Why is Apple only giving a 5600 FX card with a $3000 system?

[drag]

Whose "spec"? Do you have a link to his site.

If you are refering to the veritest's test. They aren't running optimized or unoptimized binaries in any of them... except in the cast of the actual gcc compiler. And that needs to be optimized to begin with to even work with either the G5 or the P4 or any platform... They were compiling from code. Part of the disput is how much optimization is being used while compiling code. Generally less "optimization" flags means that the code will compile faster but produce larger and slower running binaries. To much optimization you start running into problems and produce buggy and unstable binaries, or in other terms exacutables.

Otherwise I am not sure what you are talking about.

 

[drag]

Also now that I think about it, those quake3 benchmarks are kinda unfair on the G5...

Because quake3 is a 32bit program, so it can't take advantage of the 64bit stuff and other improvements and unless that Id gave code to the Mac developers so that they could compile their own code and optimize it for the G5 it would be running in 32bit-compatablity mode so you wouldn't be able to make use the improvements in the platform... So basicly you would be comparing 2- 2ghz G4 type proccessors vs a 3.0ghz pentium. Which in any normal case with equal video cards the Pentium would have a comfortable edge in performance...

 

[ViRGE]

Drag, you're way off base. One of the few(but biggest) improvements that come from upgrading an arcitecture is the additional registers, as doubling the bits tends to double the registers for most platforms. AMD, for example, gets quite a boost from this, but in comparison, the Athlon has 8 GP registers, the G4 had 32. Now, the G5 may contain a whole 64 registers, but it's reaching the point of diminishing returns, so the benefit Apple gets from doubling the registers is not as great as when AMD doubled theirs. That means the biggest benefit Apple gets from a 64bit chip, besides memory size, is that it can work with 64bit numbers in one pass; the reality of the situation is that few consumer level programs(or games for that matter) use such large numbers in the first place, since you avoid using them when you can. So, while not running in 64bit mode may have made the G5 look a bit weaker than what it can do, it is not a substaincial amount, and doesn't particularly invalidate current comparisons.

 

[drag]

That is true to a certain extent, but there are plenty of other factors besides memory registers. You see as proccessors are going faster and faster you are going to hit a speed limit to the physical nature of the metals and semiconductors you are using. Well in order for that moore's (?) of proccessor speed doubling every year or two or whatever you can't realy on pure clock rates to increase the power. So you are going to have to design the chip better and make more improvements to the design to get more out of it per clock cycle. Well that is the case with the design of IBM's G5 vs Motorola's G4. IBM didn't just increase the clock speed and allow fot morememory registers, they have a much more slick design based on their Power4 proccessors (which blow g5's,g4's pentiums, amd's etc out of the water even though they are much slower mhz wise, but they are also much more expensive).


Another way to look at would be like this..
Pentiums 4 are considured by some to be i786's and i686's by others.

Now you have the i386 proccessor. And say you have some 3-d tester program that was designed originally on that platform. Now your Pentium4 can run that binary that was created by the i386, right? Yes, because the Pentium has built in compatability for that sort of thing, but you won't be able to take advantage of things like MMX or SSE1/2 and other improvements, so if you run that program on 2ghz Pentium4 you will esentually be pretending to be a i386 running at 2000mhz instead of 20mhz. Now if you were able to get the code to that program and change is very slightly and use a compiler like GCC3.X which is designed to take advantage of those improvements then you would see a dramatic improvement in performance.

Now the G4 is produced by motorola, and the "G5" is realy a IBM power970. They are both PowerPC proccessors, like i386 and Pentium4 are both x86 proccessors (even though granted the difference between a g4 and g5 is minor compared to a i386 and PIV). So even though the G5 can run G4 code it can't run it as well as it can more "native" code.

It's along the same lines as AMD64 running XP-32bit vs XP-64bit. The difference isn't great, but it's there.

Of course the x86 line is still a much better platform were gaming is involved anyways especially once you get those Pentium5's and AMD64's come out in a couple months.

 

[zephyrprime]

as doubling the bits tends to double the registers for most platforms

Where'd you get that? I've never heard of such a thing.

Going from 16bit -> 32bit on the x86 platform didn't double the number of registers.
The IBM power4 only have 32 registers like most risc architectures and the 970 is a derivative of the power4 so I reckon that it'd have 32 registers also.

