Motorola MPC7457 Summary

[Bregalad]

Re: anxious

Originally posted by kenohki


WTF? "Agressiveness of your memory system and it's latency"? You can't adjust that. First off, latency of memory is not adjustable. It just depends on what you buy, whether it's DRAM, SDRAM, MoSys 1T-SRAM or whatever. Memory latency is determined by the materials and build of the actual memory. And I'm not sure what you're talking about by "agressiveness" but I'm assuming it's something to do with paging but could you expand a little on this?

What the original poster meant was that no matter what quality of RAM you buy for your Mac it'll still run at the slowest CAS 3 timings (2.5 in the case of the new DDR models). On the PC sticking high end CAS 2 5-2-2 1T PC2700 DDR into your PC and adjusting the memory settings in the BIOS accordingly will yield a measurable performance improvement over standard CAS 2.5 stuff.

For the overclockers there is an additional benefit to buying expensive RAM. I read a test report in which a 1.6GHz Pentium 4 that ran at 2.2GHz using ordinary DDR was able to run at 2.5GHz using top of the line PC3200.

Essentially this all boils down to choice. Whereas there are definite compatibility advantages to Apple's control of the entire widget, the sheer number of choices in the PC world results in rapid technological advances and, at the same time, very aggressive prices.

You could spend $30 on a PC case complete with a 300W power supply. You could also spend $300 on the case alone. The same holds true for many other components. It's not difficult to build a PC with very good components and a fit and finish quality that rivals that of a Mac. Notice I said build. Buying a PC off the shelf means you have to accept the component choices and manufacturing quality of whoever had the lowest bid.

PC motherboard manufacturers have to compete with one another on features like built in FireWire and ATA-133 RAID. They also compete on BIOS features like automatic updates via the internet and the ability to adjust the timing of virtually everything. The result is great for consumers. Not only are prices very low, but features are often available before you can even make use of them. Serial ATA is the latest example while FireWire is probably the only case of the PC world being behind Apple.

Finally there is one other thing that will always make building a PC, whether it be a $300 peecee of crap or a $2000 gamer's ultimate, much less expensive than a Mac: OEM parts. The serious PC builder who selects each component individually does not need the "nice" features that come with a retail box so she is able to save a lot of money getting the same hardware without the glitz.

When Mac clones and CHRP were announced people had visions of building their own Mac using the best parts they could afford. That dream was soon shattered and today we're back where we were in 1995 paying through the nose for an operating system and a better than average case design. Oh and Steve, why did you put those ridiculously loud, 60mm, high rpm fans in the new MDD case? I thought you wanted Macs to be seen and not heard.

Lest you think I'm a PC weenie, I own 4 Macs and have been running OS X since developer preview 2.

 

[g3ski]

2Ghz should be here.....

come on now folks, Mhz myth or not, you all would prefer to have a dual G5/2.8Ghz system right now.

And for processing video and 3D, mhz do make a difference. The fastest graphics boxes run Intel and AMD chips. Macs are just not there. This is not my opinion, it's not PC hype, and it's not Mhz myth. It's a fact. Our Macs at 1.25Ghz cannot compete with 2.8Ghz PCs with 533Mhz front side buses....and it's sad, sad, sad.

Blame Motorola for this issue, once again, just like when we were stuck at 500Mhz for 18 months. Apple could only PR their way out of the Motorola mess. Blame Motorola, not someone on this forum who wants a faster mac....all they want is a faster mac.

We need a faster front side bus on the PPC 74XX/75XX chips!

We need new chips....give me G5 and G6 in volume or the IBM 970!

 

[locovaca]

Re: Bus width is not fair comparison

Originally posted by Frobozz
"166 mhz bus (slower than either an average Athlon (266) or P4 (400)): Advantage PC."

