[pixelhead11]

Well I just finished setting everything up. Here is a short review after 24 hours of ownership.

Fist off this machine is silent! I have not heard a fan come on yet. This thing is Huge, it is 3" taller than my G4 and 3" deeper (you can see in the pics). The build quality is the best I have ever seen in a computer.

Lets get to the speed. I ran an xbench test. click the link to see it. I ran a Photoshop test between my G4 and G5.

Xbench Test

The test was building a gallery for Speedarena.com with a set of actions we use to build galleries. The action includes:

1. Resize Photo from 2500pix wide to 800pix
2. Crop photo to 533pix high
3. Auto Color correct
4. increase saturation 20%
5. Unsharp mask
6. Past logo bar
7. Save for web

Each photo was 12 meg in size.

The G5 1.8 has 2.5 gig of ram, the G4 Dual 1ghz has 1.5 gig of ram. Photoshop has the G5 optimized upgrade applied.

The G4 completed the test in 2 min 35 sec
The G5 completed the test in 1 min 23 sec

This is pretty amazing increase of speed. I know the G5 has 1 gig more of ram and is probably the main reason the G5 was faster but this test was an every day use test not a processor test. I may try this test later today with 1 gig of ram removed from the G5.

The over all feel is a lot snappier Dreamweaver is no longer a dog. I am amazed at how much faster Dreamweaver is.

Pictures

[moki]

Originally posted by pixelhead11:

The G4 completed the test in 2 min 35 sec
The G5 completed the test in 1 min 23 sec

[snip]
The over all feel is a lot snappier Dreamweaver is no longer a dog. I am amazed at how much faster Dreamweaver is.

Yeah that sounds about right; life will only get better once you have the OS and the app in question recompiled fully for the G5. Don't expect massive improvements, but there will be some for sure.

[TheTraveller]

Yeah I tried out a 1.6 Ghz G5 at the Apple store in San Jose. It seemed pretty snappy alright...not mind-blowingly so, perhaps, but I didn't do more than launch and toy with a bunch of the installed apps. It had 1.25 GB of RAM, but sadly, no cool games or anything to stress it with.

[blackwind]

Originally posted by pixelhead11:

This is pretty amazing increase of speed. I know the G5 has 1 gig more of ram and is probably the main reason the G5 was faster but this test was an every day use test not a processor test. I may try this test later today with 1 gig of ram removed from the G5.

Actually, I doubt that the extra gigabyte of RAM on the G5 would influence the results by any measurable amount of time. A 12-MB file would easily fit in 1.5-GB of RAM on the G4, and as such, neither computer would need to use the scratch disk.

As for the result, that is a very impressive gain, considering that it is a single-processor G5 against a dual-processor G4.

Anyway, congratulations!

[michaelb]

That benchmark test is one of the most commonsense ones I've seen here. It actually involves things that people need to do...

Two questions though: why aren't you doing the color correction and saturation gain BEFORE down-sizing the image? Wouldn't such operations have better results if there are more pixels to interpolate? (Probably not much difference for a web page, but I'm a purist. )

You know that Batch operations also depend a lot on hard drive speed? Which is of course an indirect benefit of the G5s - they come with faster drives.

Of course, since the aim of buying a G5 is to get a faster machine, it doesn't matter that the gain is not entirely down to CPU/memory bandwidth.

Anyway, thank you for posting this "here's what I do everyday" test. If only more computer users saw tests like these instead of the xBench crap...

[RooneyX]

Originally posted by pixelhead11:

The G4 completed the test in 2 min 35 sec
The G5 completed the test in 1 min 23 sec

For some reason I think a 1.8Ghz G4 wouldn't be much different from the G5 score you gave, if one existed. And 64 bit apps and an OS don't really show much in the ways of speed improvements if there is also an increasing amount of code (more lines of code are added to software yearly).

[ryarber]

Originally posted by RooneyX:
For some reason I think a 1.8Ghz G4 wouldn't be much different from the G5 score you gave, if one existed. And 64 bit apps and an OS don't really show much in the ways of speed improvements if there is also an increasing amount of code (more lines of code are added to software yearly).

First of all, doubling the clock speed of the chip doesn't equal doubling the speed of the machine. This is what Intel has been struggling with for a long time. You get to a point of diminishing returns, and a newer architecture must be designed.

Also, in his numbers, he was comparing a single processor G5 with the dual 1GHz G4. According to his numbers, if you assume that the increase in speed associated with a faster G4 would be linear, then the single G5 would be very comparable to the dual G4. Therefore, your arguement is flawed.

[superfula]

Originally posted by ryarber:
First of all, doubling the clock speed of the chip doesn't equal doubling the speed of the machine. This is what Intel has been struggling with for a long time. You get to a point of diminishing returns, and a newer architecture must be designed.

Also, in his numbers, he was comparing a single processor G5 with the dual 1GHz G4. According to his numbers, if you assume that the increase in speed associated with a faster G4 would be linear, then the single G5 would be very comparable to the dual G4. Therefore, your arguement is flawed.

You are just assuming the opposite of him. Both are assumptions, and neither can be proven right or wrong. I find it somewhat funny you try to prove his assumption wrong with your own assumption.

[ryarber]

Originally posted by superfula:
You are just assuming the opposite of him. Both are assumptions, and neither can be proven right or wrong. I find it somewhat funny you try to prove his assumption wrong with your own assumption.

