[bil207]

http://www.barefeats.com/pmddr.html

[andreas_g4]

As said on the page: depressing.

[The Placid Casual]

I guess this will answer a lot of peoples questions about the effectiveness of the Xserve style DDR in a desktop system...

In my mind this confirms this generation of Powermacs to be a a 'Yikes!' system, just waiting for a DDR compatible chip.

Also, with the new system controller (finally!), it would seem that it doesn't have to come from Motorola either...

I may now wait just that little bit longer for a new Powermac...

Peace,

Marc

[raferx]

Well I feel a whole lot better! I thought I was going to have to really think about upgrading my DP 1GHz to the new model. Not now though. I'll wait until IBM comes onboard, and Apple has a chip that can take advantage of DDR for real, AND increases it's bandwidth. I concur with previous posters here, and in other threads... I smell Yikes! all over again. I think I'll keep my box until a G5(?) and Rapid I/O, FW2, real DDR implementation etc. comes to the desktop line before I upgrade again. My DP will always be a great render box.

[Phrogman]

Sh*t!!!.....I just bought the new dual 1 ghz. I could have saved a bundle by buying one of the old ones and had a faster(slightly) computer to boot!.....Someone, Please say something to cheer me up!
:-)

[dru]

Originally posted by bil207:
http://www.barefeats.com/pmddr.html

Is Jaguar the difference?

Also, there's no specific information about each system used in the test (rather than the generic specs listed); that detail would be helpful.

Until these "tests" can be independently verified...

[solitere]

Originally posted by Marc2211:


I may now wait just that little bit longer for a new Powermac...

I did buy a Dual 800 a year ago and I am still very pleased with this system. And I have had a lots of different workstations during my years as a macuser.

If you get a dual cpu system with jaguar you will get pleased with it�s performance at least I did and I�am pretty greedy when it comes to performance.

[hmurchison2001]

Apple YOU SUCK!

[RyanG3]

Please clarify

Is it just the current G4s that are the bottleneck?

Is it because they only support 133MHz fsb at 1 or 1.3 gig per second and they (current G4s) don't support DDR ram at 2.7 gigs per second??

If that's the case then the motherboard is ready for an updated processor and it's the current G4 chip that is the bottleneck, not anything on the new motherboard...right?

Thanks for your reply.

[DarkMo0n]

Originally posted by hmurchison2001:
Apple YOU SUCK!

Oh werry mature. Lets just see some more independant test before we cry wolf.

/thomas

[bigv]

As I've said earlier the only reason I bought the new DP867 is I'm replacing an aging G3, otherwise I would not have. I am not ever going to buy another PowerMac that has a motorola processor of any kind. Apple will have to drop motorola and soon or it will be strangled by it. With Apple's attempt to improve it's acceptance in the corporate market it cannot afford to stick with motorola and expect any progress.

Right now I am furious with motorola. DDR has been around more than two years now and STILL we can't use it. To be free of motorola once and for all is a fond dream, one I fervently wish will soon come true.

[Big Mac]

Now first off many don't hold Bare Feats in very high repute, so no one should do anything drastic over this information. However, I do believe the chances of these benchmarks being true are good. What's particularly troubling is that even the extra thirty-four MHz of the new 1GHz didn't help it in the tests. DDR won't help all that much, but the thirty-four MHz definitely should have. We definitely need more testing (Macworld) before we can decide what's really going on.

[OSX Abuser]

I'm very happy with my Dual800, but I will probably wait till I can buy double the clockspeed.
I really want to see 10.2 on my box though.
I agree Apple needs to move industry standard components.
Full DDR etc.

Originally posted by solitere:


I did buy a Dual 800 a year ago and I am still very pleased with this system. And I have had a lots of different workstations during my years as a macuser.

If you get a dual cpu system with jaguar you will get pleased with it�s performance at least I did and I�am pretty greedy when it comes to performance.

