Will X, assuming it is fully optimized for the chip, run at a par at the OS level (e.g. copying, finder-tasks, etc.)? Third-party software will, it seems to me, be a grab-bag; but the OS's are what I am thinking about here....
That's not exactly taxing on the CPU though. The gripe with OSX is usually it's 3D graphics implementation be it the drivers or OpenGL. It'd be quite embarrassing for Apple if, for example, DOOM3 or Cinebench still ran much faster on Windows than on the exact same hardware (or as near as) under OSX.
DDR 2 speeds relate to the argument because they point up an area in which Apple could have easily made a necessary improvement, but didn't. There would have been practically zero cost differential. Neither of us knows why they didn't implement this one either.
What i've been arguing all along is that, in the case of the G5 FSB, Apple could not easily have made such an improvement.
Quote:
Of course I read what you wrote. But you are wrong. If you look at APPLE's bus speeds on the G5 PM's over the years, (I'm not interested in what Intel has been doing because Apple has a completely different bus and controller design), you will see that they scale perfectly with the cpu speeds.
It's not a matter of "what Intel has been doing", it's a matter of what the trend is for the entire industry -- or at least for those parts of the industry which experienced significant clock rate increases at all.
Quote:
You make the assumption, with no reason to, that at some point in time there will be some disconnect between bus and cpu speeds. There has been no evidence for that. If anything the evidence is quite the opposite.
There has been no evidence if you restrict your point of view exclusively to the few iterations of the G5 that have shipped so far (and went from 2 GHz top to 2.7 GHz tops in the process). What makes you think that Apple / IBM can somehow magically just scale the FSB speed as they please when everybody else is unable to do so?
Quote:
PCI is a 5 volt logic powered board, and PCI X is either a 3.3 or 3.5 (I forget which right now) volt logic powered board. 5 volt boards won't work on PCI X, and 3+ volt boards won't work on PCI. Some boards can use either.
It goes like this:
PCI-X cards and slots use 3.3V signalling.
For plain PCI, cards can use 3.3V signalling, 5V signalling or they can be "universal" (i.e. support both). The variants can easily be distinguished by looking at the position of the notch(es) in the PCI connector. PCI slots, on the other hand, can either support 5V signalling (most PC boards) or 3.3V (newer boards, and also slots running at 66MHz), but not both. The slots are also coded such that only cards that support the signalling voltage used by the slot physically fit.
Quote:
PCI X isn't just about AGP.
Right. In fact, I don't think there are PCI-X graphics cards at all (or at least they're extremely rare).
Quote:
Express serves no purpose for the majority of users either. The 16x slot is needed only by those buying the highest performing boards, and the rest of the slots are at least twice as fast as the slots in PCI X, which in turn are twice as fast as those in PCI.
Actually, compared to the PCI slots used in almost all PC motherboards (32 bit, 33 MHz) or on G4 Macs (64 bit, 33 MHz), PCI-X (64 bit, up to 133 MHz) is four or eight times as fast, respectively. (Of course, that's just comparing theoretical maximum numbers, not real-world throughput.)
Quote:
It's odd that you would say that, because the opinion stated many times over the years leading up to the introduction, was that Express would be more expensive than PCI X. That was one of the reasons PCI X was developed, and one of the reasons given in public statements.
It's more expensive in that, due to lack of physical backwards compatibility, it obsoletes nearly all the expansion cards out there, forcing users to buy new ones. Since this is not something that's going to go too well especially with the PC world, you need to provide some PCI slots as well for backwards comptibility, and thus need a bridge chip, additional PCB real estate, etc. -- which is not the case for PCI-X. (Also, note that I was talking about x1 slots, not x4/x8/x16 ones.)
Quote:
Everything electronic gets cheaper over the years, and sometimes by the time something is introduced, the price has dropped past expectations. Even Intel admitted that Express would be more expensive than PCI X. And I'll state again that the reason PCI X wasn't adopted wasn't because of its expense, but because Intel makes at least 50% of the mobo's used by manufacturers, and they were afraid of getting stuck with something that Intel's boards would swamp. It was politics, and the fact that the PCI X manufacturers had no clear upgrade path from version one.
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
Apple just went to DDR2 533. Why? It offers little advantage. Even cheap PC mobo's are using 667 speeds. There are even 800 speed boards coming out.
Quote:
Originally posted by RazzFazz
I don't quite see how this really relates to our argument -- it appears Apple felt it didn't make business sense for them to go with the higher speed DDR2. That, however, doesn't in any way imply that all their decisions are made purely for political / business reasons.
I assume the memory controller cannot handle the 4 DIMMs at the higher speed. In PCs the memory controller can often handle 2 DIMMs at a lower speed and only 1 at higher speed. Or has to use registered DIMMs instead.
(Of course this is always number of DIMMs per memory bank).
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
Actually, the way I recall it, the MSI Neo4 Platinum board I purchased a few months ago had one PCI x16, x4 and x1 port each, three PCI ports and finally one PCI-X port. And this is a fairly mainstream choice and not too costly either. SO I think that combining PCI, PCIe and PCI-X is in fact done by quite a few manufacturers.
What i've been arguing all along is that, in the case of the G5 FSB, Apple could not easily have made such an improvement.
It's not a matter of "what Intel has been doing", it's a matter of what the trend is for the entire industry -- or at least for those parts of the industry which experienced significant clock rate increases at all.
There has been no evidence if you restrict your point of view exclusively to the few iterations of the G5 that have shipped so far (and went from 2 GHz top to 2.7 GHz tops in the process). What makes you think that Apple / IBM can somehow magically just scale the FSB speed as they please when everybody else is unable to do so?
