GDDR5 to DDR3

Spectralware

Reputable
Feb 13, 2016
37
0
4,540
I know a lot about software (because i work with it), and i was trying to learn about hardware, but i didn't understand a thing about GPUS. One of the first things that comes to mind when comparing a GDDR5 graphics card vs a DDR3 graphics card is the speed. For an example, a graphics card with 1Gb and GDDR5 technology has a lot better performance than a 2Gb with DDR3 technology.

I asked a friend if there was a way to measure the GDDR5 with DDR3, making a "conversion". If you want to convert Kilometers to Meters, you would do something like this:

1 Km = 1.000M.

My friend said that a conversion of Graphics Card technology would be like:

1Gb GDDR5 = 4Gb DDR3

Is that right? Is there any way to "convert" a GDDR5 graphics card to DDR3? You know, about memory. Like... If a Graphics Card GDDR5 have 1Gb, if you converted it to DDR3 how many memory would it have??


###CORRECTION###

No one understood what i was asking and is vulgarly talking nonsense things, so i am going to try to be more clear:

IF (IIIIFFFF) there was a way to convert a graphics card with GDDR5 Technology to DDR3, what would be it's total memory? counting all the advantages, such as speed and bandwith, a 1Gb GDDR5 Graphics Card would be equivalent to wich type of DDR3 Graphics Card? 2Gb, 4Gb, etc...

 
Solution
I'm not sure your friend understands any better than you do. Consider this a rough break down of RAM technology.

To start with, the basic unchanging aspect of RAM is that it has a capacity and it has a clock speed, be it 100 MHz or 1000 MHz. The clock speed in part determines how much data it can transmit each cycle.
Throughout most of the 1990s, there was a type of RAM known as SDR-RAM. The "SDR" stands for "Single-Data Rate." This meant that the computer could access the RAM just once every cycle. As this RAM was much older, it was typically clocked asynchronously with the system's FSB, which was locked at either 66 or 133 MHz.

Then came DDR-RAM. The "DDR" stand for "Dual-Data Rate." This allowed the computer to access the RAM twice...

Spectralware

Reputable
Feb 13, 2016
37
0
4,540


Was that supposed to be an answer? It both didn't answer my question AND you said an OBVIOUS thing...
 
You can't "convert" the VRAM, it's a different architecture and specification. The speeds are much different as well. The amount has nothing to do with speed, the speed indicates how fast the GPU chip itself can communicate with the memory (adding and moving data as necessary), it only affects performance when there is demand for more more than you have or the speed isn't enough and causes latency.

If you have a 4gb GDDR5 GPU and a program wants 6gb of space, then the GPU has to either make room by swapping out data or by using other resources. If it's DDR3, this will happen much slower and the performance will be affected much more. Which is why no high end GPU's use DDR3 and the top end GPU's are moving to HBM.

There is no equivalency between VRAM tech. This is just a crude, basic explanation, but you could write a book on how everything interacts on a GPU to affect performance.
 
I'm not sure your friend understands any better than you do. Consider this a rough break down of RAM technology.

To start with, the basic unchanging aspect of RAM is that it has a capacity and it has a clock speed, be it 100 MHz or 1000 MHz. The clock speed in part determines how much data it can transmit each cycle.
Throughout most of the 1990s, there was a type of RAM known as SDR-RAM. The "SDR" stands for "Single-Data Rate." This meant that the computer could access the RAM just once every cycle. As this RAM was much older, it was typically clocked asynchronously with the system's FSB, which was locked at either 66 or 133 MHz.

Then came DDR-RAM. The "DDR" stand for "Dual-Data Rate." This allowed the computer to access the RAM twice per cycle. This means that RAM operating at 133 MHz for example effectively functioned as though it was RAM operating at 266 MHz. It was also able to clock higher, and it provided more bandwidth to the system as a result.

DDR2, DDR3 and DDR4 all function similar to the first DDR, but they progressively consumed less power (DDR4 being the most efficient), and they operated at higher clock speeds (DDR4 being the fastest).

Eventually we arrived at DDR5. The DDR here is misleading, as the RAM actually be accessed four times per cycle. DDR5 can be clocked roughly as high as 2000 MHz, but as the system can access the RAM four times each cycle it effectively works as though it was clocked at 8000 MHz.

The G in GDDR5 just stands for "Graphics," which typically indicates it operates with a higher voltage to enable higher clock speeds, and it doesn't really change anything else.

It should also be noted that RAM is typically labeled as operating at its effective clock speed. So GDDR5 is often listed as operating at 8000 MHz for example.

As for capacity, this doesn't change. 1GB of GDDR5 is 1GB of DDR3. GDDR5 will perform better, however, as it has more bandwidth and this allows the CPU/GPU to operate more efficiently with higher throughput.

One final note, don't be mislead in thinking that more RAM is equal to more performance. This is a common mistake. If you have a GPU that has 4 GB of GDDR3/GDDR5, but the graphics processor itself is too slow to actually use all of this RAM, it won't help you any. It will just sit unused. If you have two GPUs that are identical except one has 2 GB of RAM and one has 4 GB, and you run a game on them that only uses 1 GB of RAM, then they will perform identically. The only time the 4 GB card would get any advantage at all is if more than 2 GB of RAM was used. When comparing a card with a lot of DDR3 vs. just a small amount of GDDR5, the GDDR5 is always the better option as the increased bandwidth will boost performance more than the increase in RAM.
 
Solution

neblogai

Distinguished
The answer by IInuyasha74 was not complete (he did not include bus width), but it was a very nice and you should try reading in to it. THERE IS NO way to convert amount of GGDR5 into some other amount of DDR3, like you want. These are different metrics: amount of memory is for storing game textures, while memory speed- for moving them. Different games and game resolutions use these metrics differently. All you need to know- most of modern games use 1.5-4GB of memory for running at 1080p resolution. If GPU has less memory than that- it will have to store some data outside of graphic card memory, which will slow it down. Either way (DDR3 or 1GB GDDR5) is not recommended for current games, as many will run too slow to be playable. Some modern games might be playable- search youtube for definite answer.
 

Ivan96

Reputable
Feb 18, 2015
460
3
4,865


Well,my point was you can not compare them like you want in your example 1gb=4gb ddr3 that doesn't make sense at all because 4gb can never be like 1gb no matter how fast it is. So it's pretty simple gddr5 is faster almost a double but when buying a gpu it depends do you need fast memory or big memory and there is your answer.
 


Yes the bus width is important too and greatly effects performance, and certainly should not be forgot. I left it off out of concern that the post was already getting a bit long. The easiest thing to remember is the GPU with the highest memory clock speed and bit-width will have the most bandwidth.
 

Spectralware

Reputable
Feb 13, 2016
37
0
4,540


Okay, i think i understand now, thanks ^.^