[Link]

How much of a bottleneck is a 250mhz bus on a 2.5ghz processor?

Oh yeah right it's supposed to be 1.25ghz!�

Well, actually it's 5x250mhz. I guess you guys forgot what DDR really was *cough* marketing *cough*

[Eug Wanker]

Originally posted by Link:
Well, actually it's 5x250mhz. I guess you guys forgot what DDR really was *cough* marketing *cough*

DDR on the G4s is marketing. DDR on the G5 (esp. dual channel) makes a noticeable difference. BTW, I don't know where you get this 5x250 business.

It's 2x2x200 for the memory, which is the limiting factor.

[gizzard]

FYI, a 1.7 GHz 7447A is about equal to a 1.4 GHz 7455B in performance.

[Eug Wanker]

Originally posted by gizzard:
FYI, a 1.7 GHz 7447A is about equal to a 1.4 GHz 7455B in performance.

No it isn't. Remember that the 7447/7447A has 512 KB of L2 cache. The 7455 has only 256 KB. ie. only half the L2 cache. Thus, clock for clock, the 7447A is significantly faster than the 7455.

I'd say a 1.7 GHz 7447A is equal to about a 1.9 GHz 7455 in performance.

A 1.7 7447A might be in the same ballbark as a 1.7 GHz 7455 with 2 MB L3 though.

[P]

There's a lot of "facts" being thrown around here about L2 and L3 caches, DDR and whatnot. Perhaps it would be better to tone down the absolute truths a bit, because the world is a bit more complicated than it may seem. Different apps require different stuff from the CPU.

DDR on a G4 only helps marginally, because the system bus is slower than the memory bus. It does help because DMA for IDE and AGP transfers are essentially free (as in they don't steal bandwidth from the CPU), but that's marginal.

Don't think of the difference between data bus/system bus and the CPU speed - think about each figure in its own context. Is what you're doing CPU bound? Then the bus speed doesn't matter. Does it depend on memory bandwidth? Then an upgrade such this won't help at all, and you shoudl consider a G5. Even the slowest iMac G5 available has 4 times the system bandwidth if you have a 133 MHz system bus - and that really means 4 times the speed in some applications. Rather few operations depend on both, although of course it does happen. Video operations are notorious for requiring a lot of bandwidth. Compression and encryption applications are CPU bound, to give another example. If you're unsure, go to some benchmark site and start reading relevant benchmarks.

FWIW, I'd prefer a 512K L2 over a 256K L2 + 2 MB L3. The lower L2 latency is worth the loss of the L3 in mixed applications. The L3 does save the situation a bit when the application is bandwidth limited, but in that case you should probably consider a G5 anyway.

[gizzard]

Originally posted by Eug Wanker:
No it isn't. Remember that the 7447/7447A has 512 KB of L2 cache. The 7455 has only 256 KB. ie. only half the L2 cache. Thus, clock for clock, the 7447A is significantly faster than the 7455.

I'd say a 1.7 GHz 7447A is equal to about a 1.9 GHz 7455 in performance.

A 1.7 7447A might be in the same ballbark as a 1.7 GHz 7455 with 2 MB L3 though.

Let me clarify. A 1.7 GHz 7447A with 512 KB L2 cache and no L3 cache performs approximately the same as a 1.4 GHz 7455B with 256 KB L2 cache and 2 MB L3 cache. Also, I'm not guestimating.

[Weezer]

Originally posted by gizzard:
Let me clarify. A 1.7 GHz 7447A with 512 KB L2 cache and no L3 cache performs approximately the same as a 1.4 GHz 7455B with 256 KB L2 cache and 2 MB L3 cache. Also, I'm not guestimating.

elaborate then. If you arnt guessing, do you have one already to test?

[Eug Wanker]

Originally posted by gizzard:
Let me clarify. A 1.7 GHz 7447A with 512 KB L2 cache and no L3 cache performs approximately the same as a 1.4 GHz 7455B with 256 KB L2 cache and 2 MB L3 cache. Also, I'm not guestimating.