[bigv]

One interesting thing that makes it even worse. With the mobo designed with separate access to the memory controller I was under the impression that each processor had separate access as well, but they share access. If the processors did have separate access that would have mitigated to some degree the lack of support for DDR. I can only assume the shared access is yet another limitation of the G4 processor itself. I see no other reason why the bandwidth is shared between the two when Apple made such a big deal about not sharing memory bandwidth with the other components of the system.

Just a thought.

[99switcher]

Originally posted by RyanG3:
Please clarify

Is it just the current G4s that are the bottleneck?

Is it because they only support 133MHz fsb at 1 or 1.3 gig per second and they (current G4s) don't support DDR ram at 2.7 gigs per second??

If that's the case then the motherboard is ready for an updated processor and it's the current G4 chip that is the bottleneck, not anything on the new motherboard...right?

Thanks for your reply.

Even if the MPS7450s were the bottleneck, are we to pay extra for a CPU upgrade several months (year(s)) down the road while Steve & Co. pocket the profits from a new model that may have an improved motherboard design but is pushed out the door carrying an obsolete microprocessor? I doubt if Apple will come along in 6 months and offer a free (or "shipping and handling only") upgrade for those who purchased a "Dual Optical" machine in August, '02.

[Rickag]

Come on people. They're both 1 GHz dual machines. What did you expect? And now the dual 1 GHz machines are HOW MUCH LESS EXPENSIVE than a week ago???

All this test means is that under the test parameters used the bus and potentially the size of the L3 cache were not the limiting factors.

Some one with more technical knowledge can explain it much better, but really, think before you condemn the machines.



Just thought I'd add that everything I've read concerning the MPX bus indicates it to be extremely efficient, why would it suddenly become inefficient @ 167MHz.

[The Placid Casual]

Originally posted by solitere:


I did buy a Dual 800 a year ago and I am still very pleased with this system. And I have had a lots of different workstations during my years as a macuser.

If you get a dual cpu system with jaguar you will get pleased with it�s performance at least I did and I�am pretty greedy when it comes to performance.

I bought an EOL Dual 800 for a bargain price in February

As you say, it was great with OS X, and was a really sweet machine (I'm a graphic designer so ran alot of Adobe stuff!). I was totally impressed.

Except for one thing. It sounded like a low flying Apache attack helicoper when all the fans were going. It was my home machine and I wanted to run it 24/7 (team Macnn at night!)... At night I could hear the thing 2 rooms away, through closed doors!

It had to go I really miss it

I downgraded to a cube with Radeon card, and a stack of RAM... It is a great machine, but I do struggle now and again when I'm working from home, and I need to run more than one app at a time (Photoshop, Illustrator, indesign).

I guess we will know more about the IBM chip in October..I am going to weld shut my wallet and wait until then

Cheers,

Marc

[iamnid]

Raferx... you mentioned some dismay because you had already ordered a new dual gig -- this may cheer you up... at least you can install two optical drives easily... and there's a front speaker jack... maybe the new fan is quieter?

[The Placid Casual]

Originally posted by RyanG3:
Please clarify

Is it just the current G4s that are the bottleneck?

Is it because they only support 133MHz fsb at 1 or 1.3 gig per second and they (current G4s) don't support DDR ram at 2.7 gigs per second??

If that's the case then the motherboard is ready for an updated processor and it's the current G4 chip that is the bottleneck, not anything on the new motherboard...right?

Thanks for your reply.

It is indeed the G4 that is the bottleneck. The new (fabled!) system controller has been implemented so that in theory the board is completely DDR ready...It is just the current G4 that cannot.

This 'flaw' in the current G4 means that as there is no double data rate, there will be NO system speed up, it will perform exactly the same as an SDRAM machine.

The mother board is fine, it is a sweet design. If it had a DDR capable chip in there it would be awesome. It is like the board is designed for a more elaborate CPU...

The similarities to the Yikes! machines are striking...