This does not give the PC an advantage. It _can_ make it faster in some circumstances, but it's not a simple matter of 400 MHz bus being 3 times faster than our 166. The PC, in the case of a 400 MHz bus, is really only supplying 3 or 4 buses at lower bandwidths-- like 100 Mhz x 4. Because of how a PC is built, EVERYTHING goes through the system bus in both directions. So in the PC's case, 167 MHz a single direction, and adding another 167 for the other direction, for a total of 333. In most cases on the Mac, the CPU, memory, and hard drive/device controllers can bidirectionally talk. This means that in most cases a 167 bus Mac is as fast as a 333 Mhz bus PC. The only Caveat is the amount of information it can send at that speed. The PC will have a slight advantage because of the through-put (more GB/sec), but it's through a smaller pipe. Keep in mind that a Mac can have it's memory talk to the CPU in a more direct manner, so these types of things are not as much of an issue on the Mac as the are for a PC.

I read most of this on other rumors sites, so if there are small errors I apologize in advance.

Not to come off as a pompous ass or anything, but I really suggest that you take a computer engineering class instead of taking what someone dumbed down from what someone dumbed down from a similar situation and posted it. Not to be arrogant, but to prove that what I'm about to say is not bull****- I am currently a Senior in Computer Engineering at Iowa State University. I have taken classes on both computer architecture and processor design, as well as assembly programming and interrupt programming. We do most of our assembly on PowerPC processors, for what it's worth.

A bus transfers data. It goes in one way only at one time- there is no such thing as a bi-directional bus that transfers both ways at the same time- that would be two busses. A Pmac does this at a speed of 167 MHz. It does one thing during each cycle- it transfers data from the cpu to the system controller or from the system controller to the cpu, no more. An Athlon transfers does this very same thing at 133 MHz (soon to be 167 MHz), but it has the ability to transfer twice during a cycle- this is called DDR. So, this does twice the amount of transfers in a clock cycle in compared to the G4 bus. A P4 does this very samet hing at 100 MHz (and is now becoming 133 MHz), but it has the ability to transfer four times during the cycle- this is called QDR. So, it does four times the transfers in a clock cycle as compared to the G4 bus, or 2 times the transfers in a clock cycle as compared to the Athlon. I am not trying to twist facts in anyway. The G4 has the smallest "pipe." The G4 bus is no more efficient than the Athlon or P4 bus. Prior to the 167 MHz bus, the G4s had the same bus (size and speed wise) as the Pentium 3. The size of the busses are:

G4: 64 bit
Athlon: 64 bit
Pentium 4: 64 bit

So, bus speeds are:

G4: 64 (bits/cycle) * (1/8) (byte/bits) * 167 (cycles/second) * 1 (SDR)= 1336 MB/sec
Athlon: 64 (bits/cycle) * (1/8) (byte/bits) * 133 (cycles/second) * 2 (DDR)= 2128 MB/sec
P4: 64 (bits/cycle) * (1/8) (byte/bits) * 100 (cycles/second) * 4 (QDR)= 3200 MB/sec


A hard drive, graphics card, firewire device, usb device, keyboard, mouse, sound card, any kind of device, when it needs to transfer data, will raise an interupt with the processor. This is true of any processor in existence today with very few exceptions (and those are integrated processors that only do one function, like those in cars, and they work by polling to see when data needs to be moved around). When an interrupt is raised, the processor stops what it is doing, saves what it was doing to memory, services the interrupt, and then resumes processing. Writing to memory is a very time intensive process (relatively speaking), one that is dependant on both the FSB and memory speed. While the Pmac may have DDR 333 memory, it still has a piddly 167 MHz bus which is the bottleneck. The only difference between a pc and a mac is that the most PC chipsets are composed of two chips, with one controlling the HDD, USB, Firewire, Ethernet, etc. and the other the AGP, Memory, and CPU. That said, it does not incur any penalty versus the mac when performing CPU to Memory operations because it is still only going through one additional processing unit.


So, in retrospect, the biggest bottleneck in today's computing is... the memory bus. Memory accesses are very expensive in processors (which is why programs that run in cache run very quickly) regardless of what platform it is, and only by removing the bottlenecks between the processor and the memory can you improve performance greatly.

 

[g3ski]

533 FSB?

Great info.....

But the Pentium 4 has a 533Mhz FSB, not just a 400Mhz bus. Where is that extra 133Mhz going to, coming from?

What is the difference between the FSB and the memory bus?

 

[Telomar]

Re: Re: Bus width is not fair comparison

Originally posted by locovaca


Not to come off as a pompous ass or anything, but I really suggest that you take a computer engineering class instead of taking what someone dumbed down from what someone dumbed down from a similar situation and posted it.

