Apple introduces new Mac Pro topping out at 3.2GHz - Page 6

Originally Posted by Mr. H

I think you mean "kudos"

back on topic:

I was a bit surprised to still see FB-DIMMS. I thought Intel was go to phase out the use of FB because of the extra latency, power consumption and cost?

Originally Posted by Bergermeister

I think you mean "was going to"...

Originally Posted by Bergermeister

Still waiting until next week to decide whether I need to start thinking about how to pay for a Pro...

Originally Posted by melgross

How many pro's are REALLY going to use them? Really?

That's why there are TWO 1 GB Ethernet connections on the back.

Originally Posted by melgross

8 core vs 4 cores for the mid line models.

Faster bus, faster memory, faster cpu's etc.

It's all theoretical.

We'll find out when they land in the sites hands and we see tests.

Originally Posted by emig647

If it keeps the cost down I'm glad the 2600xt is the default card. Not everyone uses a 30". I'm on a 24" and I think the 2600xt would be fine for most things besides any semi-serious gaming. They could have done worse... Like they did with the last rev... 7300gt as default card? LOL ... they could have put a x2300pro or 8300gt as default card.

Originally Posted by melgross

When Nehalem comes out, supposedly in the 3rd quarter of this year, Apple will be required to make a massive makeover of the internals of the machine, at least.

At that time Express 2 should be completely supported, and I would expect a new case to mark that major shift. This case will be 4 years old, and that's pretty old.

Nehalem, its supporting chipset, and full support of Express 2, will also allow for more lanes, giving more options in video, among others.

I would expect eSATA by then as well.

Originally Posted by Messiah

Yeah I know, they went from a four-core 2.66GHz, to an eight-core 2.8GHz, and they only got a 1.3x increase in Photoshop?

That's a bit of a let down from '2x faster'.

Originally Posted by JeffDM

Photoshop might not even be optimized for more processors yet.

Originally Posted by taoshen1983

The standard configuration uses the Radeon HD2600XT? That's such a horrible mismatch for Intel's mighty 8 cores @ 2.8Ghz @ 1600FSB!!!!!....You will be crawling at the 30 inch resolution for anything 3d. I don't know if iTunes can use the visualization at that resolution.

I am surprised that the Geforce 8800GT 512MB wasn't the default standard configuration. On the other hand, the 200 dollar upgrade for 8800GT isn't that bad. It costs $275 retail to buy a 8800GT. Considering the retarded 2600XT costs 80 bux, it's almost trivial to upgrade to the 8800GT.

Originally Posted by philipm

Having worked on efficient coding of multiprocessor software for my PhD about a decade back (look for it at http://www.itee.uq.edu.au/~philip/Publications/), I wouldn't be too surprised if the problem is that too few people understand the performance issues. If you don't have a good grounding in computer architecture with some understanding of OS and other hardware-software interactions, it's hard to get decent speedups. This is especially true as the speed gap between DRAM and processors grows. There's been a bit of a stall in clock speed increases the last few years, but DRAM hasn't been improving that dramatically either -- transfer rates have been going up but total time to start a new random access is still pretty slow compared with CPU cycle times. To look at some numbers, if you get a DDR3 SDRAM from Micron with a 1.5ns cycle time, it has an overhead of at least 24ns before any data starts to move. If your shiny new 8 core 3.2GHz machine is only trying to deliver a conservative 1 instruction per clock per core, it can execute over 600 instructions in the time it takes this kind of DRAM to heave into life. The DDR2 stuff the Mac uses is not far off this sort of speed.

But anyway, the point is that any code that uses all the cores has to avoid touching DRAM as much as possible, otherwise memory accesses become a serious bottleneck.

Originally Posted by Mr. H

I think you mean "kudos"

back on topic:

I was a bit surprised to still see FB-DIMMS. I thought Intel was going to phase out the use of FB because of the extra latency, power consumption and cost?

