A64 Need Dual Chanel?

Kido1986

[H]ard|Gawd
Joined
Mar 10, 2003
Messages
1,180
Does anyone here know if Athlon 64s need dual channel memory? I am looking to upgrade to an A64 3000 or 3200+ and have 1 512mb Kingston Hyperx Pc3200 and don't really want to buy new memory.
 
No.
Socket754 only supports single channel RAM, and does so with an impressive level of efficiency due to the ondie controller.
 
There is no such thing as dual channel memory. Dual Channel is supported by the motherboard or in this case of the A64 the processor. The A64 3200+ does not support dual channel. Dual Channel is worthless for most areas especially gaming and gaming im sure is the reason you are buying the A64. Only heavy applications that can benefit from more bandwidth actually show improvement from dual channel. 3.2GB/s is more then enough to run most everything else fluently.
 
There is no such thing as dual channel memory.
Wrong.

Look up 32 bit RIMMs.
They are literally dual channel memory sticks.

Dual Channel is worthless for most areas especially gaming and gaming im sure is the reason you are buying the A64. Only heavy applications that can benefit from more bandwidth actually show improvement from dual channel.

You're again making absolute statements which just arent wholly true I'm afraid.

Q3a as well as other games were for years known to benefit from additional bandwidth in part because additional bandwidth can result in lower effective latency.

Here is a quote from this article at aceshardware, one of the most well-informed technical sites on the web:

The latency, expressed in clock cycles, is the time it takes to get the data send out, that has been requested. It takes some time before the memory address is decoded and the data is ready to be send back. Bandwidth is, of course, the number of megabytes that can be send per second through the memory bus. Contrary to popular belief, latency and bandwidth are very closely related. If latency is very high for the first word (4 bytes), the bandwidth is lower, especially with random memory accesses.
....
Indeed, low bandwidth will increase the latency that the CPU sees! So the last thing I claim is that high bandwidth is useless. But what we do claim is that high bandwidth memory is useless if the front side bus can not cope with it. And now we got the benchmarks to prove it! Let us see how much the front side bus and the memory bandwidth and latency affects performance. You will be amazed, I can tell you that.

Dual channel is most important in bandwidth-limited applications, but additional bandwidth reduces effective latency by allowing transfers of a given size to occur in less time.

This is why we can say that the socket 939 A64's will be extremely impressive when they arrive, because they will combine the lowest memory latencies in the industry (that comes with using unregistered RAM and an ondie memory controller) with the highest efficiency bandwidth utilization
around (also as a result of using an ondie controller rather than a northbridge).

Hopefully the aggressive SSE2 utilization already seen in rendering programs (like 3DSMAX) will continue with SSE3 to allow the Prescott to stay really competitive with the absurdly low latencies of the A64 after the PIv's bandwidth advantage has been totally eliminated and will drop prices all-around.
 
Wrong. Look up 32 bit RIMMs.
They are literally dual channel memory sticks.

Yes there is dual channel like RIMMs from Rambus but Rambus is not an equating factor here because its not supported for the motherboards. There is no such thing as dual channel DDR and you dont have to buy dual channel labled DDR kits for enableing dual channel is what i meant.

You're again making absolute statements which just arent wholly true I'm afraid.

Q3a as well as other games were for years known to benefit from additional bandwidth in part because additional bandwidth can result in lower effective latency.

Here is a quote from this article at aceshardware, one of the most well-informed technical sites on the web:

Dual channel is most important in bandwidth-limited applications, but additional bandwidth reduces effective latency by allowing transfers of a given size to occur in less time.

This is why we can say that the socket 939 A64's will be extremely impressive when they arrive, because they will combine the lowest memory latencies in the industry (that comes with using unregistered RAM and an ondie memory controller) with the highest efficiency bandwidth utilization
around (also as a result of using an ondie controller rather than a northbridge).

Hopefully the aggressive SSE2 utilization already seen in rendering programs (like 3DSMAX) will continue with SSE3 to allow the Prescott to stay really competitive with the absurdly low latencies of the A64 after the PIv's bandwidth advantage has been totally eliminated and will drop prices all-around.

For years that was true because the amount of bandwidth needed by these games hadn't been reached. Once PC3200 and higher was released its now been reached for current games and most applications.

http://www.pcstats.com/articleview.cfm?articleid=873&page=5

You can see that even with 100Mhz increase in bandwidth it scores less in Quake 3 then the RAM running 100MHz slower and 2-2-2 timings.

Of course with dual channel you dont have to sacrifice timings for bandwidth.

The launch of AMD’s 3400+, however, throws a nasty wrench into things. Because it’s a Socket 754 pin part, the 3400+ only has a “mere” 3.2 GB/s of memory bandwidth—yet manages to equal the performance of the FX-51 (or lag it by mere seconds) in virtually every test we ran. If this sounds familiar, it’s because we’ve already seen it happen twice—once with the BH7, and once with the NF2 400.

http://www.sudhian.com/showdocs.cfm?aid=480&pid=1768

The fact that the FX-51 uses registered RAM is not enough to effect the performance of its on-die dual channel memory controller to the point there is no benefit from it if games and the majority of applications actually needed it. The registered RAM from Corsair is even rated for 2-3-2-5 timings.

The reason for this is that CPUs and computer systems are designed wherever possible to avoid having to go to main memory to work on the next piece of data. CPUs are designed to get that data ahead of time to be ready when it's needed.

This low scaling-up of performance by adding faster memory is nothing new. Memory has never provided a big boost to overall system performance. For instance, when memory went from 66MHz to 100MHz, that only contributed about 3% to performance. When memory went from SDRAM to DDR, the level of improvement was about 7%

This is one of the biggest myths in this field, and it's just incorrect when it comes to the real world. The only place where faster memory gives you big improvements is in memory benchmarks. That's it. When you get to applications and games, you'll see some improvement, but nowhere near the increase in memory speed or bandwidth.

http://www.overclockers.com/tips00224/

Timings are far more important to the performance in games then bandwidth especially after the requirements have been met. The FX-51 more then doubles the bandwidth and yet just because it uses registered RAM which is a little slower in access times it completely diminishes the gains all togethor? I dont think so. Its because that amount of bandwidth it can give is not required for current processors and front side buses. The Pentium 4 with dual channel enabled has twice as much bandwidth as the A64 and yet because its latency is a little lower because of an on-die controller that double in available bandwidth means nothing in most areas and especially games. If more bandwidth lowers latency then i think its a safe bet that doubleing the bandwidth is going to give you lower latency then just an on-die controller but thats not true. You can think of it like a broadband connection. You can have a 2.0MB/s line but if your downloading from a site and your maxing out at 30KB/s then that 2MB/s connection does you no good.
 
Back
Top