Why was AMD64 good?

Casual_gamer

Distinguished
Sep 15, 2006
2
0
18,510
With the release of the Core 2 DUO, my friend and I started a debate about what made the AMD Athlon 64 series processors so good when they came out.

He says that the fact it was was able to run 64 bit didn't matter and that it would have kicked ass even if it didn't do 64 bit. I said that the reason it had a performance advantage in a 32 bit environment was because the design change allowed the chip to use the 64 bit part of the chip to do 32 bit tasks better/faster and it would not have been able to do it's magic without having the 64 bit improvements in the chip. What do you think?
 
From PC World
A 64-bit chip can run longer, more complex instructions than a 32-bit one, improving performance of data-intensive tasks such as audio and video encoding, advanced engineering design apps, and, naturally, games.

Equally key is a 64-bit CPU's ability to recognize and use a lot more RAM. Today's 32-bit chips, such as AMD's Athlon XP and Intel's Pentium 4, can address up to 4GB of RAM split between the OS and applications. Few PCs have that much memory, and even fewer apps use it. But with ever-more-complex software, that limitation may become a bottleneck, making Athlon 64's ability to address a whopping terabyte (1000GB) of physical memory very attractive.

But you will need a 64-bit-capable operating system, new hardware drivers, and 64-bit applications to fully take advantage of such a chip, and therein lies the rub.

You're both kinda right.
 

Mex

Distinguished
Feb 17, 2005
479
0
18,780
Ehh...sorry to burst your bubble, but your friend is right on this one. The processor has 64-bit registers, etc. to handle 64-bit tasks. Software today is coded to use the 32-bit registers and instuctions - it doesn't even touch the 64-bit features if the software isn't specifically designed to use them. Then add in the fact that you also need a 64-bit operating system...

I can't explain this very well - someone else who knows the design could explain it better.
Perhaps...JumpingJack?
 

trinitron64

Distinguished
Jun 25, 2006
302
0
18,780
w00t integrated memory controller on the dieeeee bitches

plus they are idiot proof as far as overclocking... unless you fuck them with voltage you can push them as much as you want and ... hell my A64 stays so damn cool I don't even bother paying attention to its temps when im pushing it... cause whats it going to do? Freeze? Throttle?

Pfft

What happened to the good ol' days when T-Bird and XP's would catch on fire faster than your mind can register a reaction of surprise or dread.
 

trinitron64

Distinguished
Jun 25, 2006
302
0
18,780
Im guessing you never owned a T-bird... nor have you ever smelled my 1900+.

Wana smell it?

EDIT:
(I was referring to overclocking, not just everyday use)
For everyday use my 1900+ was a trooper... I remember it hitting 100C on a couple occasions... plus the rubber stoppers on it were melted... and thats all before i burnt it from having the cpu fan pause for 2 seconds while overclocking it (board or psup problem).
 

Mex

Distinguished
Feb 17, 2005
479
0
18,780

konfuzd

Distinguished
May 18, 2006
42
0
18,530
What makes the K8 (A64) so good, though conroe is better, is an excess of bandwidth. A much better description of the K8 archetecture is a northbridge with an integrated processor. The actual processor core is an adapted K7 or Axp, with the main architectural differences being a 12 stage pipe for more speed (vs the xp's 10), better prefetch/branch prediction, and x86-64 (AMD64/EMT64). It's still posesses an fsb, though the distance data has to travel is only a few nm, so it can easily run at the same speed as the processor. This gives a 2GHz A64 32GB/s of theoretical bandwidth over the fsb assuming it still uses the same EV6 bus as the Axp. For the chipset parts, you have the crossbar to allow multiple cores to talk over a single interface, the integrated memory controler, and hypertransport links to other sockets or the rest of the system. This setup gives the antiquated, though still decent, processor access to more bandwidth than it can use, allowing the processor to more fully utilize it's resources an estimated 97% of them.

If you want to know how the K8 was so good compared to Intel's offerings at the time, it was more of a lack of options. At the K8's launch, Intel had to maintain roughly 1.5 times the speed of the K8 with the aging pentium 4 to stay where they were, not to take the lead. Combined with the utter failure that was prescott, Intel decided netburst was no longer in their future and built conroe.
 

Mex

Distinguished
Feb 17, 2005
479
0
18,780
Im guessing you never owned a T-bird... nor have you ever smelled my 1900+.

Wana smell it?

EDIT:
(I was referring to overclocking, not just everyday use)
For everyday use my 1900+ was a trooper... I remember it hitting 100C on a couple occasions... plus the rubber stoppers on it were melted... and thats all before i burnt it from having the cpu fan pause for 2 seconds while overclocking it (board or psup problem).
I could see one of those chips doing that. Despite the heat they put out, they could cope with it.

A little while back I came into the posession of a 900MHz Duron. While doing a disassembly to check everything, I removed the CPU heatsink, and forgot to replace the thermal paste. I turned it on and watched from the BIOS in awe as the Duron quickly climded to 90C before I shut it off. I looked at the processor - no physical damage.

That computer never did work right after that.
 

Casual_gamer

Distinguished
Sep 15, 2006
2
0
18,510
Does this point to a difference that came because of a shift from 32 to 64 bit?

The K8 basically has two operating modes : the legacy mode, that causes the K8 to behave like an IA32 CPU, and the long mode that is the 64 bits mode.

The legacy mode is the common 32 bits mode, used by current operating systems. When running under this mode, the K8 behaves like a K7, boosted with the improvements we talked about in the previous chapter; it has eight 32 bits GPR, and can address 4GB of memory. Thanks to the Physical Address Extension, the addresses are extended to 52 bits, that allows the K8 to use 4096TB, or 4PB (Petabytes) of virtual memory.

The long mode is the extended mode dedicated to the K8. To give an idea, it is to the protected mode what the protected mode is to real mode. The CPU needs to be switched in long mode by the OS, that implies a specific OS. For the moment, Linux and two versions of Windows are able to use the K8 in long mode.

The compatibility mode is a submode of the long mode, and allows a 64 bits OS to run a 32 bits program. Of course, these applications keep being 32 bits applications, that means they won't be able to access extra registers or extra memory. This mode is based on the Wow64 (Windows on Windows 64) layer, that creates a 32 bits virtual memory space, seen by the application like a 32 bits OS.
http://www.cpuid.com/reviews/K8/index.php
 

Eurasianman

Distinguished
Jul 20, 2006
883
0
19,010
hmm... I wonder what happens if I take the heatsink of my Athlon 64 3200 Newcastle. I have had the CPU fan off one time and the temp jumped up a lot, but nothing major... hmmm

other than that, sweet video!!!

oh, and I also have a Pentium 4 2.0A GHz Northwood w/ 400MHz FSB, man, the VCORE is 1.47V and idles at 42C with AS5 and stock Intel Heatsink with a case that has 2 exhaust and intake on side! I'm glad to see that Intel is improving their thermal management. As of now, with C1E and EIST both enabled, my C2D E6600 idles at 1.12V @ 1600 MHz @ 28C!

However, the other day, my Athlon 64 3200 on a S754 was idling around 32C at 1.5V @ 2200 MHz with Cool n Quiet Disabled... hmmm I do have a Zalman all copper hs on it though, so yea...