Originally Posted by taoshen1983

The standard configuration uses the Radeon HD2600XT? That's such a horrible mismatch for Intel's mighty 8 cores @ 2.8Ghz @ 1600FSB!!!!!....You will be crawling at the 30 inch resolution for anything 3d. I don't know if iTunes can use the visualization at that resolution.

Originally Posted by rosstheboss

um...well, i am a pro, as are the six people i work with, i sit here surrounded by G5 and Intels, DVCams, video capture cards and RAIDs. We use wireless mice and keyboards, we need to update work schedules to our mobile phones, use Wi-fi for remote (as in a Cat 5 wouldn't be convenient) control of audio and video equipment

Also, if you're going to supply your lowest spec machines WITH these 'basics' yet not your £3000 flagship, it SEEMS a little shite....

So, yeah, as a pro i do REALLY use them. Really.

i don't give a shit about the numbers and who's got what PCI architecture and a SCSI jock strap, i just want my video to render quickly so i can go out with some girls...

Originally Posted by melgross

Most pro's don't need a good 3D board, why charge for it? It's perfectly reasonable to expect those who do need it to upgrade.

Let's understand something. You don't get something for free. If Apple made the 8800 the standard, they would have raised the price on the computer by another $100 . The same is true for WiFi.

Then people would complain that the price was even higher.

No, it's better to have the upgrade.

I'm disappointed that the card wasn't the 8800GTX, or some higher performance version. The GT is not a great card, considering that the next one up the line costs $2,600 more.

Originally Posted by Messiah

Yeah I know, they went from a four-core 2.66GHz, to an eight-core 2.8GHz, and they only got a 1.3x increase in Photoshop?

That's a bit of a let down from '2x faster'.

Originally Posted by winterspan

Why are you so bothered (assuming a few comments up was you too.. forgot to look)
about having the ports of the Mac Pro be PCIe 2.0 ??? Is there even a video card on the market that can saturate a PCIe 1.0 x16 slot?? For that matter, is there *ANY CARD* available for the Mac Pro that would get ANY benefit from a PCIe 2.0 slot? I mean 4 gigabytes a second seems to be enough for anything that we have now. Am I missing something?

Originally Posted by Royboy

My son is excited about the WiFi option on the new Mac Pros. He got a Ninetendo Wii for Christmas and he says that if I order the Wifi option on the Mac Pro then he can use the Wifi option on his Ninetendo Wii to connect to the Internet via my Mac Pro. It seems to be reasonable, but I can't tell him for sure. Anyone got Wii and knows?

I currently have his Ninetendo Wii hooked up to the Internet with a router (hard wire) through my cable modem.

Originally Posted by Royboy

My son is excited about the WiFi option on the new Mac Pros. He got a Ninetendo Wii for Christmas and he says that if I order the Wifi option on the Mac Pro then he can use the Wifi option on his Ninetendo Wii to connect to the Internet via my Mac Pro. It seems to be reasonable, but I can't tell him for sure. Anyone got Wii and knows?

Originally Posted by melgross

That's not the fault of the machine. It's the fault of Photoshop. PS doesn't use more than 2 cores yet. Adobe has told me that it's a priority for the next release, and possibly, they might be able to release an update that would help somewhat before then.

Of course, Apple's own tests are strange. They rate machines against the Quad G5, which according to them was still shipping as of December 2007!

Look to the botton of the page, just below the Mathamatica results in grey type.:

http://www.apple.com/macpro/performance.html

But some programs will use all 8 cores, and that will make quite a difference.

Originally Posted by melgross

My daughter's Wii goes through our 1 GHz Ethernet network. Latwer this year, I might pick upo an Apple router, if I can use it with my Internet Gateway, which I got from Covad.

I would get the WiFi for this as well, since it's so cheap. Actually this could serve as a basestation, but it's not on all the time.

Originally Posted by admactanium

many people bought the quad g5 to tide them over through the intel transition, just like i did.

Originally Posted by admactanium

the part about listing the quad g5 for sale in dec2007 is odd, but i don't think it's an odd comparison. many people bought the quad g5 to tide them over through the intel transition, just like i did. i'm happy to see the comparisons. there was so much confusion when the mac pros came out because CS was running in rosetta. so the mac pros weren't faster. even when it went native, the photoshop tests were still close enough for most imaging professionals to just put off upgrading longer.