I think I realise what the problem here is.
You keep talking about multiple GHz bus's. But that's not the question at all.
Ok, I quess we do have to bring Intel into this.
Intel's memory bus's have gone from 400MHz to 1.066GHz in 2 1/2 years. The processor speeds have gone from slightly under 3GHZ to 3.8 GHz, and back down to 3.4 GHz during that same time period.
So, bus speeds have gone up vastly faster than chip speeds have. The opposite of what you said earlier.
So let's see what the situation really is.
Intel has a bus that is bidirectional, but can only pass a signal one way at a time. But when the signal does pass through, it does at the full 1.066GHZ speed on their newest bus.
Apple has a bus that at the present time runs at 1.25GHz, down from 1.35GHz on the dual 2.7.
BUT, this bus consists of two seperate bus's. Each one is unidirectional. And each one operates at half the speed of the overall bus speed. The Apple bus at 1.25GHz can pass more information over the bus at one time than Intel's 1.066GHz bus can.
But, each bus is only running at 625MHz, much slower than the 1.066GHZ that the traffic runs at in Intel's bus in either direction.
If Apple went to a 2GHz bus for the dual core 2GHZ machine, each unidirectional bus would only be running at 1GHz. Slightly slower than Intel's.
If Apple used it on the dual core 2.3GHz, it would be running at 1.15GHz, slightly faster than Intel's.
If they ran it at 2.5 GHz on the quad, each side of the bus would be running at 1.25GHz. That's just under 200MHz faster than Intel's.
None of Apple's bus's would be running at multiples of 1GHz as you've been saying, and been so concerned about.
So we are not talking about these high speeds that you are thinking about. Just slightly faster than what Intel has now.
Quote:
It goes like this:
PCI-X cards and slots use 3.3V signalling.
For plain PCI, cards can use 3.3V signalling, 5V signalling or they can be "universal" (i.e. support both). The variants can easily be distinguished by looking at the position of the notch(es) in the PCI connector. PCI slots, on the other hand, can either support 5V signalling (most PC boards) or 3.3V (newer boards, and also slots running at 66MHz), but not both. The slots are also coded such that only cards that support the signalling voltage used by the slot physically fit.
Sorry, but PCI boards are 5 volt, and PCI X boards are 3.3 volt.
If a card supports both, then that card can run in either because the card senses which voltage the bus is using.
Quote:
Right. In fact, I don't think there are PCI-X graphics cards at all (or at least they're extremely rare).
I quess it's sad to know that all of the G5 PM's have been running without graphics cards. We should all run out and tell ATI and Nvidia, so that they can make us some. Here, you are just guessing. I can't imagine how you came up with this one. We are talking 8x AGP in either PCI or PCI X. Same thing electrically.
Quote:
Actually, compared to the PCI slots used in almost all PC motherboards (32 bit, 33 MHz) or on G4 Macs (64 bit, 33 MHz), PCI-X (64 bit, up to 133 MHz) is four or eight times as fast, respectively. (Of course, that's just comparing theoretical maximum numbers, not real-world throughput.)
Yes, that's what I said, PCI X is up to twice as fast as PCI, and Express is up to twice as fast as PCI X.
Quote:
It's more expensive in that, due to lack of physical backwards compatibility, it obsoletes nearly all the expansion cards out there, forcing users to buy new ones. Since this is not something that's going to go too well especially with the PC world, you need to provide some PCI slots as well for backwards comptibility, and thus need a bridge chip, additional PCB real estate, etc. -- which is not the case for PCI-X. (Also, note that I was talking about x1 slots, not x4/x8/x16 ones.)
Well, that's what I was saying. You were saying that it was LESS expensive to implement Express.
Quote:
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
That would be a baaddd idea. Keeping ISA, and now PCI on new machines just holds developement back. It was the ISA boards, for the most part that prevented Plug-n-Play from working.
A couple of quotes from Anand, from Anandtech:
"It is irritating that Apple didn't move to DDR2-667 yet, especially on their highest end configuration (and especially because it can use the bandwidth), but given Apple's relatively conservative nature whenever it comes to memory speeds it isn't a huge surprise."
" In my opinion, the biggest improvement to the new G5s is the move to PCI Express. And here's one thing I really do like about Apple, when they move to a new technology, they really move to it.
There isn't a single parallel PCI slot in the new G5s,..."
I assume the memory controller cannot handle the 4 DIMMs at the higher speed. In PCs the memory controller can often handle 2 DIMMs at a lower speed and only 1 at higher speed. Or has to use registered DIMMs instead.
(Of course this is always number of DIMMs per memory bank).
The controller can handle it. Read the quotes at the end of my above post.
while we're on the topic.. (wait, what was the topic again??) just a comment on DDR vs DDR2 to muse over (go nuts everyone)
DDR2 for powerbooks and 2006 macintels makes sense as intel is standardising to DDR2 support, and DDR2 in portables show reduced power consumption (by how much, well, that's a matter for debate).
according to most PC sites lower-latency but slower speed DDR amd64s vs faster speed but higher-latency DDR2 intels results in very neck and neck comparisons, wrt to memory specific synthetic benchmarks (sandrasoft)
i wonder in the case of the quad g5, if the memory controller is "teh bomb", then in this case DDR with decently tight timings will do fine, DDR2 may not offer any advantage, we may not see powermac g5s with DDR2 until middle of 2006.
while we're on the topic.. (wait, what was the topic again??) just a comment on DDR vs DDR2 to muse over (go nuts everyone)
DDR2 for powerbooks and 2006 macintels makes sense as intel is standardising to DDR2 support, and DDR2 in portables show reduced power consumption (by how much, well, that's a matter for debate).
according to most PC sites lower-latency but slower speed DDR amd64s vs faster speed but higher-latency DDR2 intels results in very neck and neck comparisons, wrt to memory specific synthetic benchmarks (sandrasoft)
i wonder in the case of the quad g5, if the memory controller is "teh bomb", then in this case DDR with decently tight timings will do fine, DDR2 may not offer any advantage, we may not see powermac g5s with DDR2 until middle of 2006.