Based on what, and in what situation? It really varies on the context. People have done some tests with L3 turned on and off, and even with the same 256 KB L2 (not the 512 of the 7447), they don't lose as much performance as you suggest. Using your statement, you might expect a 1.7 GHz 7455 with the L3 turned off to behave like a 1.2 GHz 7455, and that is rarely the case.

Anyways, a 1.4 GHz 7455 is fairly hot. Sonnet doesn't make a 7455 1.4 for the Cube for example.

[gizzard]

On average.

There's this little birdy that keeps buzzing around my head whispering these things to me.

[ibookmark]

Ah but now I am starting to drool over upgrading my now puny 867 to one of these 1.7 dps!

[SilentEchoes]

I am all over getting a dual 1.7GHz upgrade for my dual 533.

After having a Dual Processor machine, I will always take the extra time to save for a Dual Processor box as long as I have the option to.

I see a lot of people tell other people don't bother going with the duals, they are only used by some applications and wont give THAT big of a speed difference. As far as I am concerned this is BS.

I have done plenty of playing around on my machine and serveral others, Though no actual benchmarking because I think benchmarking is a crock, All I care about is how fast I can get my work done, and how well I can enjoy my play. BUT I have seen this dual 533 get roughly the same performance while using it for work, as a G4 800/100 iMac, G4 1.3/100 Tower, and a G4 1.0/133 doing the same stuff on all of them. I say Roughly the same because somtimes its better sometimes its worse.

Anyways just my 11� -- adjusted for inflation.

[Simon]

Originally posted by iREZ:
I have no clue on how to install one of these in a cube, but I know of one laying around that I could get a good deal on, would something like this be easy for the everyday joe to install or should I just stick to my plan of getting a G5 iMac later down the road?

It looks as if you could use the complete disassembly guide for the Cube from Apple. That should give you a good idea of what you'll be having to do in there.

Good luck!

[brucejy]

Look at this:

http://www.gigadesigns.com/newsMDD7447.html

I'm still going for the Dual 1.7 due to power/heat issues and the fact that everyone recommended a dual for me in my topic a few weeks ago.

[Weezer]

"From the tests we have performed to date with the 133MHz bus AGP equipped Power Macs, we have determined the performance of a 7447A processor running at 1.733GHz to be equivalent to that of a 7455 running at 1.467GHz. The 7447A does not support an L3 cache, and in most applications requiring a heavy CPU load this has a big impact on performance. "


blah, that's a little disappointing.

[gizzard]

The little birdy was right!

[MORT A POTTY]

so, what was this rumor about PowerLogix again? nobody ever elaborated on it.

I could really use a speed boost though on my single 867Mhz Quicksilver. I've already got an SATA controller and HD and a gig of RAM and a 4x SuperDrive... next is CPU and GPU... or a whole new machine which would cost roughly twice as much and yield that much performance increase

[Cadaver]

Originally posted by Weezer:
blah, that's a little disappointing.

Well, at least now I can stop trying to figure out how to afford a new processor for my G4. Its doesn't really need an upgrade at all (currently running a 7455B @ 1.4GHz + 2MB L3), but I too am a MHz whore.

[Eug Wanker]

I still have my doubts about the benches. I've seen benches on G4 Power Macs which do not agree. If they had said 1.4 7457 with L3 equalling 1.7 7447A, I'd believe, it, but a 1.4 7455 with L3 equalling a 1.7 7447A is not in line with what I've seen.

It could be true in specific situations, but I'd highly doubt it to be true overall.

Then again, most of what I've seen is with 1 MB L3 per CPU. This is 2 MB per CPU.

Whatever the case, I'm getting the 1.7.

[iomatic]

I hope I'm not thread-jacking too much, and sorta low-end in relation to this topic, but here's a question:

I have a (Sawtooth? AGP.) 350MHz I'd like to upgrade, say to 500MHz, or a 1x multiplier per bus speed. Are there cheapy upgrades, free or the like available-- it's for a home file/web server, and would help I think in improving speed. The HD is already fast-- 7200rpm?

Thanks!

