They (I believe) have the same motherboard/PCH/Video etc. Identical just different displays.