Originally Posted by shetline

Me too. And I decided to wait for the second generation of Mac Pros before buying one.

It'll be interesting to see what a Quad G5 will sell for on eBay, which is where mine is headed as soon as my new Mac Pro arrives.

Originally Posted by shetline

Supposing that the only difference between the 3.2 and 2.8 GHz chips is sheer clock speed -- no architectural differences, no difference in cache sizes -- the 3.2 is 14% faster than the 2.8 for any task that's purely bound to CPU speed (and the 3.0 is 7% faster, of course).

Anything you do that has elements of disk I/O speed, bus speed, RAM speed, etc. -- i.e. most things you probably do -- won't see all of that 14% increase. Only a solid set of benchmarks will tell you how much for what tasks.

I really, really do like to buy the top-of-the-line models Apple makes. I'm typing this right now on my once top-of-the-line Quad G5. But I have to admit my craving for speed is not at all justified by any real-life business justification, or even any hobbyist uses that push my computer very hard. I just like nice toys.

But even though I like to spoil myself, in this case I just couldn't see spending the extra $1600, or even $800 dollars. I settled for the basic 8x2.8, with an extra optical drive, the NVIDIA GeForce 8800 (which sadly bumps my shipping time from 3-5 days up to 3-5 weeks -- I keep thinking about canceling my order and changing that!), and a 500GB hard drive instead of 320GB. I'll add an additional 4GB of RAM myself at a much more reasonable price than Apple's price.

Originally Posted by admactanium

the part about listing the quad g5 for sale in dec2007 is odd, but i don't think it's an odd comparison. many people bought the quad g5 to tide them over through the intel transition, just like i did. i'm happy to see the comparisons. there was so much confusion when the mac pros came out because CS was running in rosetta. so the mac pros weren't faster. even when it went native, the photoshop tests were still close enough for most imaging professionals to just put off upgrading longer.

Originally Posted by Joe_the_dragon

just get a real WiFi router like this with N and 5 gig-e ports
1 for the internet side and the other 4 for your network. http://www.tigerdirect.com/applicati...&Sku=D700-5426

Originally Posted by shetline

Me too. And I decided to wait for the second generation of Mac Pros before buying one.

It'll be interesting to see what a Quad G5 will sell for on eBay, which is where mine is headed as soon as my new Mac Pro arrives.

Originally Posted by solipsism

I agree. We all know that a new Intel platform will be faster than an older Intel platform and there are plenty of PC sites that will compare the most current and 2nd most current chips long before Apple puts them in their machines. But only Apple and its staunch consumer base will want to compare performance between PPC and Intel platforms.

Originally Posted by labelexec

how come you just don't order it with the ATI graphics board and get it within several days and go out and buy a $285 8800GT retail and put it in. I'd pay an extra $85 to get it before 3-5 weeks.

Originally Posted by melgross

Yeah, but it's pretty old now. It would make more sense to compare against an Intel based machine.

Originally Posted by melgross

I don't mind the comparison. It just isn't helpful when deciding whether to move from an older Intel model.

I agree that this will be done in other places. But a lot of people will see what Apple has put up and think, "Whut?".

Originally Posted by melgross

8 core vs 4 cores for the mid line models.

Faster bus, faster memory, faster cpu's etc.

It's all theoretical.

We'll find out when they land in the sites hands and we see tests.

Originally Posted by melgross

That's one reason why these processors have so much cache.

Originally Posted by philipm

Yup. It's called Amdahl's Law. Study it before you believe "up to" speedup figures.

Originally Posted by philipm

That doesn't help you if you are doing interprocessor communication which in any form needs to go through memory. If your processes or threads very seldom communicate and don't have shared data structures, big caches will keep data traffic off lower levels of the hierarchy. Otherwise, you have a big scalability problem.

