I received the following email response from the guy over at Barefeats:
"I ran the Photoshop test with a 400MB file and got the same time on the Power Mac SDRAM. I'll try it on the DDR model but I doubt that's the issue."
[On Thursday, August 15, 2002, at 08:15 AM, Eirik Iverson wrote:
Would you mind running some significant more memory intensive tests? I believe the Photoshop test you ran involved a 30 MB file. Frankly, I thought that would be memory intensive. Maybe something 90 MB or bigger might reveal a greater performance delta.]
Are the flaws with these tests no completely apparent? Tests like "Altivec Fractal" are tied to the processor, and do not require that information be moved from RAM to the CPU. They are
completely bound by the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...
The simple bump from 133Mhz bus to 167Mhz bus should improve some things regardless. These limited tests prove absolutely nothing.
<strong> [quote]The simple bump from 133Mhz bus to 167Mhz bus should improve some things regardless. These limited tests prove absolutely nothing.</strong><hr></blockquote>
I think that's what we're wondering about though - we had an FSB speed increase but no corresponding increase in performance with these tests. Many are saying it's due to the smaller L3 cache, but whatever the case maybe, if these tests are accurate (and I believe they are), then anyone in the market for anything other than the 1.25GHz should buy one of the previous line if possible.
<strong>[qb]anyone in the market for anything other than the 1.25GHz should buy one of the previous line if possible.</strong><hr></blockquote>Why? The new one is substantially cheaper than the old one.
<strong>Are the flaws with these tests no completely apparent? Tests like "Altivec Fractal" are tied to the processor, and do not require that information be moved from RAM to the CPU. They are
completely bound by the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...
The simple bump from 133Mhz bus to 167Mhz bus should improve some things regardless. These limited tests prove absolutely nothing.</strong><hr></blockquote>
That is what I needed to see. A logical explination.
[quote]Tests like "Altivec Fractal" are tied to the processor, and do not require that information be moved from RAM to the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...<hr></blockquote>
The altivec unit did not change so why would an altivec test change what so ever between these machines at the same processor speed.?
Just because the BareFeats guy doesn't know what number the new G4 is does not mean that he can't run benchmarks. I guess we'll soon enough have independent corroboration or contradiction, but I'm wiling to believe these numbers. Still though, I am surprised.
Now some here are complaining that these aren't the right benches to run. But look: he has a Photoshop test, a 3d test, an audio (well, itunes) test, and an altivec test. I mean, if the new G4 doesn't speed up any of these tests, I don't think it would speed up any of the work I would do with it.
The reason why the Altivec fractal test should be surprising is that for months here, there has been a contingent who have claimed that Altivec is data starved. If this were true, then a faster bus should have returned significant improvements here (with or without ddr). If we believe this test, it did not. Hey: where is Team Bandwidth today?
I said that if the tests are true, anyone in the market for a Power Mac right now should try to get one of the previous models, unless said individual(s) are looking for the 1.25GHz machine.
[quote]Originally posted by BRussell:
<strong>Why? The new one is substantially cheaper than the old one.</strong><hr></blockquote>
I should have said anyone looking for a dual 1GHz should look for the old model rather than the new one. (Obviously if you're on the low end, then you should get the new 867 since it's a dual.) But why would someone want to get the old dual over the new dual? If their speeds are equivalent, then why wouldn't one want to buy the old dual? The new one is listed at macmall.com (a major reseller) for $2494, while the old one is listed for $2194. It looks like anyone who buys one of the new dual 1GHz will be cheated, as long as we assume Bare Feats is right.
Most of the points I outline beneath have already been stated.
It seems like a lot of you are missing the point here. Bearfeats ran tests of the old "Fastest" configuration against the new "Faster" configuration. It makes sense to me that the Altivec Fractal results would be the same. It also makes sense to me that Apple would not include as much RAM in the new mid-range model as in the older high-end model. Wait a couple of months for the 1.25 dual to ship. The extra MB of L3 cache/processor as well as all of the new motherboard improvements should really shine through and give us some truly amazing results.
Wait for XLR8yourmac's review whenever that comes out. I consider that to be the most accurate place to obtain performance numbers.
[quote]According to Motorola sources, a tweaked version of the Apollo 7450 G4, the 7470, will be ready for volume production shortly after the end of Q2 [of 2002], in time for a summer ramp. The 7470 will be manufactured on a 0.13 micron process, allowing for a smaller die size with room for 512K of L2 cache, and support up to 4MB of DDR-SDRAM L3 cache. <hr></blockquote>
And that bit right there indicates that Apple declined this chip, asked for overclocked G4s to "wrap things up" for the Fall, and is now dancing with a new partner for Q1 2003.
Barefeats = made up pseudo tests without connection to reality.
The same guy used to rave about how fast Macs where, still under RDF, and when that failed all of a sudden PCs were always faster...
And no he even posts stuff liek new machines being slower than old ones. Even if it's possible (which it is, due to the smaller cache), who gives a damn?
Real benchmarking with comparable results is a very hard thing to do. I've done it myself and it takes several days to finish a series, not several minutes.