Blah blah blah...

Just to correct a few things. The G4, Athlon and PIV all use bidirectional buses. Data can be passed in either direction along the bus, albeit not necessarily at the same time.

A unidirectional bus is something akin to what the PPC970 uses. It uses two uni-directional buses. One streams in each direction. One of the major advantages being that it allows streaming simultaneously in both directions and lowers overhead.

Second you will find the bus of the G4 is considerably more efficient that those of the Athlon or PIV. The design of it is such that the overhead is lower and efficiency greater. Unfortuantely it has the lowest theoretical bandwidth of the lot, which makes it a weak point in the desktop G4.

One of the reasons you will never see a DDR MPX bus is because to do so would mean trading off some efficiency (along with some other benefits).

Finally certain parts of the computer rely on DMA. Bandwidth from CPU to memory is not necessarily the limiting factor in those cases. This is one of the reasons that the Xserve can be a very good file server.

 

[locovaca]

Re: 533 FSB?

Originally posted by g3ski
Great info.....

But the Pentium 4 has a 533Mhz FSB, not just a 400Mhz bus. Where is that extra 133Mhz going to, coming from?

What is the difference between the FSB and the memory bus?

Up until recently the P4 had a 400 mhz FSB and a 533 mhz memory clock- so it went to the same place the extra 167 mhz goes on a mac- bandwidth for other devices (which, admittidly, doesn't help very much because they still depend on the CPU for most things, and it's still stuck on that 400/167 mhz bus). Wehn Intel releases their 3.06 they will officially be moving to a 533 mhz FSB and run it at the same speed as the memory. That said, it's possible to run the memory bus and the FSB at same speeds, since they are one chip removed from each other, but the benefits are very small.

The FSB is the bus that the CPU communicates with the system controller/north bridge with. Anything that a CPU touches comes through this thing. In P4's and Macs there is only one FSB for whatever number of CPUS you throw at it (so, a dual processor doesn't have dual FSB)- the Athlon has a FSB per cpu (which isn't as grand as it would sound, but it does help).

The Memory Bus is what connects the memory subsystem to the system controller/north bridge. It operates at whatever speed your memory is set at, and while this usually runs at the same speed as the FSB, it isn't required to. The reason why it usually runs at the same speed is because there are few benefits to not running it at the same speed: if it's slower than the FSB then the processor will always be waiting on memory, and if it's faster then the memory will be waiting on the FSB. The only thing that gets sped up if the memory is faster are devices that do direct memory writes, however, even these still require CPU interaction, so they don't get a magical speedup of 2x just because there's all this extra bandwidth lying around. The memory bus doesn't even have to be the same width as the CPU- in the P4 (RAMBUS) there are actually two 16 bit memory busses or a 64 bit bus (SDR, DDR), although there are dual ddr busses coming out soon. With the Athlon this is typically just 64 bits, although NVIDIA's chipset can use a 128 bit bus (because of their integrated graphics, one of the few things that can effectively make use of extra bandwidth. An integrated graphics card that shares system memory reserves a portion of the memory for the card, and it gets exclusive rights to this memory, so it can operate on it without the CPU). The G4 has a 64 bit bus that runs at the same speed as the FSB, only this runs DDR.

So the main difference between the two is the fact that they are two seperate things with a chip in between them. While a majority of what travels on the FSB is just stuff from the memory bus, the FSB does much more stuff than the memory bus.

 

[mr evil brkfast]

Motorola Performance question

We shall see if an when motorola gets the chips out the door... I think 2003 will be a long year like this one in terms of upgrades.

The new 970 will hopefully be the savior in these dark times.

 

[Chryx]

Re: Re: 533 FSB?

Originally posted by locovaca
Up until recently the P4 had a 400 mhz FSB and a 533 mhz memory clock

Nope, PC800 Rdram clocks at 400Mhz, PC2100 clocks at 133Mhz, PC133 clocks at 133Mhz, it's not possible (to my knowledge) to use PC1066 RDram (533Mhz) with a 400Mhz-FSB P4

with the 2.4Ghz P4's introduction, they raised the FSB from 100 QDR, to 133 QDR (400 effective to 533 effective)