But all of ViRGE's other points are right on except running in 32bit mode might be a little faster than running at 64bit mode because the instruction stream size would be somewhat less thus making the L1 code cache a little more effective. Or maybe not if 32bit performance was neglected during design and more effort put into 64bit performance.

 

[ViRGE]

drag, if you're talking about SIMD stuff, AltiVect, SSE, and whatnot, you are right to some extent, but not completely. Among some of the changes between the 386 and the P4 included out of order execution, branch prediction, pipelined FPUs, multiple FPUs, multiple ALUs, etc. While SSE was also added, it is the only thing that the program can not take advantage of, everything else the CPU will use on its own(albeit not as effeciently as a recompiled program). In Q3 the G5 was at a disadvantage compared to the P4 when you talk about SIMD instructions(id is just now trying their hand at AltiVect stuff), but it is not a night and day difference; the P4 will always reap the benefit of being Q3's native platform, and will always have a slight advantage as a result.

 

[cmdrdredd]

Originally posted by: drag
So your PC with a 1024x784 at 32 bits color scored 375 fps? That's pretty good. A tomshardware.com a 3.0 ghz gets 400fps. So I beleive you.

They were probably comparing the 3d performance with a dell desktop which comes with the geforce4 MX line of cards standard and if you read the pages the Mac had the 9800 pro upgrade installed. Even with those fast proccessors (dell) I could see that happening.

And apple was vauge enough that they weren't actually dishonest about it. Sorta along the lines that Intel/Amd/Microsoft uses for their benchmarks.

THat's is some bad mojo, but still that's not what I was talking about. I was talking about the test veritest did. They are a benchmarking company and their reputations depend on it. If they screw up then nobody can use them again, because their reputation will bring along bad vibes.

But still it doesn't change the fact that because the platform isn't designed for gaming, but for other uses, I will take the veritest tests at face value until I see some proof otherwise. Meaning professional-level benchmarks proving them wrong, and not just idle vitriolic speaches from AMD-fanboy sites.

yes but you can tell that domething is wrong when my PC (around 3Ghz) with a 9700Pro which Dell's gaming PCs come with is getting much more than Apple says the Dell would. Like I said...probably the standard 3Ghz $900 system or whatever

 

[zephyrprime]

spec

 

[Pariah]

Originally posted by: cmdrdredd

Originally posted by: drag
So your PC with a 1024x784 at 32 bits color scored 375 fps? That's pretty good. A tomshardware.com a 3.0 ghz gets 400fps. So I beleive you.

They were probably comparing the 3d performance with a dell desktop which comes with the geforce4 MX line of cards standard and if you read the pages the Mac had the 9800 pro upgrade installed. Even with those fast proccessors (dell) I could see that happening.

And apple was vauge enough that they weren't actually dishonest about it. Sorta along the lines that Intel/Amd/Microsoft uses for their benchmarks.

THat's is some bad mojo, but still that's not what I was talking about. I was talking about the test veritest did. They are a benchmarking company and their reputations depend on it. If they screw up then nobody can use them again, because their reputation will bring along bad vibes.

But still it doesn't change the fact that because the platform isn't designed for gaming, but for other uses, I will take the veritest tests at face value until I see some proof otherwise. Meaning professional-level benchmarks proving them wrong, and not just idle vitriolic speaches from AMD-fanboy sites.

yes but you can tell that domething is wrong when my PC (around 3Ghz) with a 9700Pro which Dell's gaming PCs come with is getting much more than Apple says the Dell would. Like I said...probably the standard 3Ghz $900 system or whatever

Apple claims both the G5 and Dell system had a 9800Pro in them. I don't know what the rest of the specs for the Dell were beyond a 3GHz P4, but something is oddly wrong with the system they tested if it performed that poorly. Maybe they reduced the RAM to 64MB or something.

 

[drag]

What exactly are cpu registers anyways? I know that each of the G5 processors has a individual 64bit databus going to the system controller this bus operates at 1/4 the speed of the proccessor 500mhz or 1ghz in ddr language for this specific proccessor (Faster cpu's will have corrisponding faster buses, and out of the system controller it has a 128bit databus going to the memory, were it can access 2 banks at once (ala the nforce2's dual memory channel for AMDs...)