Sunil, I really hate to break this to you , but the new PMG5's Express machines use DDR2 now.
Now if I were a l33t piratez, I just might keep my mouth shut UNTIL the production MacTel's are released, then and only then would I get the word out. Basic human nature, people talk, loose lips sink ships!
Meh, at some point you would have to release your work at which point Apple would change their anti=piracy strategy and we would all be back to square 1.
There's really no point in delaying releases of such hacks. All you do is delay how long people will be able to use the current versions. The next version will always be incompatible and we start the whole process over again.
You keep talking about multiple GHz bus's. But that's not the question at all.
Ok, I quess we do have to bring Intel into this.
Intel's memory bus's have gone from 400MHz to 1.066GHz in 2 1/2 years. The processor speeds have gone from slightly under 3GHZ to 3.8 GHz, and back down to 3.4 GHz during that same time period.
I do think I put the "before the great brick wall was hit" qualifier in my statement? I was talking about the long run.
Quote:
So, bus speeds have gone up vastly faster than chip speeds have. The opposite of what you said earlier.
Up to and including the 486DX, processor and FSB were at 1:1. Then came the DX2, DX4, the Socket 7 processors, whatnot. With each generation, the FSB multiplier grew on average. It wasn't until recently that CPU frequency just kinda stopped scaling.
Quote:
So let's see what the situation really is.
Intel has a bus that is bidirectional, but can only pass a signal one way at a time. But when the signal does pass through, it does at the full 1.066GHZ speed on their newest bus.
No, it does not. Intel's FSB is quad-pumped, the actual clock rate is 266MHz.
Quote:
Apple has a bus that at the present time runs at 1.25GHz, down from 1.35GHz on the dual 2.7.
The G5's EI is double-pumped, so the actual clock rate is half of what you state.
Quote:
BUT, this bus consists of two seperate bus's. Each one is unidirectional. And each one operates at half the speed of the overall bus speed. The Apple bus at 1.25GHz can pass more information over the bus at one time than Intel's 1.066GHz bus can.
But, each bus is only running at 625MHz, much slower than the 1.066GHZ that the traffic runs at in Intel's bus in either direction.
If Apple went to a 2GHz bus for the dual core 2GHZ machine, each unidirectional bus would only be running at 1GHz. Slightly slower than Intel's.
You're really confusing something here. "Operates at half the speed" does not mean "each direction operates at half the clock rate" (that doesn't even make sense). A "2 GHz bus" runs at, well, 2 GHz. Not "1 GHz per direction" or some such. The thing that's halved per-direction is the bus's width.
The math for the current incarnation of the G5's FSB goes like this:
You have two unidirectional 32-bit buses, clocked at 4:1 but double-pumped so it's effectively 2:1. That gives you about 750MHz x 2 (DDR) x 32 bits/direction x 2 directions = 12GBps aggregate throughput.
Intel's current FSB is a bidirectional 64-bit bus that runs at 266 MHz. It's quad-pumped, so the effective clock rate is 1066 MHz. Thus, you get 266 MHz x 4 (QDR) x 64 bits = 8.5GBps aggregate throughput.
Quote:
If Apple used it on the dual core 2.3GHz, it would be running at 1.15GHz, slightly faster than Intel's.
If they ran it at 2.5 GHz on the quad, each side of the bus would be running at 1.25GHz. That's just under 200MHz faster than Intel's.
You should really drop the whole "each side runs at clock rate x, so the whole bus runs at twice that clock rate" argument. It's the throughput that doubles, not the clock rate. If one lane runs at 1.15GHz, then two such lanes still run at the same clock speed, the whole bus is just wider now.
[QUOTE]
Sorry, but PCI boards are 5 volt, and PCI X boards are 3.3 volt.
If a card supports both, then that card can run in either because the card senses which voltage the bus is using.
[/QUOTE
This is getting fairly annoying. The PCI spec does allow for 3.3V and 5V slots, and 3.3V, 5V and "universal" cards. Go ask Google, Wikipedia or whatever other source you like.
Quote:
I quess it's sad to know that all of the G5 PM's have been running without graphics cards. We should all run out and tell ATI and Nvidia, so that they can make us some. Here, you are just guessing. I can't imagine how you came up with this one. We are talking 8x AGP in either PCI or PCI X. Same thing electrically.
AGP in PCI/PCI-X? What are you even talking about? So, for you, AGP is just a different physical slot form factor and otherwise just PCI(-X)? If that really is what you're suggesting, I suggest you consult any kind of source you'd like, and you'll see that AGP uses different voltages, uses different signal speeds, doesn't have muxed A/D lines, etc., pp.
Also, make sure you take a good look at the Apple hardware tech docs -- the AGP port is attached directly to U3 (i.e. the north bridge), whereas PCI-X is provided by a HT-to-PCI-X-Tunnel that's attached to the HyperTransport port. There isn't even a direct connection between the two.
Quote:
Well, that's what I was saying. You were saying that it was LESS expensive to implement Express.