Comments
I AM SciFi
-onlooker
"I ran the Photoshop test with a 400MB file and got the same time on the Power Mac SDRAM. I'll try it on the DDR model but I doubt that's the issue."
[On Thursday, August 15, 2002, at 08:15 AM, Eirik Iverson wrote:
Would you mind running some significant more memory intensive tests? I believe the Photoshop test you ran involved a 30 MB file. Frankly, I thought that would be memory intensive. Maybe something 90 MB or bigger might reveal a greater performance delta.]
Eirik
<strong>He'll try it? Does that mean he hasn't yet? How did he get those benchmarks without trying it?</strong><hr></blockquote>
I've had no further contact and have no further info from him so I can not answer your good questions.
completely bound by the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...
The simple bump from 133Mhz bus to 167Mhz bus should improve some things regardless. These limited tests prove absolutely nothing.
I think that's what we're wondering about though - we had an FSB speed increase but no corresponding increase in performance with these tests. Many are saying it's due to the smaller L3 cache, but whatever the case maybe, if these tests are accurate (and I believe they are), then anyone in the market for anything other than the 1.25GHz should buy one of the previous line if possible.
<strong>[qb]anyone in the market for anything other than the 1.25GHz should buy one of the previous line if possible.</strong><hr></blockquote>Why? The new one is substantially cheaper than the old one.
<strong>Are the flaws with these tests no completely apparent? Tests like "Altivec Fractal" are tied to the processor, and do not require that information be moved from RAM to the CPU. They are
completely bound by the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...
The simple bump from 133Mhz bus to 167Mhz bus should improve some things regardless. These limited tests prove absolutely nothing.</strong><hr></blockquote>
That is what I needed to see. A logical explination.
[quote]Tests like "Altivec Fractal" are tied to the processor, and do not require that information be moved from RAM to the CPU. Since both machines have the exact same CPUs it's not suprising that they give the same results...<hr></blockquote>
The altivec unit did not change so why would an altivec test change what so ever between these machines at the same processor speed.?
Makes sence
Now I feel better.
With or without a signature
I Am still SciFi
-onlooker
Some people still refuse to accept the truth: 7470 is just a product of MacOSRumors.
Now some here are complaining that these aren't the right benches to run. But look: he has a Photoshop test, a 3d test, an audio (well, itunes) test, and an altivec test. I mean, if the new G4 doesn't speed up any of these tests, I don't think it would speed up any of the work I would do with it.
The reason why the Altivec fractal test should be surprising is that for months here, there has been a contingent who have claimed that Altivec is data starved. If this were true, then a faster bus should have returned significant improvements here (with or without ddr). If we believe this test, it did not. Hey: where is Team Bandwidth today?
[quote]Originally posted by BRussell:
<strong>Why? The new one is substantially cheaper than the old one.</strong><hr></blockquote>
I should have said anyone looking for a dual 1GHz should look for the old model rather than the new one. (Obviously if you're on the low end, then you should get the new 867 since it's a dual.) But why would someone want to get the old dual over the new dual? If their speeds are equivalent, then why wouldn't one want to buy the old dual? The new one is listed at macmall.com (a major reseller) for $2494, while the old one is listed for $2194. It looks like anyone who buys one of the new dual 1GHz will be cheated, as long as we assume Bare Feats is right.
[ 08-16-2002: Message edited by: Big Mac ]</p>
It seems like a lot of you are missing the point here. Bearfeats ran tests of the old "Fastest" configuration against the new "Faster" configuration. It makes sense to me that the Altivec Fractal results would be the same. It also makes sense to me that Apple would not include as much RAM in the new mid-range model as in the older high-end model. Wait a couple of months for the 1.25 dual to ship. The extra MB of L3 cache/processor as well as all of the new motherboard improvements should really shine through and give us some truly amazing results.
Wait for XLR8yourmac's review whenever that comes out. I consider that to be the most accurate place to obtain performance numbers.
-Dual 867
<strong> <img src="graemlins/oyvey.gif" border="0" alt="[No]" />
Some people still refuse to accept the truth: 7470 is just a product of MacOSRumors.</strong><hr></blockquote>
nonono it's a rumor from the register.
here: <a href="http://theregister.co.uk/content/archive/24018.html" target="_blank">http://theregister.co.uk/content/archive/24018.html</a>
And that bit right there indicates that Apple declined this chip, asked for overclocked G4s to "wrap things up" for the Fall, and is now dancing with a new partner for Q1 2003.
Screed ...CONFIRMED!!!
I'd still like to know who fed them that series of stories. I'm betting it was someone here.
<img src="graemlins/bugeye.gif" border="0" alt="[Skeptical]" />
The same guy used to rave about how fast Macs where, still under RDF, and when that failed all of a sudden PCs were always faster...
And no he even posts stuff liek new machines being slower than old ones. Even if it's possible (which it is, due to the smaller cache), who gives a damn?
Real benchmarking with comparable results is a very hard thing to do. I've done it myself and it takes several days to finish a series, not several minutes.
G-news