It is Motorola that has shafted Apple again...

However, the beauty of the new system controller is that it can be changed very easily to take a chip other than the G4...changes are afoot...

Peace,

Marc

[The Placid Casual]

Originally posted by Big Mac:
Now first off many don't hold Bare Feats in very high repute, so no one should do anything drastic over this information. However, I do believe the chances of these benchmarks being true are good. What's particularly troubling is that even the extra thirty-four MHz of the new 1GHz didn't help it in the tests. DDR won't help all that much, but the thirty-four MHz definitely should have. We definitely need more testing (Macworld) before we can decide what's really going on.

The barefeats figures are usually quite accurate...its the inference drawn from them, and the comments they make that I have a problem with

Marc

[Morpheus]

Originally posted by Marc2211:


I downgraded to a cube
Marc

I would say Apple has yet to deliver a computer from which you can "downgrade" to a cube! I'm sure you meant "upgrade".

[The Placid Casual]

Originally posted by Morpheus:


I would say Apple has yet to deliver a computer from which you can "downgrade" to a cube! I'm sure you meant "upgrade".

Indeed! It is the best Mac I have ever owned...I'll be sorry to see it go

Marc

[echoes]

Here's another negative writeup on the new systems. The author suggests that Moto had no faster CPUs for Apple to release (and that there's no info about the bumped chips on Moto's site) so Apple took the old 1MB L3 G4 800 and overclocked it's FSB to 167MHz to make the new dual 1GHz and took the old 2MB L3 G4 1GHz and did the same to create the dual Dual 1.25GHz.

I'm not sure what to make of it...

[normyzo]

Could the Barefeats people actually run some _memory intensive_ tests to see what that would do? I mean, none of those tests were anything for the Xserve. You don't run an Xserve as a normal machine, you run it as a web/database/scientific server, and applications such as those would benefit much more from the (admitedly half-assed) DDR than a simple iTunes rip... These test only show one side of the issue, and don't show where the new machines actually _would_ be better.

Dan

[echoes]

Bryce rendering should be memory intensive.

[99switcher]

Originally posted by echoes:
Here's another negative writeup on the new systems. The author suggests that Moto had no faster CPUs for Apple to release (and that there's no info about the bumped chips on Moto's site) so Apple took the old 1MB L3 G4 800 and overclocked it's FSB to 167MHz to make the new dual 1GHz and took the old 2MB L3 G4 1GHz and did the same to create the dual Dual 1.25GHz.

I'm not sure what to make of it...

I do. And it does get increasingly ugly as I read some of the more enlightened posts. It follows an old engineering adage: "If you can't fix it, feature it". Something that the Sales/Marketing weenies do routinely when the engineering types like me throw up our hands in despair.

533 MHz (OC'd to 600 MHz) "Digital Audio", OS X.1.5

[OreoCookie]

Most of the Photoshop stuff fits right into the L3 cache used in the PowerMacs. Some memory intensive apps would surely clarify that (e. g. the swim benchmark of the SPEC suite).

Otherwise, the test revealed what we already knew all along: same CPU, same MHz, same speed.

No doubt the 7450 is the bottleneck, but the system could (and should) perform differently when memory intensive stuff is concerned.

[RGB]

But isn't this new G4 situation the exact opposite of the Yikes! situation? With the Yikes! it was a Yosemite mobo with the new G4 chip. Here, we have a new mobo with the old G4 chip. Am I mistaken?

[GnOm]

Originally posted by Phrogman:
Sh*t!!!.....I just bought the new dual 1 ghz. I could have saved a bundle by buying one of the old ones and had a faster(slightly) computer to boot!.....Someone, Please say something to cheer me up!
:-)

the new one is much cheaper than the old one, that should make you feel better now.




bye.

[OreoCookie]

Yup, you are right, with Yikes, it was the other way around.