I'm saying a PCIe x1 slot is cheaper to implement than a PCI-X slot.
Quote:
That would be a baaddd idea. Keeping ISA, and now PCI on new machines just holds developement back. It was the ISA boards, for the most part that prevented Plug-n-Play from working.
a) I wasn't advocating it, I was just saying board makers wouldn't necesarrily have been "stuck" with PCI-X and PCI-X alone.
b) From the software's point of view, PCI-X and PCIe are both just "really fast PCI", so the ISA analogy is flawed.
In any case, this is getting rather tiresome, and I'm about to head off & enjoy my weekend. Guess we can ust agree to disagree...
Meh, at some point you would have to release your work at which point Apple would change their anti=piracy strategy and we would all be back to square 1.
There's really no point in delaying releases of such hacks. All you do is delay how long people will be able to use the current versions. The next version will always be incompatible and we start the whole process over again.
PSP ROM 2.0 anyone?
EXACTLY!
Apple, of course, knows this (i. e. "release, revise, repeat as necessary"). The hackerz think their playing their own game, when in fact their playing right into Apple's game plan! The OSX86 hackerz are really in fact just iToolz. You know the hackerz should be really doing this on some kinda darknet, maybe they are? But, by the time Apple releases production MacTels, my gut tells me that the hackerz will need to take a soldering gun to a generic PC motherboard to get their OS'z, appz, and warez to work. Yeah, they may get things to work, but will it ever be mainstream? That's my whole point, and I think it will be Apple's also.
In fact, what's to stop Intel from making "slightly tweaked" versions of their x86 CPU's? Intel's making literally 100's now? Devote a few 1000's transistors on these Apple specific CPU's that REQUIRE an Apple custom IC (they would of course be working closely together). Apple has a sole supplier of said custom IC's (top secret and with severe penalties for any contract violations), and it's GAME OVER!
And to fuel this futile hackerz discussion further, the SW dev cycle and Apple's HW/SW cycle will go hand-in-hand. Introduce low end HW (i. e. iBook) WITH low end native SW. See what happens. Revise said HW/SW (Mac mini released also), release, repeat as necessary. Release PB several months later with this learning curve incorporated and with a FEW high end apps. See what happens. Several months later, release PM's (with Leopard?), revised high end apps (and perhaps a few more), and see what happens. Now at this stage major 3rd parties begin to release their high end SW. So by 2007 we'll have secure HW and SW and the transition hits critical mass, and in 2008 Microsoft files for Chapter 11 .
In fact, I'm feeling so good about this scenario (except for the very last part), that instead of getting that quad PM, I think I'll invest in APPL. After all, I'd rather make mo money than that quad PM will ever give me!
...But, by the time Apple releases production MacTels, my gut tells me that the hackerz will need to take a soldering gun to a generic PC motherboard to get their OS'z, appz, and warez to work. Yeah, they may get things to work, but will it ever be mainstream?...
wouldn't be the first time it has been done, I remember that you could buy boards to turn an Amega into a Mac, but you had to provide the ROM's to make it work. With computers at a much lower cost, and Refirbs and used computers more prevalent I could see a day when some company could offer an "altered" motherboard with Mac ROM's (or whatever) presoldered to them, or even making a "card" that holds them. If Apple is slow to adopt faster chips or leaves a product sector blank then I could definatly see this happening, and upgrade manufacturers have found some very inventive ways to get new buisness from "non'upgradable" Macs in the past.
wouldn't be the first time it has been done, I remember that you could buy boards to turn an Amega into a Mac, but you had to provide the ROM's to make it work. With computers at a much lower cost, and Refirbs and used computers more prevalent I could see a day when some company could offer an "altered" motherboard with Mac ROM's (or whatever) presoldered to them, or even making a "card" that holds them. If Apple is slow to adopt faster chips or leaves a product sector blank then I could definatly see this happening, and upgrade manufacturers have found some very inventive ways to get new buisness from "non'upgradable" Macs in the past.
Ain't going to happen! Just because something has happened in the past, doesn't automatically imply that things will be the same going forward. This is (figuratively) just like starting over, with all of the previous history KNOWN! Now maybe this could happen in say the PRC where this kind of thing is SOP. But Apple legal would be so all over this kind of scenario, that it'll never happen here (or much of the "western" world for that matter). You know "The rule of law."
Don't worry, RazzFazz, you're not the first one who's given up arguing with melgross.
Melgross is relentless. He's a messageboard terminator T3000 liquid poly superbit alloy. Better bring your hard hat.
I'm really hoping that an Intel Mac mini is announced. For once in my life I'm willing to buy a 1st gen product. Hell all I need is school stuff. Then I'll be looking at a Powerbook or Powermac next. Cool stuff.
I bought a 1st gen G4 desktop the moment I saw Steve Jobs display it on a pedestal, like some kind of shining crystal. I got the pre-downgrade price on the 450Mhz model. Didn't regret it one bit.
Apple will have less wiggle room for 1st gen disasters given the architecture is basically Intel's. How could they screw that up?
Comments
Originally posted by Mandricard
Will X, assuming it is fully optimized for the chip, run at a par at the OS level (e.g. copying, finder-tasks, etc.)? Third-party software will, it seems to me, be a grab-bag; but the OS's are what I am thinking about here....
That's not exactly taxing on the CPU though. The gripe with OSX is usually it's 3D graphics implementation be it the drivers or OpenGL. It'd be quite embarrassing for Apple if, for example, DOOM3 or Cinebench still ran much faster on Windows than on the exact same hardware (or as near as) under OSX.