Now does the registrers have to do with the amount of physical wires going to cpu? Or is simply a place to store bits inside the cpu until it is used?

 

[zephyrprime]

Processor registers are a critical part of the von neumann computer architecture and hence all practical computing devices since they're all von neumann machines.

Basically, a register is a place to store data but not "until it's needed" like a cache. A register is the foremost data storage mechanism in a processor. And it doesn't have anything to do with the bus width of a processor (although it is corellated.)

Look at this bit of assembly code I got from IBM.
http://www-106.ibm.com/developerworks/library/l-ppc/?t=gr,lnxw16=PowPC

msg:
.string "Hello, world!\n"
len = . - msg # length of our dear string

li 0,4 # syscall number (sys_write)
li 3,1 # first argument: file descriptor (stdout)
# second argument: pointer to message to write
lis 4,msg@ha # load top 16 bits of &msg
addi 4,4,msg@l # load bottom 16 bits
li 5,len # third argument: message length
sc # call kernel

Here's what it means:

li 0,4 load the number '4' into register 0
li 3,1 load the number '1' into register 3
lis 4,msg@ha load the first 16bits of the address of "msg" to register 4
addi 4,4,msg@l add the second 16bits of the address of "msg" to register 4
li 5,len load the length of the string "msg" ino register 5
sc call kernel

 

[wetcat007]

Hmm for the price of one of those macs i could find a nice overpriced alienware with a P4 3.2GHz 1GB of memory and a GeForce FX 5900 Ultra, and a much better looking case, and bigger hd, and an OS that dominates the industry, as well as the standard type platform, not an oucast type that only suports 1 OS. On PC there are multiple OS's available unlike Mac being locked to well Mac, on PC u can use Linux or windows, the 2 big ones and then theres BeOS and other os's out there as well. I used my friends new G4 2000 dollar computer, and I must say it's damn slowwww, it takes too long to open IE or Mozilla or any browser for that fact, and it's just slwo for menus and choppy. The interface for MacOS for it is rather bloated and clunky. Games do not run VirtualPC he has on there runs XP as if it were a 386 (literally the mouse moves at 1FPS). What exactly the point of Macs these days are I don't know... MacOS X in some ways was a step up as well as it was much more bloated.. Anyways I've said enough, MacOS wouldn't be too bad, but Mac's attempt to use there own hardware wont last, Microsoft, and Linux do not produce their own hardware or computers, with the exception of mice and keyboads I guess. Macs really havnt come all that far, they are still a troubled company that doesn't have a place in the future, in fact I've pretty much ruled them out as a competitor the only competitior as far as I know to Windows, is Linux, a much more worthy alternitive, and if u want usefriendly how bout Lindows? Long live X86.

-Mark

 

[cmdrdredd]

Originally posted by: Pariah

Originally posted by: cmdrdredd

Originally posted by: drag
So your PC with a 1024x784 at 32 bits color scored 375 fps? That's pretty good. A tomshardware.com a 3.0 ghz gets 400fps. So I beleive you.

They were probably comparing the 3d performance with a dell desktop which comes with the geforce4 MX line of cards standard and if you read the pages the Mac had the 9800 pro upgrade installed. Even with those fast proccessors (dell) I could see that happening.

And apple was vauge enough that they weren't actually dishonest about it. Sorta along the lines that Intel/Amd/Microsoft uses for their benchmarks.

THat's is some bad mojo, but still that's not what I was talking about. I was talking about the test veritest did. They are a benchmarking company and their reputations depend on it. If they screw up then nobody can use them again, because their reputation will bring along bad vibes.

But still it doesn't change the fact that because the platform isn't designed for gaming, but for other uses, I will take the veritest tests at face value until I see some proof otherwise. Meaning professional-level benchmarks proving them wrong, and not just idle vitriolic speaches from AMD-fanboy sites.

yes but you can tell that domething is wrong when my PC (around 3Ghz) with a 9700Pro which Dell's gaming PCs come with is getting much more than Apple says the Dell would. Like I said...probably the standard 3Ghz $900 system or whatever

Apple claims both the G5 and Dell system had a 9800Pro in them. I don't know what the rest of the specs for the Dell were beyond a 3GHz P4, but something is oddly wrong with the system they tested if it performed that poorly. Maybe they reduced the RAM to 64MB or something.

the only thing I can think of is the Mac was tested with MP and sound off. This is fine, but the dell was probably not tested with sound off (MP wouldn't be a factor since Q3 is not SMT aware only SMP). My PC with sound off produces very close to those numbers. 280fps default 32bit 1024x768. I'll wait for some independent sites to review and benchmark the mac in Q3, UT2k3 etc and compare then. Right now it's all up in the air and I'd be willing to bet that the Mac is still slower in gaming than the PC even with the new systems.

