Just a matter of time... G4 @ 2ghz

[mgargan1]

Do programmers really concern themselves with the presence of the L3 cache? Isn't that something the OS writers would take care of when they design the memory manangement system? How are apps written specifically to take advantage of the caches?

I'm not a programmer, so I don't really know. But I do know that some apps take advantage of certain types of cache. For instance, certain game engines like cache in general, certain game engines like fast cache. The OS itself is configured for dual processor machines, yet some apps aren't. Applications have to be written to take advantage of everything, from the processors IPC, to its cache.

 

[Abstract]

The dual-core G4 in development supposedly is up to almost 2ghz with a 200mhz bus if I remember reading correctly on Freescale's website.

Anyone else think that the FSB on this dual core G4 is also too slow? I mean, 200 MHz isn't that much better than 133MHz or 167MHz. It just seems so bad when compared to what you see in G5s and Windows computers.

How much of an effect does this slow FSB have on performance if, despite the FSB, they manage to reach high clock speeds? I'm no expert at all. 😕

Maybe I will regret it when Apple or whoever bring out something that needs a G5. Ah well thats my thoughts. Does anyone else feel that people are going over the top about these G5's.

Smartest thing I've heard in a while. Who cares if its a G4 as long as its fast. It doesn't seem like the 64-bit G5, and the futureproof-ness of getting a G5 is worth it right now. It has been a while since the G5 came out, and the payoff of getting one isn't that great despite the FSB and clock speeds. I don't know why not, but it just seems as though a DP 1.42GHz G4 from several years ago delivers quite a punch.

 

[Little Endian]

It's been 5 and a half years since the first G4 74XX series processor shipped in a mac at 350Mhz. Today the 74XX series is up to 2Ghz!! Talk about design longevity the 74XX series may very well hold the record for the most overclocked desktop CPU to exist. Sure today's 7447 and 7457 CPUs have come along way design wise from the original 7400 and 7410 however they still reamain more or less the same CPU. 350Mhz to 2Ghz is almost a 6x increase in clockspeed, I can't really think of any CPU design that is as long lived and seen such large clock speed increases as the G4.

2Ghz seems wasted though especially considering lower clocked Chips with L3 cache can perform roughly the same. I suppose Gigadesigns released this overclocked beast just for the psychological marketing factor of having the highest clocked G4 chips. I think Motorola/Freescale should have incorporated the L3 cache into the 7447 as the 512Kb L2 cache does'nt come near to compensating a 1MB or 2MB L3 cache.

 

[Lord Blackadder]

Smartest thing I've heard in a while. Who cares if its a G4 as long as its fast. It doesn't seem like the 64-bit G5, and the futureproof-ness of getting a G5 is worth it right now. It has been a while since the G5 came out, and the payoff of getting one isn't that great despite the FSB and clock speeds. I don't know why not, but it just seems as though a DP 1.42GHz G4 from several years ago delivers quite a punch.

People get too caught up in the notion that the G4 is dead-end old tech. It isn't. I can easily imagine Freescale continuing to develop low-power/high performance G4 derivative CPUs for years, and Apple continuing to use them in their portable lines. Increase the clock and FSB and it will remain a competitive processor. Personally, I think that (in the short term) the hurdles Freescale is faced with improving the G4 to near G5 performance may be potentially smaller than the hurdles IBM faces getting the G5 into a laptop...

And little Endian is right, Giga is just doing the overclock for the bragging rights (I admit it is tempting), but I would be happier with a 1.467 with L3. A 400MHz overclock is a bit scary. I really wish Freescale would put that L3 cache back.

 

[ewinemiller]

Do programmers really concern themselves with the presence of the L3 cache? Isn't that something the OS writers would take care of when they design the memory manangement system? How are apps written specifically to take advantage of the caches?

I have seen some of this and have actually done some optimization myself where I've used prefetches so that my working dataset is always available in cache. You couldn't do too much prefetching because then you'd start pushing other stuff out of cache so the size of the cache was an important consideration in that optimization. I was targeting a 512k l2 cache PIII at the time.