Originally posted by melgross
DDR 2 speeds relate to the argument because they point up an area in which Apple could have easily made a necessary improvement, but didn't. There would have been practically zero cost differential. Neither of us knows why they didn't implement this one either.
What i've been arguing all along is that, in the case of the G5 FSB, Apple could not easily have made such an improvement.
Of course I read what you wrote. But you are wrong. If you look at APPLE's bus speeds on the G5 PM's over the years, (I'm not interested in what Intel has been doing because Apple has a completely different bus and controller design), you will see that they scale perfectly with the cpu speeds.
It's not a matter of "what Intel has been doing", it's a matter of what the trend is for the entire industry -- or at least for those parts of the industry which experienced significant clock rate increases at all.
You make the assumption, with no reason to, that at some point in time there will be some disconnect between bus and cpu speeds. There has been no evidence for that. If anything the evidence is quite the opposite.
There has been no evidence if you restrict your point of view exclusively to the few iterations of the G5 that have shipped so far (and went from 2 GHz top to 2.7 GHz tops in the process). What makes you think that Apple / IBM can somehow magically just scale the FSB speed as they please when everybody else is unable to do so?
PCI is a 5 volt logic powered board, and PCI X is either a 3.3 or 3.5 (I forget which right now) volt logic powered board. 5 volt boards won't work on PCI X, and 3+ volt boards won't work on PCI. Some boards can use either.
It goes like this:
PCI-X cards and slots use 3.3V signalling.
For plain PCI, cards can use 3.3V signalling, 5V signalling or they can be "universal" (i.e. support both). The variants can easily be distinguished by looking at the position of the notch(es) in the PCI connector. PCI slots, on the other hand, can either support 5V signalling (most PC boards) or 3.3V (newer boards, and also slots running at 66MHz), but not both. The slots are also coded such that only cards that support the signalling voltage used by the slot physically fit.
PCI X isn't just about AGP.
Right. In fact, I don't think there are PCI-X graphics cards at all (or at least they're extremely rare).
Express serves no purpose for the majority of users either. The 16x slot is needed only by those buying the highest performing boards, and the rest of the slots are at least twice as fast as the slots in PCI X, which in turn are twice as fast as those in PCI.
Actually, compared to the PCI slots used in almost all PC motherboards (32 bit, 33 MHz) or on G4 Macs (64 bit, 33 MHz), PCI-X (64 bit, up to 133 MHz) is four or eight times as fast, respectively. (Of course, that's just comparing theoretical maximum numbers, not real-world throughput.)
It's odd that you would say that, because the opinion stated many times over the years leading up to the introduction, was that Express would be more expensive than PCI X. That was one of the reasons PCI X was developed, and one of the reasons given in public statements.
It's more expensive in that, due to lack of physical backwards compatibility, it obsoletes nearly all the expansion cards out there, forcing users to buy new ones. Since this is not something that's going to go too well especially with the PC world, you need to provide some PCI slots as well for backwards comptibility, and thus need a bridge chip, additional PCB real estate, etc. -- which is not the case for PCI-X. (Also, note that I was talking about x1 slots, not x4/x8/x16 ones.)
Everything electronic gets cheaper over the years, and sometimes by the time something is introduced, the price has dropped past expectations. Even Intel admitted that Express would be more expensive than PCI X. And I'll state again that the reason PCI X wasn't adopted wasn't because of its expense, but because Intel makes at least 50% of the mobo's used by manufacturers, and they were afraid of getting stuck with something that Intel's boards would swamp. It was politics, and the fact that the PCI X manufacturers had no clear upgrade path from version one.
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
Originally posted by melgross
Apple just went to DDR2 533. Why? It offers little advantage. Even cheap PC mobo's are using 667 speeds. There are even 800 speed boards coming out.
Originally posted by RazzFazz
I don't quite see how this really relates to our argument -- it appears Apple felt it didn't make business sense for them to go with the higher speed DDR2. That, however, doesn't in any way imply that all their decisions are made purely for political / business reasons.
I assume the memory controller cannot handle the 4 DIMMs at the higher speed. In PCs the memory controller can often handle 2 DIMMs at a lower speed and only 1 at higher speed. Or has to use registered DIMMs instead.
(Of course this is always number of DIMMs per memory bank).
Originally posted by RazzFazz
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
Actually, the way I recall it, the MSI Neo4 Platinum board I purchased a few months ago had one PCI x16, x4 and x1 port each, three PCI ports and finally one PCI-X port. And this is a fairly mainstream choice and not too costly either. SO I think that combining PCI, PCIe and PCI-X is in fact done by quite a few manufacturers.
Originally posted by RazzFazz
What i've been arguing all along is that, in the case of the G5 FSB, Apple could not easily have made such an improvement.
It's not a matter of "what Intel has been doing", it's a matter of what the trend is for the entire industry -- or at least for those parts of the industry which experienced significant clock rate increases at all.
There has been no evidence if you restrict your point of view exclusively to the few iterations of the G5 that have shipped so far (and went from 2 GHz top to 2.7 GHz tops in the process). What makes you think that Apple / IBM can somehow magically just scale the FSB speed as they please when everybody else is unable to do so?
I think I realise what the problem here is.
You keep talking about multiple GHz bus's. But that's not the question at all.
Ok, I quess we do have to bring Intel into this.
Intel's memory bus's have gone from 400MHz to 1.066GHz in 2 1/2 years. The processor speeds have gone from slightly under 3GHZ to 3.8 GHz, and back down to 3.4 GHz during that same time period.