 

[wetcat007]

Originally posted by: drag
Also now that I think about it, those quake3 benchmarks are kinda unfair on the G5...

Because quake3 is a 32bit program, so it can't take advantage of the 64bit stuff and other improvements and unless that Id gave code to the Mac developers so that they could compile their own code and optimize it for the G5 it would be running in 32bit-compatablity mode so you wouldn't be able to make use the improvements in the platform... So basicly you would be comparing 2- 2ghz G4 type proccessors vs a 3.0ghz pentium. Which in any normal case with equal video cards the Pentium would have a comfortable edge in performance...

Why is anyone even concerned with Quake 3 performance anyways, lol 300FPS not enuff???? lmao!

 

[cmdrdredd]

Originally posted by: Sideswipe001

Now that the G4 Quicksilvers are two generations old their prices should go down quite a bit, and maybe some of the Mac haters will actually go out and TRY one for a month to see if they really hate it as much as they have convinced themselves that they do.

I've had to troubleshoot and take care of a few Macs at work. I would never, ever use one at home. Say what you want about their stability or speed, I've had nothing but problems making the things work right. Our company uses a large Canon lazerjet printer for most of its printing. MacOS X? No drivers. Doesn't work. We have an Exchange 2000 server for E-mail. Outlook for Mac? Only in OS9 compatability mode, which needs a completely different set of printer drivers. And is there a version of Quickbooks for Mac? Sorry.

One of the employees uses their old Mac laptop. That's on OS9 and it takes it literally about 2 minutes from power on before you can do anything on it - and even then there's a good chance of it locking up on you.

My experiences with them have never been good. I'm glad some people enjoy them - they have their place in teh world - just not *my* world.

aside from drivers not available for OSX I've had no trouble with my Macs at home. Maybe cause I use them alot, but if you had a Mac first you'd probably not have a whole heap of trouble with them and what trouble you did have you could fix without problem from your experience. Much can be said about the PC market.

 

[Pariah]

the only thing I can think of is the Mac was tested with MP and sound off. This is fine, but the dell was probably not tested with sound off (MP wouldn't be a factor since Q3 is not SMT aware only SMP). My PC with sound off produces very close to those numbers. 280fps default 32bit 1024x768. I'll wait for some independent sites to review and benchmark the mac in Q3, UT2k3 etc and compare then. Right now it's all up in the air and I'd be willing to bet that the Mac is still slower in gaming than the PC even with the new systems.

Looking at the Apple performance PDF, they say the Dell system is a Dimension 8300. Visiting Dell's site and customizing an 8300 with an 3GHz P4 upgrade and 9800Pro upgrade as well as dumping the monitor, the cost is $1,529. That seems like a fair comparison, a $1,529 system vs a $3299 system. The Dimension has 256MB RAM which is rather skimpy for an XP system. I don't know if Quake 3 will run over that limit, probably not, but that may be an issue.

 

[drag]

Originally posted by: wetcat007

Originally posted by: drag
Also now that I think about it, those quake3 benchmarks are kinda unfair on the G5...

Because quake3 is a 32bit program, so it can't take advantage of the 64bit stuff and other improvements and unless that Id gave code to the Mac developers so that they could compile their own code and optimize it for the G5 it would be running in 32bit-compatablity mode so you wouldn't be able to make use the improvements in the platform... So basicly you would be comparing 2- 2ghz G4 type proccessors vs a 3.0ghz pentium. Which in any normal case with equal video cards the Pentium would have a comfortable edge in performance...

Why is anyone even concerned with Quake 3 performance anyways, lol 300FPS not enuff???? lmao!

