We reported on this a few days ago, but now the article is updated 8/17/02 with ‘streamlined’ graphs, tweaked verbiage, and addition of a very interesting Memory Speed Graph. Page Two added on 8/20/02 with additional tests. Updated 8/21/02 with another bottleneck theory and Memory Speed Graph interpretation. Read the updated benchmarks here and here. What will it take to fix the bottleneck, the conclusion asks? The author replies: A CPU than can handle the full speed of DDR memory (like the fabled PPC 7470) and a redesigned motherboard with separate bus for each CPU.
So are these systems cheaper or the same cost as the previous systems? Otherwise the buyers are truly getting screwed. Can Apple afford the R&D required to either “fatten the pipe” or go “point to point” to create a truly useful SMP system?
The new dual 1ghz they tested is about $700 less than the previous dual 1 ghz.
The dual 867’s are 1700$, the price of a single 867 from last week.
Well I agree that the prices are cheaper, still no reason putting DDR if you are not going to take advantage of it. Let us see the fabeled CPU or the IBM chip. Ya I know, ever since the Power4 decendent announced I have been pushing for it. Sorry. So we should at least seperate pipers for each CPU and the 7470.
Hey how did I miss the first post ๐
:>)
How would they handle cache invalidation and atomic operations if the CPUs have separate busses and can’t snoop each others memory accesses?
It’s probably possible somehow, but I really can’t see a cost-effective way to do it. Just making the bus faster and adding more cache (to reduce memory bus traffic) should be enough, and would make a whole lot more sense.
although the 1 second slower benchmarks in almost everything arent great, but its ALOT cheaper then the previous powermacs, and all powermacs are dual processor now.
The new Powermacs are alot better then the old, but are slightly slower (1-2 seconds? who cares when it is cheaper?).
Stop that stupid and unrelevant testing!
I think it is strange to compare the last high-end model with the current mid-range model.
Hey benchmark freaks out there – just wait one month and compare the current hight-end with the old hight-end model. In the meantime compare the old midrange-model with the new midrange model and stop confusing the people.
Ralf.
apple didn’t add ddr memory to make the machines faster
apple added ddr to make the machines cheaper
with ddr memory production far ahead of sdr, it saves major $$$
multiply the ram cost by the number of machines
and it easily pays for apple’s hacked memory asic
remember, apple takes pc hardware and makes it slow
apple maintains margin
mac buyer gets swindled again
do you think ellen feiss knows the difference?
uhhh…. like my memory speed was half gone….
?
It strikes me that this guy/girl isn’t really all that scientific about this benchmark. The two systems are miles apart, so you can’t promise the OS loads are the same (at least, they don’t mention “the oS was loaded from scratch”). The machines have different video cards, and they have different amounts of RAM (old==1Gb, new==768Mb). Call me crazy, but with OSX’s memory footprint, plus they’re probably using big files, they may be getting close to the physical ram. Even if not, it reduces the sapce left over for file cacheing.
What does this benchmark-test tell us?
Well, ist tells us that the new mid-range Power Mac is as fast as the old high-end Power Mac. Here in europe the new dual 1gig Power Mac is app. 1000EUR cheaper than the old dual 1gig Power Mac.
Alle I can say is:
Well done Apple! This is a true performace boost through the whole Power Mac line.
I will get one of these machines this year.
Ralf.
How would they handle cache invalidation and atomic operations if the CPUs have separate busses and can’t snoop each others memory accesses?
They send messages to each other via a dedicated bus, this is what things like Hypertransport is for, Motorola have their own take on the idea called RapidIO.
Oddly enough Apple is a member of the HyperTransport consortium so a shared bus wont be around forever.
Then is was not L3 cache size.
Seems that just like I said, knowing an OS developer(s) does not make you an OS expert just yet.
That is what HE says. If Apple had the 2 MB of L3 cache still there, the new model would have been better than the older model. Of course, having a CPU that supports DDR and a bus that does the job right, of course and it will help. But half-ing the L3 cache was butchering compared to the older model.
Now, why don’t you go and shove it before starting critisizing someone you don’t know?
Well finally you tell these people that try and get personal where to put it ๐
can’t we all ust get along
Anyway. I undertand that we have not tested the new 1.25, i am wondering if Apple is going to try and pull a fast one and sneak a CPU that can handle better bandwidth…..
I told you not to critisize people you don’t know.
Either way, is that “that of the month again?”
If answer is yes, I won’t even take it personally.
I remember a while back when I went to Apple’s online store to config a dual 1 GHz PowerMac, I asked for a CDRW instead of a DVDR, and a less expensive ATI card instead of the GF4. The total came to about $2299. A price I might have paid a few weeks ago. Now I just did the same thing but with the new G4 dual 1 GHz, and the exact total is $2199.
If the performance difference isn’t that big, I wouldn’t cry about this, otherwise I’m questioning how much performance would I get for that extra $100 for the older model, or in other words, by saving $100 how much performance am I sacrificing?
This is hard to believe. The old 1GHz model was the “fastest” model in the Apple store. It was and is in the 3200$ price region. I can not believe that you get 1000$ off only with a reduced DVD-R to CD-RW and a reduced graphics card. Maybe you confise something. Is it possible that you reduced the old “faster” model (the 933 singe G4) by DVD-R and graphice card?
Ralf.