[Eug Wanker]

Originally posted by iomatic:
I hope I'm not thread-jacking too much, and sorta low-end in relation to this topic, but here's a question:

I have a (Sawtooth? AGP.) 350MHz I'd like to upgrade, say to 500MHz, or a 1x multiplier per bus speed. Are there cheapy upgrades, free or the like available-- it's for a home file/web server, and would help I think in improving speed. The HD is already fast-- 7200rpm?

Thanks!

Sawtooth upgrades

[Zubir]

The new 1.7 chips are equal in speed to the old 1.4's. There's an article on xlr8yourmac.com that has Giga confirming this. The only advantage the 1.7 has is lower power consumption, which would make it a better choice than the 1.4 if your system has a low wattage power supply, like a cube or sawtooth.

[Lateralus]

Originally posted by Zubir:
The new 1.7 chips are equal in speed to the old 1.4's. There's an article on xlr8yourmac.com that has Giga confirming this. The only advantage the 1.7 has is lower power consumption, which would make it a better choice than the 1.4 if your system has a low wattage power supply, like a cube or sawtooth.

Thank you Dr. Timely.

[Eug Wanker]

Originally posted by Zubir:
The new 1.7 chips are equal in speed to the old 1.4's. There's an article on xlr8yourmac.com that has Giga confirming this. The only advantage the 1.7 has is lower power consumption, which would make it a better choice than the 1.4 if your system has a low wattage power supply, like a cube or sawtooth.

This post from a PowerLogix dual 1.2 tester suggests that the L3 has nowhere near that speed boost. I don't know what software he tested though.

[QuadG5Man]

I'm curious about a dual 1.7 G4 Cube setup. Can any G4 Cube be upgraded?

[Lateralus]

Originally posted by QuadG5Man:
I'm curious about a dual 1.7 G4 Cube setup. Can any G4 Cube be upgraded?

Well, there isn't a dual upgrade for the Cube.
And there is only one G4 Cube.

[QuadG5Man]

Got it. A G4 Cube is only upgradable to a single 1.7. Doh! Thanks Lats...

[MORT A POTTY]

Originally posted by QuadG5Man:
Got it. A G4 Cube is only upgradable to a single 1.7. Doh! Thanks Lats...

I would hardly say "Only" a 1.7 in a G4 Cube... who would have thought that two years ago?

[Eug Wanker]

Originally posted by Lateralus:
Well, there isn't a dual upgrade for the Cube.

There is one (or two) coming supposedly... I decided not to wait for it. I'd rather have a single 1.7 than a dual 1.2 anyway. A dual 1.4 might be nice, but the power and heat are a little concerning.

[Lateralus]

I'm indecisive about which machine to upgrade again. I had been gung-ho on the Dual 1.73 for the Digital Audio, but the Cube arrived today. It is mint. I am in love.

What I am wondering is if, given the supposedly small power consumption of these 7447A upgrades, there would be any reason why a Dual 1.7GHz Sawtooth/GigE upgrade wouldn't work in a Cube?

The 7447A @ 1.7GHz apparently uses less than half the power of a 1.4GHz 7457, so there's certainly enough juice in the Cube to power a dual. And from what I can tell, all of Giga's 7447A upgrades use the same PCB, so the dimensions would be the same.

[Eug Wanker]

Originally posted by Lateralus:
I'm indecisive about which machine to upgrade again. I had been gung-ho on the Dual 1.73 for the Digital Audio, but the Cube arrived today. It is mint. I am in love.

What I am wondering is if, given the supposedly small power consumption of these 7447A upgrades, there would be any reason why a Dual 1.7GHz Sawtooth/GigE upgrade wouldn't work in a Cube?

The 7447A @ 1.7GHz apparently uses less than half the power of a 1.4GHz 7457, so there's certainly enough juice in the Cube to power a dual. And from what I can tell, all of Giga's 7447A upgrades use the same PCB, so the dimensions would be the same.

Huh? The 7447A and 7457 actually have about the same power ratings at the same clock speed, at least on paper according to Freescale. Now it's probably true that the 7447A has been more tweaked and it's lower power, but with no inside info I'd expect the 1.7 GHz 7447A not to use that much less power than the 1.4 GHz 7457 - they'd probably be in the same ballpark, unless a very low voltage variant was used. The problem with the low voltage variants though is that they're $$$.