So, bus speeds have gone up vastly faster than chip speeds have. The opposite of what you said earlier.
So let's see what the situation really is.
Intel has a bus that is bidirectional, but can only pass a signal one way at a time. But when the signal does pass through, it does at the full 1.066GHZ speed on their newest bus.
Apple has a bus that at the present time runs at 1.25GHz, down from 1.35GHz on the dual 2.7.
BUT, this bus consists of two seperate bus's. Each one is unidirectional. And each one operates at half the speed of the overall bus speed. The Apple bus at 1.25GHz can pass more information over the bus at one time than Intel's 1.066GHz bus can.
But, each bus is only running at 625MHz, much slower than the 1.066GHZ that the traffic runs at in Intel's bus in either direction.
If Apple went to a 2GHz bus for the dual core 2GHZ machine, each unidirectional bus would only be running at 1GHz. Slightly slower than Intel's.
If Apple used it on the dual core 2.3GHz, it would be running at 1.15GHz, slightly faster than Intel's.
If they ran it at 2.5 GHz on the quad, each side of the bus would be running at 1.25GHz. That's just under 200MHz faster than Intel's.
None of Apple's bus's would be running at multiples of 1GHz as you've been saying, and been so concerned about.
So we are not talking about these high speeds that you are thinking about. Just slightly faster than what Intel has now.
Quote:
It goes like this:
PCI-X cards and slots use 3.3V signalling.
For plain PCI, cards can use 3.3V signalling, 5V signalling or they can be "universal" (i.e. support both). The variants can easily be distinguished by looking at the position of the notch(es) in the PCI connector. PCI slots, on the other hand, can either support 5V signalling (most PC boards) or 3.3V (newer boards, and also slots running at 66MHz), but not both. The slots are also coded such that only cards that support the signalling voltage used by the slot physically fit.
Sorry, but PCI boards are 5 volt, and PCI X boards are 3.3 volt.
If a card supports both, then that card can run in either because the card senses which voltage the bus is using.
Quote:
Right. In fact, I don't think there are PCI-X graphics cards at all (or at least they're extremely rare).
I quess it's sad to know that all of the G5 PM's have been running without graphics cards. We should all run out and tell ATI and Nvidia, so that they can make us some. Here, you are just guessing. I can't imagine how you came up with this one. We are talking 8x AGP in either PCI or PCI X. Same thing electrically.
Quote:
Actually, compared to the PCI slots used in almost all PC motherboards (32 bit, 33 MHz) or on G4 Macs (64 bit, 33 MHz), PCI-X (64 bit, up to 133 MHz) is four or eight times as fast, respectively. (Of course, that's just comparing theoretical maximum numbers, not real-world throughput.)
Yes, that's what I said, PCI X is up to twice as fast as PCI, and Express is up to twice as fast as PCI X.
Quote:
It's more expensive in that, due to lack of physical backwards compatibility, it obsoletes nearly all the expansion cards out there, forcing users to buy new ones. Since this is not something that's going to go too well especially with the PC world, you need to provide some PCI slots as well for backwards comptibility, and thus need a bridge chip, additional PCB real estate, etc. -- which is not the case for PCI-X. (Also, note that I was talking about x1 slots, not x4/x8/x16 ones.)
Well, that's what I was saying. You were saying that it was LESS expensive to implement Express.
Quote:
There's nothing that would prevent motherboard manufacturers from offering PCI-X and PCIe slots on the same board (like they do right now with PCIe and plain old PCI) -- well, nothing except there's really not much of a business case for doing so. Also, in the areas that see actual benefit from PCI-X, i.e. the server and high-end workstation market, PCI-X was adopted.
That would be a baaddd idea. Keeping ISA, and now PCI on new machines just holds developement back. It was the ISA boards, for the most part that prevented Plug-n-Play from working.
A couple of quotes from Anand, from Anandtech:
"It is irritating that Apple didn't move to DDR2-667 yet, especially on their highest end configuration (and especially because it can use the bandwidth), but given Apple's relatively conservative nature whenever it comes to memory speeds it isn't a huge surprise."
" In my opinion, the biggest improvement to the new G5s is the move to PCI Express. And here's one thing I really do like about Apple, when they move to a new technology, they really move to it.
There isn't a single parallel PCI slot in the new G5s,..."
Originally posted by smalM
I assume the memory controller cannot handle the 4 DIMMs at the higher speed. In PCs the memory controller can often handle 2 DIMMs at a lower speed and only 1 at higher speed. Or has to use registered DIMMs instead.
(Of course this is always number of DIMMs per memory bank).
The controller can handle it. Read the quotes at the end of my above post.
DDR2 for powerbooks and 2006 macintels makes sense as intel is standardising to DDR2 support, and DDR2 in portables show reduced power consumption (by how much, well, that's a matter for debate).
according to most PC sites lower-latency but slower speed DDR amd64s vs faster speed but higher-latency DDR2 intels results in very neck and neck comparisons, wrt to memory specific synthetic benchmarks (sandrasoft)
i wonder in the case of the quad g5, if the memory controller is "teh bomb", then in this case DDR with decently tight timings will do fine, DDR2 may not offer any advantage, we may not see powermac g5s with DDR2 until middle of 2006.
Originally posted by sunilraman
while we're on the topic.. (wait, what was the topic again??)