If you are just running a multitasking workload with lots of busy little processes that have nothing to do with each other, as long as Apple has fixed the OS bottlenecks, you should be fine.

BTW I find the terminology of a "processor" as meaning a collection of cores unnecessarily obfuscatory. Apple's figures say 12MB per processor but if you check the fine print, this means 3MB per core, with 6MB shared between 2 cores (still good but not as huge as 12MB sounds). A "processor" in every computer architecture text and research paper I've read is the same thing as a core. The only technical difference between a multicore design and a multiprocessor design is that you eliminate some off-chip latencies between cores as compared with the same design with the cores on separate chips.

In Apple's web description, the statement "6MB of cache is shared between pairs of processor cores, allowing an individual core to use all the available shared cache at any one time" is kind of interesting. This means that if both cores sharing the cache are active they can both use all of the cache at the same time. Interesting. Must be some new spin on quantum computing

I wonder if Apple's scheduler is clever enough to schedule a pair of related processes or the first 2 threads of one process on cores with a shared cache... and up to 4 related processes or threads on one chip?

Originally Posted by aegisdesign

Fridge? That looks more like a small apartment.

Originally Posted by labelexec

Is there a big performance difference between 2.8 vs 3.0 and between 3.0 vs. 3.2.

I know there should be a significant performance boost between 2.8 vs. 3.2 but is it really worth 1600? or is it really worth an extra $800 for a 200Mhz bump?

I don't do anything too hardcore graphics or video but I just want the biggest and best. I have two other PCs with quad cores and 8GB of ram and barely use its potential.

Originally Posted by philipm

That doesn't help you if you are doing interprocessor communication which in any form needs to go through memory. If your processes or threads very seldom communicate and don't have shared data structures, big caches will keep data traffic off lower levels of the hierarchy. Otherwise, you have a big scalability problem.

If you are just running a multitasking workload with lots of busy little processes that have nothing to do with each other, as long as Apple has fixed the OS bottlenecks, you should be fine.

BTW I find the terminology of a "processor" as meaning a collection of cores unnecessarily obfuscatory. Apple's figures say 12MB per processor but if you check the fine print, this means 3MB per core, with 6MB shared between 2 cores (still good but not as huge as 12MB sounds). A "processor" in every computer architecture text and research paper I've read is the same thing as a core. The only technical difference between a multicore design and a multiprocessor design is that you eliminate some off-chip latencies between cores as compared with the same design with the cores on separate chips.

In Apple's web description, the statement "6MB of cache is shared between pairs of processor cores, allowing an individual core to use all the available shared cache at any one time" is kind of interesting. This means that if both cores sharing the cache are active they can both use all of the cache at the same time. Interesting. Must be some new spin on quantum computing

I wonder if Apple's scheduler is clever enough to schedule a pair of related processes or the first 2 threads of one process on cores with a shared cache... and up to 4 related processes or threads on one chip?

Originally Posted by acr4

You won't see much of a difference under normal applications. Here's a few points to mention:

- CPU core clock (thus FSB speed to the North Bridge and DDR speed) are the same in all the systems. This means that, regardless of CPU multiplier, your data transfer rate outside the CPU is fixed.
- All three CPU steppings have the same cache size. Going to a faster CPU isn't gaining any internal storage space (which *would* speed up many applications).
- At relatively small CPU speed deltas, it is my opinion that RAM size/speed plays a larger role in perceived speed/response than the CPU speed. Money would be better spent upgrading to at least 4GB of RAM. HD swapping is a UI response time's worst enemy. Eliminate swap and you're computer will *feel* faster.
- Again with the RAM: If you can offload as much video computation to the video card, you're system will have more time to crunch other numbers - get a bigger video card if you like high-demand video applications and you're performance will go up.

Of course, if you just like having the biggest and best, then go for the 3.2GHz, but I think the better option is the spend half the money and bump the RAM.

Originally Posted by newnew

What do you mean when you say:

"HD swapping is a UI response time's worst enemy. Eliminate swap and you're computer will *feel* faster."

Appreciate if you could elaborate