hey is there any difference with nvidia 128 bits and 256 bits or gb makes difference more and how about amd ?
It affects memory speed, as it (apparently) is how many bits the GPU can send out to the DRAM at once.
For example. in the following chart, you may notice that as the interface jumps from 64 to 128 bit, the memory bandwidth doubles, would should be better especially for high resolution displays, but probably a lot of other speed related things.
Also interesting that from 128 to 256, there isn’t such a marked jump, so it seem that the GPU isn’t then so much a speed bottleneck.
More shaders help too, but increasing the shader units probably wouldn’t have linear gains if the the memory access was plugging up things.
-about gpu bus width.pretty simple really.-more data faster down a wider road.
Hello @savasuctu , This is a rather complex question because the number of bit on it’s own doesn’t really mean anything.
Yes, Higher bit rate is better i.e. As you go up in bit rate in general the cards perform faster… 64bit, 128bit, 192bit, 256bit, 512bit but other factors play a part in performance in combination with bit rate.
The other factors at play are the GPU clock speed and also the RAM type and bus speed.
Basically this all equates to how many roads are available and how fast is the traffic driving down those roads.
Higher ram clock speeds mean more traffic per second.
Higher bit rates mean more roads at that speed.
Traffic is what feeds the GPU data back and forth between the RAM.
If you are trying to compare cards I have found one of the best ways to compare is to look at Bandwidth throughput in GB/s.
Older cards like the GTX280 etc. had much slower RAM than today so to get data throughput to the GPU they had to hike up the bit rate which is why a GTX280 has a 512bit data bus but runs at 602Mhz to give it a Bandwidth of 141GB/s, which is actually really good.
In comparison a new GTX960 has a 128bit memory bus, but the ram runs at 7000Mhz to give an effective 112GB/s data throughput to the GPU.
This doesn’t always mean that the GPU on the Graphics card is capable of using all this bandwidth, but it does mean that this is the maximum data throughput that the card could handle if it needed to.
Normally this balance is worked out by Nvidia/AMD when making the chip which is why you will see many card variations, but the different manufacturers cannot change the memory bandwidth. Normally all the 3rd party card manufacturers can do is put faster ram in eg. GDDR5 instead of GDDR3, or they can push the clock speed of the assigned ram closer to it’s maximum, and in that way increase the GB/s bandwidth throughput.
Hope this helps clear up a few things. Michael VS