DDR2 for powerbooks and 2006 macintels makes sense as intel is standardising to DDR2 support, and DDR2 in portables show reduced power consumption (by how much, well, that's a matter for debate).
according to most PC sites lower-latency but slower speed DDR amd64s vs faster speed but higher-latency DDR2 intels results in very neck and neck comparisons, wrt to memory specific synthetic benchmarks (sandrasoft)
i wonder in the case of the quad g5, if the memory controller is "teh bomb", then in this case DDR with decently tight timings will do fine, DDR2 may not offer any advantage, we may not see powermac g5s with DDR2 until middle of 2006.
Sunil, I really hate to break this to you
Originally posted by melgross
Sunil, I really hate to break this to you
Yes that was a good one.
BTW DDR 2 has an higher latency than DDR, but the memory bandwitch is sligthy superior.
In conjonction or because of a greater L2 cache, the memory performance of the new G5 are superior than the older.
Originally posted by melgross
Sunil, I really hate to break this to you
lol WTF i am stuck in AMD64 land
Originally posted by franksargent
Now if I were a l33t piratez, I just might keep my mouth shut UNTIL the production MacTel's are released, then and only then would I get the word out. Basic human nature, people talk, loose lips sink ships!
Meh, at some point you would have to release your work at which point Apple would change their anti=piracy strategy and we would all be back to square 1.
There's really no point in delaying releases of such hacks. All you do is delay how long people will be able to use the current versions. The next version will always be incompatible and we start the whole process over again.
PSP ROM 2.0 anyone?
Originally posted by melgross
I think I realise what the problem here is.
You keep talking about multiple GHz bus's. But that's not the question at all.
Ok, I quess we do have to bring Intel into this.
Intel's memory bus's have gone from 400MHz to 1.066GHz in 2 1/2 years. The processor speeds have gone from slightly under 3GHZ to 3.8 GHz, and back down to 3.4 GHz during that same time period.
I do think I put the "before the great brick wall was hit" qualifier in my statement? I was talking about the long run.
So, bus speeds have gone up vastly faster than chip speeds have. The opposite of what you said earlier.
Up to and including the 486DX, processor and FSB were at 1:1. Then came the DX2, DX4, the Socket 7 processors, whatnot. With each generation, the FSB multiplier grew on average. It wasn't until recently that CPU frequency just kinda stopped scaling.
So let's see what the situation really is.
Intel has a bus that is bidirectional, but can only pass a signal one way at a time. But when the signal does pass through, it does at the full 1.066GHZ speed on their newest bus.
No, it does not. Intel's FSB is quad-pumped, the actual clock rate is 266MHz.
Apple has a bus that at the present time runs at 1.25GHz, down from 1.35GHz on the dual 2.7.
The G5's EI is double-pumped, so the actual clock rate is half of what you state.
BUT, this bus consists of two seperate bus's. Each one is unidirectional. And each one operates at half the speed of the overall bus speed. The Apple bus at 1.25GHz can pass more information over the bus at one time than Intel's 1.066GHz bus can.
But, each bus is only running at 625MHz, much slower than the 1.066GHZ that the traffic runs at in Intel's bus in either direction.
If Apple went to a 2GHz bus for the dual core 2GHZ machine, each unidirectional bus would only be running at 1GHz. Slightly slower than Intel's.
You're really confusing something here. "Operates at half the speed" does not mean "each direction operates at half the clock rate" (that doesn't even make sense). A "2 GHz bus" runs at, well, 2 GHz. Not "1 GHz per direction" or some such. The thing that's halved per-direction is the bus's width.
The math for the current incarnation of the G5's FSB goes like this:
You have two unidirectional 32-bit buses, clocked at 4:1 but double-pumped so it's effectively 2:1. That gives you about 750MHz x 2 (DDR) x 32 bits/direction x 2 directions = 12GBps aggregate throughput.
Intel's current FSB is a bidirectional 64-bit bus that runs at 266 MHz. It's quad-pumped, so the effective clock rate is 1066 MHz. Thus, you get 266 MHz x 4 (QDR) x 64 bits = 8.5GBps aggregate throughput.
If Apple used it on the dual core 2.3GHz, it would be running at 1.15GHz, slightly faster than Intel's.
If they ran it at 2.5 GHz on the quad, each side of the bus would be running at 1.25GHz. That's just under 200MHz faster than Intel's.
You should really drop the whole "each side runs at clock rate x, so the whole bus runs at twice that clock rate" argument. It's the throughput that doubles, not the clock rate. If one lane runs at 1.15GHz, then two such lanes still run at the same clock speed, the whole bus is just wider now.
[QUOTE]
Sorry, but PCI boards are 5 volt, and PCI X boards are 3.3 volt.
If a card supports both, then that card can run in either because the card senses which voltage the bus is using.
[/QUOTE
This is getting fairly annoying. The PCI spec does allow for 3.3V and 5V slots, and 3.3V, 5V and "universal" cards. Go ask Google, Wikipedia or whatever other source you like.
I quess it's sad to know that all of the G5 PM's have been running without graphics cards. We should all run out and tell ATI and Nvidia, so that they can make us some. Here, you are just guessing. I can't imagine how you came up with this one. We are talking 8x AGP in either PCI or PCI X. Same thing electrically.
AGP in PCI/PCI-X? What are you even talking about? So, for you, AGP is just a different physical slot form factor and otherwise just PCI(-X)? If that really is what you're suggesting, I suggest you consult any kind of source you'd like, and you'll see that AGP uses different voltages, uses different signal speeds, doesn't have muxed A/D lines, etc., pp.
Also, make sure you take a good look at the Apple hardware tech docs -- the AGP port is attached directly to U3 (i.e. the north bridge), whereas PCI-X is provided by a HT-to-PCI-X-Tunnel that's attached to the HyperTransport port. There isn't even a direct connection between the two.
