Slashdot Mirror
G5 vs Opteron, Finally
metfoo writes "It's been months since the G5 and Opterons have been available for purchase. When the G5 systems were first released, many Mac bashers and AMD nuts discredited the G5's performance. They always ended their comments with 'Wait until its compared to an Opteron, then we'll talk.' Well, it's finally time to talk. Barefeats has posted an article comparing the two systems. The G5 line was compared to a Dual 2GHz Opteron and the results are impressive. In gaming, the Opteron system proved to be superior, which is partly due to the superior 9800XT over the base Radeon 9800. The G5 spanks the Opteron in many of the non-gaming tests, except for the Photoshop tests."
If they would have used a linux platform instead of Windows...
Jason Faulkner
Old Os Administrator
jason@oldos.org
oldos.
One main issue with the UT 2003 tests. It doesn't say if they are running UT 2k3 2225.1 or 2225. 2225.1 brings MASSIVE performance increases. From the notes:
It's much, much faster. Several optimizations have been made, lots of Altivec code has been added, and the entire sound subsystem has been rewritten. Performance improvements of 25% or more over the original retail version are typical, with single CPU systems achieving a more noticible gain.
O.K., hopefully this will put to bed all those folks who cry about Apple computers being so damned expensive. Feature for feature, the G5 is about $600 cheaper than the Opteron. I certainly found this out when I was pricing workstations from Dell and other Wintel manufacturers and the G5's from Apple. I went with a fully loaded G5 and the price delta was $1200 cheaper going with the G5. Plus, OS X is soooooo nice.
Visit Jonesblog and say hello.
Unreal Tournament 2003 runs in Direct X mode on the Opteron and OpenGL mode on the G5. Some say this isn't a fair test but if you are choosing between the two systems, you need to know how it runs your favorite game.
If a G5 running in OGL gets such low scores something is wrong. D3D renders slower and requires more processing power than OGL.
-illumina+us "I put on my robe and wizard hat..."
I'm geniunely curious - how? A 64-bit application can't do much without 64-bit system calls. Does mach-o let you dynamically load 64-bit code in a 32-bit program (all the 64-bit code would be able to do is computation, given the lack of system calls). Or is there a windows64-on-windows-like 64-bit wrapper over libc?
Price comparisation:
- comparisation of artificially low memory systems as Apples prices are where Apple makes the most on. On the one hand claiming 'we don't want to build ourselves as Apples can't be build, and then going to another store to add memory, just isn't fair when comparing prices.
- Boot-testing the Mac for performance difference with other the HD is a good thing, but the test in the other direction (booting the PC with the other HD might reveal that the bottleneck is in the other direction).
- MacOS X is certainly better in 64bit environments than not wanting to run beta software on a system bought for performance.
- The problem with the Mac is also that the graphics subsystem is already dated. The release cycle of Macs is just too long. When they're first released they -arguably- beat most of the fastest PC's. But the next version is only released at quickest 6 months later, if you compare at that time with latest hardware. Macs just can cope up.
- I also assume that near the end of the cycle, Apple's profit margins are incredible high. It's a very good marketing tactic to keep hardware and software tied to each other, keeping it all under control.
- As I'm typing this on my top-equipped 12" PowerBook, I must admit that MacOS X is a good OS and the hardware is very good (this laptop was cheaper than any comparable hardware at the time I ordered it - not any more at the time when it got delivered)
- And as a rule of thumb, I always say it's better to buy a less expensive system and upgrade it quicker than to go for the fastest and be stuck with it for an extra year.
- Macs also have a better second hand value, and that shouldn't be forgotten when taking the price into account.
- But most performance comparisations clearly SUCK because they tend to be optimised for a certain system (because of lack of knowledge of the party), or highly dependent on release schedules of involved hardware or software.
I think what you really mean is now that my parents don't buy my computer. ;-) I definitely agree though. I recently upgraded my 450MHz G3 to a PowerBook, but that was only because I wanted a laptop. The whole time, I felt sort of dirty, because my G3 is by no means a bad machine. It runs everything I do just fine. I just can't take it to work with me, like I can my PowerBook.
irb(main):001:0>
It's running Windows XP, so we have a bunch of slow down right there. I had a 200ish mhz laptop with FreeBSD and a 400mhz WinNT machine running the same program, and the FreeBSD machine ran it about twice as fast. Moral of the Story: Operating systems can add a lot of overhead.