Oh and that's why I'm not exactly confident in the rumoured dual Cube upgrades from PowerLogix. With an updated video card and a hotter hard drive than the stock one, a dual G4 Cube at this point is really pushing it.

P.S. It's nice to reclaim that space under my desk that used to be taken up by a PIII. It'd be a shame to fill again with a tower Power Mac.

[LeeG]

Originally posted by Zubir:
The new 1.7 chips are equal in speed to the old 1.4's. There's an article on xlr8yourmac.com that has Giga confirming this. The only advantage the 1.7 has is lower power consumption, which would make it a better choice than the 1.4 if your system has a low wattage power supply, like a cube or sawtooth.

Do the sawtooth machines actually have a problem with the 1.4's? Power wise?

After the giga article about 1.7 is roughly the old 1.4 with L3, I can't see spending the extra cash. Even if those benchs are slightly off, maybe in SOME situations, its marginally better. for 25% price increase - not worth it. Save the $100 for some other mac toy - thats my plan. OWC 1.4 - I plan to order this week, and I am psyched.

Lee

[gizzard]

First:
The 512 KB L2 cache of a 7457 gives it about a 5% speed boost (in games) over a 7455 with 256 KB L2 cache.

Second:
A processor upgrade with 2 MB L3 cache was between 20-30% faster (in games) than an upgrade without.

[Eug Wanker]

Originally posted by gizzard:
First:
The 512 KB L2 cache of a 7457 gives it about a 5% speed boost (in games) over a 7455 with 256 KB L2 cache.

Second:
A processor upgrade with 2 MB L3 cache was between 20-30% faster (in games) than an upgrade without.

OK, I will assume this is true with games then. However, in terms of the Cube, it's useless for gaming anyway, because of the GPU.

If it's representative of other actions overall, then you (and Giga Designs) are right, it doesn't really matter either way... It's 6 of one and a half dozen of the other:

1.4 GHz 7455 + 25% boost from added L3 = 1.75 GHz equivalent*
1.7 GHz 7447A + 5% boost from added L2 = 1.79 GHz equivalent*

* - Compared to 7455 without L3.

Except that 7447A will be much lower power and heat, which is important for the Cube. Plus 1.7 looks better under Xbench, cuz it doesn't test cache very well.

I'd like to see some numbers though (CHUD?).

[Eug Wanker]

Here are some numbers.

The statement that in games the boost from 2 MB L3 can make up to a 30% boost for 7455 chips is accurate, at least when comparing a Giga Designs 800 MHz 7455 with no L3 vs. an 800 MHz 7455 with 2 MB L3.

OTOH, in things like iMovie and iTunes, the difference is less than 8%. In Photoshop, the 2 MB L3 gives about an 11% boost.

So I guess we're both right.

[MORT A POTTY]

what I'd be interested in is the 7448 which will have some key improvements. it will be a die shrink to 90nm, support for OoOE in AltiVec, have 1MB of L2 Cache (can be ECC if desired) and will use less than ten watts at 1.4Ghz. I say it will be perfectfor a Cube. it's also a drop-in replacement for the 7447 and 7447A. Freescale says it will be "At Least 1.5Ghz"

I'm quite optimistic about this chip and might hold off on an upgrade till one utilizing this processor is availible. the extra cache alone will be very important.

It also has 10 year reliability at 105 degrees centigrade.

[MORT A POTTY]

Originally posted by Eug Wanker:
Here are some numbers.

The statement that in games the boost from 2 MB L3 can make up to a 30% boost for 7455 chips is accurate, at least when comparing a Giga Designs 800 MHz 7455 with no L3 vs. an 800 MHz 7455 with 2 MB L3.

OTOH, in things like iMovie and iTunes, the difference is less than 8%. In Photoshop, the 2 MB L3 gives about an 11% boost.

So I guess we're both right.

and that won't be entirely accurate at all. the one without L3 cache still only has 256K of L2, whereas the 7447A has twice that. this means the performance numbers are not even close to accurate when being related to the difference between a 7455B w/ 2MB of L3 and a 7447A with no L3, but 512K of L2.