[davecom]

You're all thinking too much. The barefeats tests are probably right, but there's a very good reason why. The new Dual Gigs only have HALF of the L3 cache that the old ones did. 1MB versus the old 2MB. Perhaps the lower cache negates whatever speed difference would be expected.

[::maroma::]

I have a coworker who is in the market for a new PM. She is going to be getting into some pretty heavy video editing (FCP/DVD Pro). She was initially going to get the original Dual 1GHz box, but I recommended she wait until these new ones were released. Now she debating between the new Dual 1GHz, the old 1GHz and a New 1.25GHz.

So, my dilemma is should I recommend the new 1GHz, the new 1.25GHz, or the old 1GHz. Or should I tell her to wait until more official, more thorough tests are released? Or are these tests pretty indicative of what we'll see on further testing?

Signed,
Confused Consultant

[markphip]

Here is what I do not understand, perhaps someone can answer some of this for me?

These tests show that the old and new Dual-Gigs have roughly the same performance on these benchmarks. The conclusion that is drawn from that is that Apple implemented a half-baked system that doesn't reap any of the benefits of DDR.

So here are my questions?

1) Assuming Apple had a perfect DDR implementation, what would the expected results from these benchmarks have been? How much faster should the DDR systems be on these tests? Do we know that these benchmarks should reveal such differences?

2) Have their ever been any comparable benchmarks done in the PC world between two systems that are largely the same other than DDR vs SDRAM? What kind of performance gain does DDR yield?

3) Is it possible that differences in the P-IV/Athlon architectures and the PowerPC architecture cause the former to be more "RAM-dependent"? In other words, if the PowerPC is more efficient in managing its pipeline and more-intelligent in its use of cache, then is it possible that PowerPC platforms get less of a boost from RAM improvements?

Thanks

Mark

[markphip]

Originally posted by ::maroma:::
I have a coworker who is in the market for a new PM. She is going to be getting into some pretty heavy video editing (FCP/DVD Pro). She was initially going to get the original Dual 1GHz box, but I recommended she wait until these new ones were released. Now she debating between the new Dual 1GHz, the old 1GHz and a New 1.25GHz.

So, my dilemma is should I recommend the new 1GHz, the new 1.25GHz, or the old 1GHz. Or should I tell her to wait until more official, more thorough tests are released? Or are these tests pretty indicative of what we'll see on further testing?

Signed,
Confused Consultant

The price difference between the new and old Dual-Gigs is not that great, and the new one offers some extra's that have nothing to do with DDR, such as room for 2 optical drives, 4 hard drives (2 with ATA/100), an audio in port and a front-side headphone jack.

To me, those may be reasons enough to purchase the new box over the old. Also, if she was prepared to buy the dual-gig before the announcements, then why isn't she also looking at the dual-1.25 which presumably will offer some real performance gains on video apps. After all, it is an extra 25% of CPU speed.

Mark

[::maroma::]

Originally posted by markphip:


The price difference between the new and old Dual-Gigs is not that great, and the new one offers some extra's that have nothing to do with DDR, such as room for 2 optical drives, 4 hard drives (2 with ATA/100), an audio in port and a front-side headphone jack.

To me, those may be reasons enough to purchase the new box over the old. Also, if she was prepared to buy the dual-gig before the announcements, then why isn't she also looking at the dual-1.25 which presumably will offer some real performance gains on video apps. After all, it is an extra 25% of CPU speed.

Mark

Good points. She is indeed considering the 1.25GHz model, but she isn't sure if the extra cash is worth the extra Hz. I told her my philosophy about buying a new computer, which is to decide on the most money you are willing to spend on a computer, and buy the best one you can get for that amount of money.

She's trying to save as much as she can, but wants to get the most for her money (and no, she isn't willing to get a Wintel )

Before seeing these preliminary test results, I was steering her away from the older models, and towards the new ones. She told me that The Computer Store (an Apple reseller out here) has the old 1GHz boxes on sale, which is why she is considering the older models.