As much of a voodoo quest as that was though (modify prefetches, render, time, rinse, repeat), I would be really surprised if someone went as far as to check the systems cache and then adjusted their behavior to take advantage of more cache. The original poster's point was correct that depending on what you are doing there may be performance penalties (or benefits) when comparing the same CPU with 512k of l2 cache vs. one with 256k l2 cache and some amount of L3. The PC side for the most part got rid of l3 cache a long time ago, but I remember when the PIIIs went from 256k to 512k, there was about a 10% boost in performance just for that if the chips were running at the same clockspeed.

Back to the original thread, I notice they have a dual 1.8 available. It might be time to upgrade yea ole dual 800. I will have to see what the next couple of months brings on in the G5 line.

 

[ewinemiller]

350Mhz to 2Ghz is almost a 6x increase in clockspeed, I can't really think of any CPU design that is as long lived and seen such large clock speed increases as the G4.

Um, PIII, that core started in the pentium pros at 150mhz and got to 1.4 ghz. 9.33x increase in clock speed, not sure about the time, but I think it took them less than 5 and half years (which is a good thing). You could argue that the Pentium Ms are still just a modified PIII, though the differences are pretty radical. They are at something like 2.1 ghz these days and if you're talking about overclocks I think I've seen those running up to 2.4.

 

[mgargan1]

Um, PIII, that core started in the pentium pros at 150mhz and got to 1.4 ghz. 9.33x increase in clock speed, not sure about the time, but I think it took them less than 5 and half years (which is a good thing). You could argue that the Pentium Ms are still just a modified PIII, though the differences are pretty radical. They are at something like 2.1 ghz these days and if you're talking about overclocks I think I've seen those running up to 2.4.

Intel came out with the Pentium Pro in 1995, and the 1.4GHz Pentium III in 2001/2002ish. Also don't forget that the 750 processor aka. G3 is the same core as the G4, the G4 just has altivec unit(s). And the G3 started at 233MHz i belive, not too sure of that.

 

[Little Endian]

Um, PIII, that core started in the pentium pros at 150mhz and got to 1.4 ghz. 9.33x increase in clock speed, not sure about the time, but I think it took them less than 5 and half years (which is a good thing). You could argue that the Pentium Ms are still just a modified PIII, though the differences are pretty radical. They are at something like 2.1 ghz these days and if you're talking about overclocks I think I've seen those running up to 2.4.

You are right about the Pentium III being the decendant of the Pentium Pro. However they would be considered two different families even though they share cores from the same roots. While the P III cores are simialar and originate from the Pentium Pro they have a greater differences in design than differences within the Motorola 74XX series chips to be considered the same chip.

If you want to count the Pentium Pro, PIII, and Centrino in the same family then I could go back to 1995 for the PowerPC chips as well.

As mgargan1 states: "G3 is the same core as the G4, the G4 just has altivec unit(s). And the G3 started at 233MHz i belive, not too sure of that."

That statement is true the G3 shares the same core design as the Alitvec enabled G4 and started at 233Mhz. I can go further to state that the G3 was essentially the same core as the PowerPC 603 and 603e chips that started at 75Mhz. The biggest difference from the G3 compared to the 603e was the G3 had a backside L2 and improved FPU other than that the core was virtually identical to the G3 which is the same as a G4 which has Alivec added and ability to use L3 cache.

Point is we can go far far back in design lineages but from a marketing standpoint and in many ways performance standpoint we can't consider all of these to be the same chip. So from the 75Mhz PowerPC 603 of 1995 to the 2Ghz Power PC 7447 of 2005 we have seen a whopping 26.6X clock speed increase!!! Shame on Intel and it's crappy X86 architecture! a 9.33x increase from 1995-2002 lets see..... in that same time frame the Power PC 603/G3/G4 family of chips saw a 13.33x increase in clock speed in those same seven years and a 26.6x increase over 10 years.