[Eug Wanker]

Originally posted by MORT A POTTY:
what I'd be interested in is the 7448 which will have some key improvements. it will be a die shrink to 90nm, support for OoOE in AltiVec, have 1MB of L2 Cache (can be ECC if desired) and will use less than ten watts at 1.4Ghz. I say it will be perfectfor a Cube. it's also a drop-in replacement for the 7447 and 7447A. Freescale says it will be "At Least 1.5Ghz"

I'm quite optimistic about this chip and might hold off on an upgrade till one utilizing this processor is availible. the extra cache alone will be very important.

I'm interested in the 7448 also, but I can't wait that long to upgrade this Cube. I just scrapped my PC, and the Cube's G4 450 is far too slow to take its place as my primary home desktop.

After I get the 7447A for the Cube, I think my next CPU upgrade will be a G5 iMac.

Originally posted by MORT A POTTY:
and that won't be entirely accurate at all. the one without L3 cache still only has 256K of L2, whereas the 7447A has twice that. this means the performance numbers are not even close to accurate when being related to the difference between a 7455B w/ 2MB of L3 and a 7447A with no L3, but 512K of L2.

I agree completely. I was just commenting on the statement earlier that 2 MB L3 provides a 20-30% boost. I didn't believe it, but it is true... for games. It doesn't overall, and I'd be using iMovie way more than I'll be playing Quake III on my Cube.

The doubled L2 cache is a separate issue. It should provide at least a 5% boost, but in some cases it might be much more than that.

[MORT A POTTY]

yeah. I will try to play some higher end games on mine though, that's why I'm also upgrading my video card to a Radeon 9800 Pro. I've already got a fairly large SATA HD (and controller) in my G4 and a gig of RAM... so after these two major upgrades, it'll be pretty maxed out on upgrades.

[Weezer]

Im still debating what to do on this. Is the dual 1.7 going to be like a new machine for me, coming from a single 933?

I think Im now debating between the single 1.467, dual 1.2 (7455) or dual 1.7 (7447)

I was hoping for a price drop on the dual 1.33 (7455) but its still the same price as the dual 1.7

[Eug Wanker]

Originally posted by Weezer:
Is the dual 1.7 going to be like a new machine for me, coming from a single 933?

Of course. Night vs. day. It's a dual, with almost twice the MHz for each individual CPU, too.

[Chinasaur]

Weezer and I have the same machine. I'm really looking forward to this upgrade Glad I waited.

Thanks for all the information all

[Weezer]

Originally posted by Chinasaur:
Weezer and I have the same machine. I'm really looking forward to this upgrade Glad I waited.

Thanks for all the information all

have you placed your order yet? did they give you an estimated ship date?

The price of the old duals has dropped 100 each, so the dual 1.2 is now 500 and the dual 1.3 is 600. Would the dual 1.3 play games better because of the 2 meg cache than the dual 1.7?

[MORT A POTTY]

I'd say the dual 1.7 would game a little bit better than a dual 1.3, but I really don't feel comfortable saying that without any evidence whatsoever... I'm waiting for some benchmarks and for PowerLogix to reveal whatever it is they have up their sleeves... but, I will be buying a dual processor upgrade very soon, and I will make damn sure it's the fastest one available at the time.

[QuadG5Man]

Weezer, a dual CPU G4 at a higher clock WILL seem like a new machine compared to your single 933, especially if you multi-task with heavy CPU load in different programs! Dual G4(5) systems are much more productive because of this. The dual systems never 'slow down to a crawl.'

Does the digital audio series Single G4 466 support Dual G4 upgrades? I was thinking that might be cheaper unit to upgrade to get a dual G4 1.2-1.7 setup with a 133 Mhz bus, instead of upgrading a sawtooth. That makes sense, no?

Qg5

[Link]

Originally posted by QuadG5Man:
Does the digital audio series Single G4 466 support Dual G4 upgrades? I was thinking that might be cheaper unit to upgrade to get a dual G4 1.2-1.7 setup with a 133 Mhz bus, instead of upgrading a sawtooth. That makes sense, no?