I clearly stated mine was an assumption. He also inferred that his was an assumtion. My assumption was actually based on "his" assumption of a linear relationship between clock speed and real world speed. I was using his arguement to show him that his logic was flawed. If I used the assumption that the increase in performance were nonlinear, then I would come to the conclusion that a 1.8GHz G5 would in fact be substantially superior to a 1.8GHz dual G4.

These however, can be proven in time, when and if the G4 ever realizes a clock speed of 1.8GHz. With Motorola, that is a HUGE IF. However, they will undoubtedly change other things in chip design by that time, so we might actually not be comparing apples to oranges any more, but rather pears to oranges. As I stated above, they do change architecture a bit with new chip design and historically newer G4's will attempt to outperform older G4's. Also historically, there is a diminishing return even with these architectural improvements. They just don't seem to be able to keep pace with the diminishing returns unless there is a major architectural leap in which case, the chip is no longer of the same generation as the older chip. It might be a G6 by Apple nomenclature. Who knows, maybe even a G4.5.

This is beginning to sound like my logic class from years ago. I must go to bed as I'm obviously getting delirious.

[RooneyX]

Originally posted by ryarber:
I clearly stated mine was an assumption. He also inferred that his was an assumtion. My assumption was actually based on "his" assumption of a linear relationship between clock speed and real world

No. I didn't even look at the clock speeds of the slower machine. I looked at the XBench result. I'm just not that impressed with it and think its either more or less the same or not that much better than what a G4 could do at the same speed.

[blackwind]

Originally posted by RooneyX:
For some reason I think a 1.8Ghz G4 wouldn't be much different from the G5 score you gave, if one existed.

I wouldn't be surprised if a (mythical) 1.8-GHz G4 would equal a 1.8-GHz G5 in Photoshop, although this is speculation.

Most of what Photoshop does (well, involving the CPU) involves integer and Altivec work without much floating-point work. As such, the biggest advantage of the G5 over the G4, its floating-point implementation, is unflaunted. The G4 has superior integer performance per MHz, and the G5 has a slightly weaker Altivec implementation (albeit offset by the G5's far superior bandwidth).

As for drive access, I am not sure whether the G5's Serial ATA drive is actually faster. It depends on whether its noise-reducing feature is on or not. If it is on, a Serial ATA drive is actually slower than an Ultra ATA/100 drive. Of course, if this is a Dual 1-GHz Quicksilver that we are comparing to, the Serial ATA would still have the advantage (since Quicksilvers came with Ultra ATA/66 drives).

[juanpacolopez]

Originally posted by RooneyX:
No. I didn't even look at the clock speeds of the slower machine. I looked at the XBench result. I'm just not that impressed with it and think its either more or less the same or not that much better than what a G4 could do at the same speed.

It has been stated on these boards more times than I can count, and is particularly evident in this thread.

XBench is useless for determining the real-world "speed".

We finally get someone posting some excellent "real world" numbers for the G5, and still people bitch about xbench scores

[slipjack]

If you consider that the G4 that G5 was running against was a dual machine... I think that's a pretty good sign.

Basically, the Dual 2.0 machine looks to be a killer.

155secs --> 83 secs... That's a x2 speed jump, so likely we can expect < 40 secs for a dual 2Ghz machine, which would be better than x4. Not much to complain about IMHO.

This is the first time in my memory that a new-release single processor machine has destroyed a dual proc. machine... that's good isn't it?

[osiris]

Originally posted by slipjack:

This is the first time in my memory that a new-release single processor machine has destroyed a dual proc. machine... that's good isn't it?

Heck yeah. Things should only get better from here.

[D'Espice]

If you could get me that 12MB picture, I'll be willing to run these tests on my Opteron to see which one's faster: 1.8 GHz Opteron or 1.8 GHz G5

[RooneyX]

Originally posted by D'Espice:
If you could get me that 12MB picture, I'll be willing to run these tests on my Opteron to see which one's faster: 1.8 GHz Opteron or 1.8 GHz G5

But is there a plugin for the Opteron or any OS modifications to take advantage of the CPU in the way 10.2.7 does for the G5?

[pixelhead11]

Originally posted by D'Espice:
If you could get me that 12MB picture, I'll be willing to run these tests on my Opteron to see which one's faster: 1.8 GHz Opteron or 1.8 GHz G5

I can send you a file just let me know your email addi.

[AssassyN]

Originally posted by pixelhead11:
I can send you a file just let me know your email addi.

Just FYI, if you don't have an inbox w/ 12MB of space, get a email addy at www.spymac.com for something like 25MB of inbox storage for free

[D'Espice]

There is no plugin and I'd be running Photoshop either on 32-bit Windows or on 32-bit Debian Linux w/ Wine.

pixelhead: I'll send you my eMail address at uni via PM, not gonna post it here

[3.1416]

Originally posted by RooneyX:
For some reason I think a 1.8Ghz G4 wouldn't be much different from the G5 score you gave, if one existed.

The G5 was 87% faster than the G4 with only an 80% clock speed advantage, *and* it was a single against a dual. This is an oustanding result for the G5 and bodes very well for its performance in real world tasks as opposed to synthetic benchmarks.

[Eug]

One of the reasons a single 1.8 does so well vs. a dual 1 GHz may be because many Photoshop actions are not dual optimized. I mentioned this in my dual 1.25 vs single 1.8 thread, saying that many people would be happier in real-life with a single 1.8 rather than the dual G4, despite the belief by some that no matter what, a dual is always better.

Originally posted by RooneyX:
For some reason I think a 1.8Ghz G4 wouldn't be much different from the G5 score you gave, if one existed. And

Yeah, maybe, but it's irrelevant, since a G4 1.8 GHz does not exist.