Well, that's what I was saying. You were saying that it was LESS expensive to implement Express.
I'm saying a PCIe x1 slot is cheaper to implement than a PCI-X slot.
That would be a baaddd idea. Keeping ISA, and now PCI on new machines just holds developement back. It was the ISA boards, for the most part that prevented Plug-n-Play from working.
a) I wasn't advocating it, I was just saying board makers wouldn't necesarrily have been "stuck" with PCI-X and PCI-X alone.
b) From the software's point of view, PCI-X and PCIe are both just "really fast PCI", so the ISA analogy is flawed.
In any case, this is getting rather tiresome, and I'm about to head off & enjoy my weekend. Guess we can ust agree to disagree...
This has mostly been useless, we are obviously talking past each other.
I was just going to bed myself.
Have a nice weekend.
Originally posted by strobe
Meh, at some point you would have to release your work at which point Apple would change their anti=piracy strategy and we would all be back to square 1.
There's really no point in delaying releases of such hacks. All you do is delay how long people will be able to use the current versions. The next version will always be incompatible and we start the whole process over again.
PSP ROM 2.0 anyone?
EXACTLY!
Apple, of course, knows this (i. e. "release, revise, repeat as necessary"). The hackerz think their playing their own game, when in fact their playing right into Apple's game plan! The OSX86 hackerz are really in fact just iToolz. You know the hackerz should be really doing this on some kinda darknet, maybe they are? But, by the time Apple releases production MacTels, my gut tells me that the hackerz will need to take a soldering gun to a generic PC motherboard to get their OS'z, appz, and warez to work. Yeah, they may get things to work, but will it ever be mainstream? That's my whole point, and I think it will be Apple's also.
In fact, what's to stop Intel from making "slightly tweaked" versions of their x86 CPU's? Intel's making literally 100's now? Devote a few 1000's transistors on these Apple specific CPU's that REQUIRE an Apple custom IC (they would of course be working closely together). Apple has a sole supplier of said custom IC's (top secret and with severe penalties for any contract violations), and it's GAME OVER!
And to fuel this futile hackerz discussion further, the SW dev cycle and Apple's HW/SW cycle will go hand-in-hand. Introduce low end HW (i. e. iBook) WITH low end native SW. See what happens. Revise said HW/SW (Mac mini released also), release, repeat as necessary. Release PB several months later with this learning curve incorporated and with a FEW high end apps. See what happens. Several months later, release PM's (with Leopard?), revised high end apps (and perhaps a few more), and see what happens. Now at this stage major 3rd parties begin to release their high end SW. So by 2007 we'll have secure HW and SW and the transition hits critical mass, and in 2008 Microsoft files for Chapter 11
In fact, I'm feeling so good about this scenario (except for the very last part), that instead of getting that quad PM, I think I'll invest in APPL. After all, I'd rather make mo money than that quad PM will ever give me!
Originally posted by franksargent
...But, by the time Apple releases production MacTels, my gut tells me that the hackerz will need to take a soldering gun to a generic PC motherboard to get their OS'z, appz, and warez to work. Yeah, they may get things to work, but will it ever be mainstream?...
wouldn't be the first time it has been done, I remember that you could buy boards to turn an Amega into a Mac, but you had to provide the ROM's to make it work. With computers at a much lower cost, and Refirbs and used computers more prevalent I could see a day when some company could offer an "altered" motherboard with Mac ROM's (or whatever) presoldered to them, or even making a "card" that holds them. If Apple is slow to adopt faster chips or leaves a product sector blank then I could definatly see this happening, and upgrade manufacturers have found some very inventive ways to get new buisness from "non'upgradable" Macs in the past.
Originally posted by JCG
wouldn't be the first time it has been done, I remember that you could buy boards to turn an Amega into a Mac, but you had to provide the ROM's to make it work. With computers at a much lower cost, and Refirbs and used computers more prevalent I could see a day when some company could offer an "altered" motherboard with Mac ROM's (or whatever) presoldered to them, or even making a "card" that holds them. If Apple is slow to adopt faster chips or leaves a product sector blank then I could definatly see this happening, and upgrade manufacturers have found some very inventive ways to get new buisness from "non'upgradable" Macs in the past.
Ain't going to happen! Just because something has happened in the past, doesn't automatically imply that things will be the same going forward. This is (figuratively) just like starting over, with all of the previous history KNOWN! Now maybe this could happen in say the PRC where this kind of thing is SOP. But Apple legal would be so all over this kind of scenario, that it'll never happen here (or much of the "western" world for that matter). You know "The rule of law."
Originally posted by RazzFazz
AGP in PCI/PCI-X? What are you even talking about? ...
In any case, this is getting rather tiresome, and I'm about to head off & enjoy my weekend. Guess we can ust agree to disagree...
Don't worry, RazzFazz, you're not the first one who's given up arguing with melgross.
Originally posted by bikertwin
Don't worry, RazzFazz, you're not the first one who's given up arguing with melgross.
Melgross is relentless. He's a messageboard terminator T3000 liquid poly superbit alloy. Better bring your hard hat.
I'm really hoping that an Intel Mac mini is announced. For once in my life I'm willing to buy a 1st gen product. Hell all I need is school stuff. Then I'll be looking at a Powerbook or Powermac next. Cool stuff.
Apple will have less wiggle room for 1st gen disasters given the architecture is basically Intel's. How could they screw that up?
Originally posted by hmurchison
[B]Melgross is relentless. He's a messageboard terminator T3000 liquid poly superbit alloy. Better bring your hard hat.