Anyway, I'll fill her in on the pluses you mentioned. I think she wants to see some numbers though.

[supernature]

Well, i guess this is why Apple went with the large heat sinks for the new case. I guess they really had to cool that sucker down to get it running reliably at 1.25ghz.

That's pretty sad. I am looking forward to Apple switching over to IBM's new chip later this year hopefully. I read somewhere that the vector set is similar to Motos. I wonder if Apple insisted that Moto license that technology out to IBM.

All I know is that IBM's most current PowerPC processor can run up to 1Ghz with 512K backside cache, and bus speeds of up to 200mhz. I'm sure Apple could stick that sucker right on to their iBooks, but that wouldn't be prudent for them to do.

So if anything, at least their iBooks future are well laid out for the near future. I'm hoping to see the Power4 chips at MWSF.

Most undoubtedly, Moto isn't going any further in development with their PowerPC chips.

[The Placid Casual]

Originally posted by RGB:
But isn't this new G4 situation the exact opposite of the Yikes! situation? With the Yikes! it was a Yosemite mobo with the new G4 chip. Here, we have a new mobo with the old G4 chip. Am I mistaken?

Although the actual technical situation is reversed, the parallel lies in the fact that they are both 'stop gap' machines.

They are a 'mix and match' of old and new, only used together until all the pieces of the larger puzzle fall into place at a later date...

Marc

[rhogue islander]

Originally posted by ::maroma:::


So, my dilemma is should I recommend the new 1GHz, the new 1.25GHz, or the old 1GHz. Or should I tell her to wait until more official, more thorough tests are released? Or are these tests pretty indicative of what we'll see on further testing?

Signed,
Confused Consultant

I don't know how much of a hurry she is in, but I might wait until more test results are available.

It would seem as if DDR sdram does little to boost performance. The fact that the new dual 1gig has half the L3 cache of the old DP1000, however, may have a considerable impact on performance in favor of the old machine.

Unless she's willing to spring for the 1.25 dual I might, in her place, go for the old DP1000 for $2200 and spend the difference on hard disks

NOTE: I am the owner of the old DP1000 and hence am biased in its favor.

[HitByASqurrel]

i dont know if any one has said this but the new dual 1Ghz does have something that makes its power so much more than that of the previous one....


ITS CHEAPER!!!

[NeoMac]

Making her wait is silly. The 'old' DP1000 is the best deal. Have her buy one while they are still in stock. She'll be happy with it and that's that.

[suhail]

For those who still do not believe that the new Duals are really not fast or doubt the bearfeats tests, here is something from the horse's mouth:


Found at Apple's own website

Yes! the NEW Dual 1.25GHz is only three times the speed of an OLD single 500MHz.

Yes again! the NEW Dual 867MHz is only twice as fast as the OLD single 500MHz.

And according to Apple, FCP was "the perfect application to measure dual processor performance�"

[tmornini]

Originally posted by Marc2211:

This 'flaw' in the current G4 means that as there is no double data rate, there will be NO system speed up, it will perform exactly the same as an SDRAM machine.

This is NOT true.

The issue with this test is this:

The old dual GHz had 2MB L3 cache, the new one has 1MB.

The old dual GHz had 133MHz bus, the new one has 167.

What's happening is this: This test must be borderline for the 1MB cache. The new system is doing more memory access, slowing it down, even though the memory access IS FASTER than it used to be.

The new system would perform significantly better if some other high-bandwidth tasks were going on, like copying from internal IDE drive to external Firewire. Add in some heavy network I/O and the new system will just pull away from the old one. Add in some heavy video processing with Quartz Extreme, and again, the new system will just get faster and faster than the old one.

Bottom line for me is this: New 1GHz system is $700 cheaper than old one. Even at the "same speed" it would be a good deal.

[tmornini]

Originally posted by suhail:
Yes! the NEW Dual 1.25GHz is only three times the speed of an OLD single 500MHz.