Of course, I don't know what the overhead of OSX is against XP, but I think it's something I had to point out.
This article strictly compares the 970 to the G5 using GAMING benchmarks.
The Opteron and the G5 (IBM PowerPC 970) are two disparately different chips meant to serve two different purposes. The Opteron is AMD's server chip designed to handle for the most part, 64 bit high performance database applications and applications which require large memory models in which the 64 bit memory addressing is needed-- NOT 64 bit desktop applications or games. That's reserved for the Athlon 64 which is clocked significantly higher and has a much smaller L2 cache than the Opteron. On Gaming and desktop content creation benchmarks the Athlon64 is a much better match for the 970.
If you want to compare apples to apples I would compare the IBM Power4 to the Itanium2 to the Opteron, hook them all up to an EMC storage array using fiberoptic SAN connections, and run a few million row length Oracle and DB2 databases and some SQL database benchmarks -- and for load up a few gigantic thermodynamic simulations up into main memory and see how quickly they can run through them. THAT would be an appropriate test for these server chips.
You had to sell the Sparc? How and why? An 80 MHz Sparc returns a sad, sad amount of money on eBay. I sold a 64-bit Sun Ultra 1 a few months ago. It was, I think, a 166 MHz. I got all of $28 for it.
I can't believe you 'had' to sell the Sparc, and if you 'had' to I can't believe you got enough money for it to justify selling it.
I love my Sun hardware, but right now it's not hardly even worth putting up for sale.
they were totally debunked as full of crap. they used Adobe premiere for god sakes...a program that is not even available on OS X and had to be ran in EMULATION MODE.
I am the Alpha and the Omega-3
That page doesn't explain anything. The real question is whether you can pass 64-bit pointers to system calls. Like, can I do a write() from a memory buffer that's above the 4GB limit? Otherwise, its more of a Windows NT PAE-type hack rather than actual support for 64-bit apps.
A deep unwavering belief is a sure sign you're missing something...
Or: if you're a gamer, own a console. While we aren't quite at the point where all games are "better" and "first" on the console, that day seems to be growing closer. Game stores are stocking fewer and fewer PC games, and signs point to stores like Game Stop eliminating them completely.
Then its not 64-bit at all, it just lets you do 64-bit integer math. Well, hopefully a not-to-distant revision of OS X will get proper 64-bit support. Its kind of embarrasing that Apple's new UNIX-derived OS is having problems moving to 64-bit, when the big UNIX vendors like Sun and SGI made the 32-bit -> 64-bit transition rapidly and uneventfully years ago.
A deep unwavering belief is a sure sign you're missing something...
The AMD64 has much more than 16 of each type of register. It has 16 user-visible integer registers, which makes for better processor optimizations. It has many more internal rename registers. I don't know how many the AMD64 has, but the Pentium4 draws its 8 integer and 8 floating-point registers from a bank of 128 internal GPRs. The AMD64 is probably comparable.
And in RISC systems, its 32 of each, seperated into integer and floating-point banks.
A deep unwavering belief is a sure sign you're missing something...
Uneventfully? Sun and SGI didn't have millions if idiots running their systems either. Who would have noticed if there were problems and why?
Answer this question: will final cut pro run on an x86 based machine?
To me, and most Mac users, gaming is irrelvant. Most people that use Macs are in a profession such as desktop publishing, video production, or graphic arts. Sure they may play a game or two, but their machine is used for work.
I do a little bit of everything with my G3 700Mhz 14.1" iBook, but mostly its MS Office, Mail, Safari, and Quark that I use. Along with Final Cut Pro and Photoshop when need be.
Our office is 95% Mac and 5% FreeBSD, which we run on Althon white boxes, and we have beat out competition because of productivty. We are not spending loads of time with viruses and patching security issues on a weekly basis. Our machines rarely lock up, none have crashed (knock on wood), and that helps with the bottom line.