I think that, my slightly dense freind, is that you miss the point of the argument completely. In hardware comparesions Quake3 is used to measure the releative performance of the CPU since doesn't make use of the better video card processor's of late to do the proccessing like more modern games like UT2003. Since you can eliminate variables such as the video card by giving the 2 machines identical cards it is a valid tool for comparision since it is such a well known and understood benchmark.

And since the dual G5's showed as much as 194% cpu performance advantage over a single 3ghz in tests designed to show pure cpu power. Well it was pointed out to me that the quake3 benchmarks showed a considurable lack of cpu pontential since the Mac scored a 335fps while a Pentium4 can easily score 375fps at the assumed same settings.

Since I beleive the veritest results (although I do take them with a grain of salt, I figure the dual g5's to be closer to 145% as powerfull as the Pentuim4's due to historical clock efficientcy compariasons.), I was trying to take into account the fact that q3 for mac is designed to run on a G4 platform so in order to run in a g5 is has to operate in compatability mode, thus losing some of the performance advantages of the new platform.

Now I feel slightly embarased to have to explain this to you... or in your own words:


Why is must I feel compelled to explain myself to this lamer anyways, lol 4-6 posts of discussion not enuff???? lmao!

 

[drag]

[q

the only thing I can think of is the Mac was tested with MP and sound off. This is fine, but the dell was probably not tested with sound off (MP wouldn't be a factor since Q3 is not SMT aware only SMP). My PC with sound off produces very close to those numbers. 280fps default 32bit 1024x768. I'll wait for some independent sites to review and benchmark the mac in Q3, UT2k3 etc and compare then. Right now it's all up in the air and I'd be willing to bet that the Mac is still slower in gaming than the PC even with the new systems.[/quote]

I agree with you to a point (plus I agree to the fact that x86 is better suited for gaming), the "competative" 3.0ghz p4, is very easy able to score +350fps in quake3 at 32bit color and 1024x786 resolution. I don't think sound on would cause it to be rated at 233fps for the 3ghz. The only thing I can explain this is that the 3.0 was a stock lower-end model with a crappy 32 or 64 meg video card. A very very bad comparasion. And anyways the mac only scoring 335 fps with a ATI 9800 pro is a disapointment.

 

[Pariah]

Again, you're not listening, the Dell system was a Dimension 8300 using a 9800Pro. You can read it yourself straight from the horse's mouth:

Performance White Paper (PDF)

Also, it should be noted that Quake III doesn't gain any performance from SMP systems, so basically Apple is claiming an unoptomized single 2GHz G5 is 30% faster than a 3GHz P4 (P4's known to be the king of Quake III) which is rather suspect.

 

[ViRGE]

Originally posted by: drag
I was trying to take into account the fact that q3 for mac is designed to run on a G4 platform so in order to run in a g5 is has to operate in compatability mode, thus losing some of the performance advantages of the new platform.

Drag, there is no "compatibilty mode", the G5 runs all PowerPC code natively; you are not taking a hit by running G4 code on a G5. That said, a proper recompile could boost performance by a small margin, but if you take a comparison of the G3 vs. the G4, the G4 ran G3 code just fine, much like the G5 runs G4 code. In fact, because the G5 doesn't do anything new with AltiVect, there is not a lot of tweeking to be done, only a smidge here and a smidge there. If you're doing hard-coded loops, or something else programed via assembly on a per-processor basis(Dnet client comes to mind), then you might lose a lot of speed, or be able to gain it through tweeking, but in something as generic as Q3, a more optimized compiler isn't going to be a magic bullet, it's just another small boost in the big picture.

 

[drag]

I wasn't aware of the whitepaper pdf I was just going off of the html page on apple's website. They didn't make any mention of what exactly type of computer they were comparing the g5 against on the html spec's, just that it was a "competator's 3ghz machine".

That 3d comparansion is obviously suspect.

As I stated many times before, tom's hardware stated in it's benchmarks with the ATI 9700 that the 3ghz scored 400fps at 1024x786 at 32bits color. If I listened to apple's comparision then the Pentium4 3ghz with the 9800 is only slightly faster then my AMD 1700+ with the geforce2 gts-v video card, which is laughable. It can't be just the sound, unless it's a crappy sound card that has no chance of keeping up with the benchmarks. Apple is obviously misrepresenting it.