Qg5

Cheaper? A 466 DA is worth a bit more than a sawtooth since it has a 133mhz bus instead of a 100mhz bus

and yes, it'll work.

[CincyGamer]

<<Sorry didnt get a chance to flame back until now>>
Yes, the G5 can do more simultaneously than the G4. The G5 can have like 217 simultaneous instructions in-flight than the 10 or so the G4 can do. The Megahertz myth is a complete joke from Apple, because now if you look at apple.com in the G5 hardware section they totally contradict themselves from the Megahertz Myth.

People that said they do not want a 1.6 G5 versus a faster 1.7 Ghz G4 PB in this thread..
The current G4 tops out at system bus of 167 MHZ!!. It has been shown on barefeats that DDR memory does not make a difference in G4 PowerMacs even with the latest 1.4 GHZ models. In fact, the there were a couple of benchmarks where the DDR memory was slower than the SDRAM. This fact is why Apple could never come out with a Quad-CPU G4 system because they could not get past the system controller bottleneck. This has been well-documented.

I have done several personal benchmarks vs my Intel computers at work and home and with regular computing tasks (non-mathematical) like working with XML or regular programming
the G4 is pretty close to a 1:1 linear with a P4. So, on a non-numerical tasks, A 3 GHZ P4 takes about 3x as long as my 1 GHZ G4 PB. Alot of the numerical comparisons are based on the compilers such as comparing IBM XLC vs GCC is huge difference.

I am irritated that my minimum spec G4 1 GHZ with 1 GB RAM cannot play Knights of Old Republic. It was too slow after the first couple levels. I analyzed what was happening using Shark (developer profiler) and Thread Viewer and the code was written to be mostly single threaded. 95% of the program is running on one thread (open loop). There were some other threads running but they did almost nothing. I looked at a better written game such as Warcraft III and there is a huge difference. So, this KotOR is CPU bound according to Shark. (Check for yourself) Like 12% of the code was running in strlen().. and this was with the newest patch 1.03b.

So, where I am going with this is that I need some RAW CPU speed for some of these games that are CPU bound or not ported very well. KotOR is not ported well versus a game like WoW that plays much better on the same hardware and is a more complex game. The G4 has reached the point where it does not show any improvement no matter how much faster it is because the system bus is too slow. This has been documented on Ars-Technica and some other people have worked out this using EE formulas. So, 1.7 GHZ or 2.0 GHZ G4 is not going to make your apps faster. Also on bear feats they have done massive tests showing that on a Powermac G4 the new video cards are a waste of money and do not go any faster. Do the search on bearfeats.com

The initial G5 SINGLE 1.6 Ghz PowerMacs were slightly slower than the Dual 1.42 G4 PowerMacs. But that's good!! This is similiar to getting a 2.2 GHZ G4 CPU if things were so linear in the one CPU PB world. Fact: a new PB is not even matching the minimum Doom 3 specs. Doom 3 is advanced game, but it does run on P3 and older video cards. G4 is really an old architecture that cannot scale anymore due to system bus limitations.

[CincyGamer]

sorry I wrote bearfeats instead of barefeats (I do that often) :-)

[Lateralus]

I smell a G5 fanboy...

[CincyGamer]

According to Apple-history.com
The first G4 was in the PowerMacintosh G4 in September 1999 at a speed of 400 MHZ. Fast forward a little over 5 years later, the G4 is stuck at 4.25x the orginal speed in 5 years!!. This is why its time to make the true jump to GHZ speed that only the G5 can provide (we should see 3 GHZ announced soon).

Think about how much personal computing has changed and how much software has changed since 1999. Its time for a new era. I own a G4 (and still paying it off on my Apple Loan @$200 a month - ouch !!).
Not a fanboy, but just excited that we have a kick butt machine again.

[Weezer]

Originally posted by CincyGamer:
According to Apple-history.com
The first G4 was in the PowerMacintosh G4 in September 1999 at a speed of 400 MHZ. Fast forward a little over 5 years later, the G4 is stuck at 4.25x the orginal speed in 5 years!!.

So the first G5s clocked in at a max of 2 GHZ. This means that in 5 years you will upset if we have a 8.5 ghz G5?