Does it help in video rendering to have the extra speed and power of the 64-bit G5? Yeah, the faster a project is rendered, the quicker we move on to the next. But for everyday business use, our older G4 500's, 867's, and Dual 1.25gz will serve us for years to come and even though Apples cost more up front, we know we have saved time and money by using macs for our desktops.
"The problem with socialism is eventually you run out of other people's money" - Thatcher.
Huh?
For the 3 games you can play on a G5?
Why not compare more relevant things? Okay, you say, let's do some applications tests. Okay, photoshop filters. A couple rendering jobs. Yawn. Who the hell does this on their machines all the time? Okay, now the five of you, leave the room.
Why not give me some more information about the guts of the machine, like how fast memory access is or how each bus design handles contention issues, explain why they're relevant in various facets of operating system or application execution, and provide some anecdotal evidence by way of application benchmarks. Hell, run them in a debugger so we can see if our assumptions about system behavior are correct in real-world situations.
These people get a dual G5 and a dual Opteron and all they do is run Photoshop and Quake 3 on it and call it a night. What the hell? Where's the investigation, the effort? How much more boring could that article have been? (Okay, maybe they could've lost the graphs and numbers and just told us, "Trust us, this one's faster", but that would've seemed like they were phoning it in.)
In summary, I was a little disappointed.
- A.P.
"Remember when the U.S. had a drug problem, and then we declared a War On Drugs, and now you can't buy drugs anymore?"
Maximum PC, Dec2003 issue, has a showdown of the Athlon64 F-51 vs. Intel P4ExtremeEdition vs. Mac dual G5 2.0ghz.
Basically, they came up with the conclusion that the P4EE, though it's very hard to find one, is a bit faster in the majority, the Athlon64 takes up almost all the rest of the field, and the Mac G5 comes in first in one or two areas. They also concluded that 1.) the mac is a good deal for a fast computer, 2.) the intel chip looks more like vaporware than anything else, as you can actually get AMD's and PPC's, and 3.) the difference in speed b/t the athlon64 and the P4EE is negligible.
~Will
sig?
here, or search on www.ibm.com for "XLC OS X beta" without the quotes.
Next year it will be 128-, 256-, 512-, 1024-bits.... see a trend? Where does it end? And, do people always believe more expensive and bigger numbers (faster clock) is necessarily better?
Is the migration to 64-bit part of the scheme of planned obselecence? Or merely an excuse to fix the 2038 32-bit Unix Y2K problem by a hardware hack? Haven't we learned from the wintel platform that making hardware/software purposely incompatible? M$ft/intel ostensibly maintains backward compatibly, but in truth, Falcon AT is not going to run on native DOS on Athlon64 FX or Pentium EE. Either do code-morphing transmeta style, build an emulator, or break compatibility and design a usable, general-purpose, stable system to last for the foreseeable future and stop changing things!
1) 64-bit chips are vastly more complex, requiring 1 more gate delay for mux/dmux's.
2) Clock skew becomes an even bigger issue, as most of the cpus (excluding ROMs/RAMs/cache/CAM) is dedicated to clock synchronization by area.
3) What does one need w/ 64-bit buses? A 32-bit bus is sufficient, word size can be increased by multiples of 32-bits regardless. A 64-bit signed add/sub can be done in two 32-bit add/subs, it can even be done in parallel using carry prediction and double the hardware, save the carry-out choice. All this extra precision is unnecessary in most cases... it would be better to have an uber-efficient 4 x 8-bit operand parallel CPU that could execute code as if it were microcode. That way, you could do variable precision operations, possibly saving cycles. Otherwise, you'll have that honkin' 64-bit UINT calculated every time you want to increment by one. Let's see.... increment overflow will only take 585000 years at 1 MPS. Darn that off-by-one bug!
4,5,...) Pipeline stall cycle length (deeper pipelines), more complex functional units, larger footprint, cycle efficiency, etc/etc.
Serial busses are potentially faster than any parallel bus, since a serial bus can have embedded clock (manchester, diff manchester, etc.). And the differential serial bus is always nice in the noise-immunitity department.
Change isn't always progress, but an excuse to push product out the door.
"Marketing is the art of trickery and deceit to convince people to buy your product, much like dating and job interviewing."
The biggest trick the devil pulled was letting lawyers become politicians so they can write the laws.