Sorry I missed the bit were it was said that the dell had a 9800, too

 

[Acanthus]

Please, play your 6 available games at competitive speeds on your apple... I'll be playing everything out on my PC

 

[drag]

what are talking about?

The only time that I use mac's is to show students how to use different high-end graphical programs, which is what the Mac is best at, or just configuring them and keeping them updated.

As far as games, I realy am not much of a gamer anymore anything I like to play is either on the playstation2 or Quake3 mods like urbanterror or truecombat on my AMD-based linux box.

So you go ahead and play your multitudness number of games.

 

[Fencer128]

Here's something from slashdot that made me think twice about apple's benchmarks

Greg Joswiak, vice president of hardware product marketing at Apple, in a phone interview today, defended Apple's performance claims for its upcoming Power Mac G5, after they came under fire in the wake of yesterday's announcement. Read on for the details.
Joswiak went over the points in turn, but first said that they set out from the beginning to do a fair and even comparison, which is why they used an independent lab and provided full disclosure of the methods used in the tests, which would be "a silly way to do things" if Apple were intending to be deceptive.

He said Veritest used gcc for both platforms, instead of Intel's compiler, simply because the benchmarks measure two things at the same time: compiler, and hardware. To test the hardware alone, you must normalize the compiler out of the equation -- using the same version and similar settings -- and, if anything, Joswiak said, gcc has been available on the Intel platform for a lot longer and is more optimized for Intel than for PowerPC.

He conceded readily that the Dell numbers would be higher with the Intel compiler, but that the Apple numbers could be higher with a different compiler too.

Joswiak added that in the Intel modifications for the tests, they chose the option that provided higher scores for the Intel machine, not lower. The scores were higher under Linux than under Windows, and in the rate test, the scores were higher with hyperthreading disabled than enabled. He also said they would be happy to do the tests on Windows and with hyperthreading enabled, if people wanted it, as it would only make the G5 look better.

In the G5 modifications, they were made because shipping systems will have those options available. For example, memory read bypass was turned on, for even though it is not on by default in the tested prototypes, it will be on by default for the shipping systems. Software-based prefetching was turned off and a high-performance malloc was used because those options will be available on the shipping systems (Joswiak did not know whether this malloc, which is faster but less memory efficient, will be the default in the shipping systems).

As to not using SSE2, Joswiak said they enabled the correct flags for it, as documented on the gcc web site, so that SSE2 was enabled (the Veritest report lists the options used for each test, which appears to include the appropriate flags).

In this light, the comparison seems more balanced, don't you think?

Cheers,

Andy

 

[drag]

One of the big deals that people make out of these benchmarks is how they should of used windows vs linux or used gcc vs icc as a compiler for the tests. That's kinda a double edged sword, first off icc is specificly designed to run on intel proccessors, it won't work on any thing else. Gcc on the other hand will work on pretty much anything and that's what it was designed to do, plus it is coded in such a way that it uses higher amounts of cpu power to produce much smaller binaries then the Intel compiler does, although the intel produces faster running binaries. IBM doesn't have a power970 specific compiler like the Icc so Gcc is the closest your going to get. Plus the Gcc 3.4 is actually a slower then the previously Gcc 2.* generation. That is the reason the spec2000 benchmarks differ from Apple's test and the older comparision. That's about the only reason. Trying to compare icc on a intel vs gcc on a G5 would be like comparing fps performance with Quake3 on a pentium4 vs UT2003 on a AMD.

The speed difference between gcc on intel and icc on intel can be as much as 40% difference in compile time.

However that being said, Apple supposedly has put a lot of effort into Gcc since that is the compiler it uses for it's own OS, so that may produce scewing in Apple's favor... However gcc was originally designed on 386-style hardware and that's what most it's developers use and concitrate on... So who knows?

As far as windows vs linux, that's pretty pointless too. Something as direct as compiling code isn't going to be that big of difference. Both Linux and Windows is heavily optimized for the hardware and it's a trade off between one is faster in some cases and the other is faster in others.


What I am waiting for is performance comparisions that compare the stuff that realy matters to end users. Like mp3 encoding or video compression tests, running comparative games, rendering 3d scences and image manipulation and other end-user stuff like that. I think that will be more telling, then just cpu tests.