3 times. That's HUGE!

Yes again! the NEW Dual 867MHz is only twice as fast as the OLD single 500MHz.

Twice as fast, 1/2 the cost! That's 4x price/performance!

All you've shown here is that the numbers don't mean Jack!

[suhail]

Gee. Didn't mean to make anyone mad, but the new Apple specs are not performing.

If the new stuff were working as they should, the Dual 1.25 should be AT LEAST 4x faster than the older 500MHz. Especially under an application that is fully optimised for a Dual.

500MHz is slow!

[benh57]

As others have mentioned, these tests were all CPU bound. That's not going to be faster. Duh. Same processor, same Mhz, same bus (from controller to CPU).


You will see improvements on stuff that is disk to RAM intensive but not CPU intensive, like serving web pages or serving warez hotline servers . Server tasks. People WILL use these as servers.

I would like to see some server oriented benchmarks from the guys who did the XServe tests. Those tests showed the XServe kicks the old dual g4's ass. And the DP 1Ghz is very similar to the XServe.

[neilw]

It is sufficiently seldom that I make a good prediction that I feel the need to point out when it happens. Feel free to ignore, flame, whatever. This talk of Yikes reminded me of my old post, written on 1/29/02 in this thread:

I'm thinking about a kind of upside-down "Yikes" design for MWNY. This time, the new motherboard is ready but the new processor isn't, so they'll put the "old" processor (whatever the Apollo is up to by then) on the new motherboard (DDR, etc.)

As for the BareFeats test results, they are completely unsurprising, but that doesn't make them uninteresting. The trick will be to come up with tests that show the current architecture is beneficial. Hopefully it would not only be server applications, which are not of as much interest to most users.

What these tests mean is not that the new machines suck or anything, just that for most applications (until proven otherwise) they will mainly represent merely a minor speed bump (and a major cost reduction, as many have pointed out.) And I don't blame Apple for that, as they are completely at the mercy of Motorola for now. Of course, you can blame Apple for marketing the new architecture like it's the second coming, but what's the point.

These machines certainly lay the groundwork for a 7470-based speed bump in January, leaving us to hope for a G5 or IBM equivalent at next year's MWNY.

[lngtones]

Originally posted by neilw:
It is sufficiently seldom that I make a good prediction that I feel the need to point out when it happens. Feel free to ignore, flame, whatever. This talk of Yikes reminded me of my old post, written on 1/29/02 in this thread:



As for the BareFeats test results, they are completely unsurprising, but that doesn't make them uninteresting. The trick will be to come up with tests that show the current architecture is beneficial. Hopefully it would not only be server applications, which are not of as much interest to most users.

What these tests mean is not that the new machines suck or anything, just that for most applications (until proven otherwise) they will mainly represent merely a minor speed bump (and a major cost reduction, as many have pointed out.) And I don't blame Apple for that, as they are completely at the mercy of Motorola for now. Of course, you can blame Apple for marketing the new architecture like it's the second coming, but what's the point.

These machines certainly lay the groundwork for a 7470-based speed bump in January, leaving us to hope for a G5 or IBM equivalent at next year's MWNY.

How has Apple marketed the architecture like the second coming? They just released them without a big trade show or anything. The Apple stores around here still don't have them up as display models, only to buy for the next few days.

I think they are meant to be in-between revisions to calm down nay-sayers, but not attract TOO much attention until the big thing is ready.

There are no TV ads, no billboard ads, just some pictures and stuff on their web site.

[milhous]

Three letters:

I B M

[Superchicken]

I just feel like pointing out... uhh... NEW MOBO!
How come no one's mentioned this?
Once upgrades for these new IBM proccessors come out, won't it be easier to fit one in one of these than in a Quick Silver?

[dab007]

I could not make up my mind on the old or new 1ghz until! I saw this Great deal So, had to get it.....