Jump to content

When does Bit-Rate bottleneck VRAM?

Go to solution Solved by D2ultima,

General rule of thumb is 125 bit per 1gb.

That rule of thumb makes no sense...

 

OP, the end-result is memory bandwidth. If you have not enough memory bandwidth to fill an overly large buffer, you have a problem.

 

Technically, I would not put more than 2GB of vRAM on a 128-bit memory bus, as you'd need a very fast memory clock to make good use of more than that. 256-bit memory buses can deal with 4GB and over if their memory clocks are fast enough. If your card has a 192-bit memory bus (or higher) and 5000MHz memory clock (or higher), then you probably won't be bottlenecked at all.

 

It's very difficult to determine an actual "bottleneck" for vRAM really... but the full story is only told by memory bandwidth in total, not clock speed or memory bus alone.

 

To calculate memory bandwidth, multiply the effective clock speed by the memory bus then divide that by 8. The number should be in Megabytes per second.

Example: 6000MHz memory clock on GTX 680 * 256-bit memory bus / 8 (cuz 8 bits = 1 byte) = 192,000. Then divide by 1000 for Gigabytes per second, so 192GB/s. If you check nVidia's specs for this card, you'll see that's what they list the bandwidth as.

 

If you need more info, you can click the vRAM information guide link in my signature.

That's not exactly bit-rate, it's a memory bus we're talking about here. Also, it's not as simple as you might think. Older GPUs have had a pretty wide memory bus already (GTX 480 = 384bit, GT8800 256bit). It only makes sense to directly compare the memory bus of GPUs from the same architecture, a broad statement like you posted doesn't work.

Link to post
Share on other sites

General rule of thumb is 125 bit per 1gb.

Case: Corsair 4000D Airflow; Motherboard: MSI ZZ490 Gaming Edge; CPU: i7 10700K @ 5.1GHz; Cooler: Noctua NHD15S Chromax; RAM: Corsair LPX DDR4 32GB 3200MHz; Graphics Card: Asus RTX 3080 TUF; Power: EVGA SuperNova 750G2; Storage: 2 x Seagate Barracuda 1TB; Crucial M500 240GB & MX100 512GB; Keyboard: Logitech G710+; Mouse: Logitech G502; Headphones / Amp: HiFiMan Sundara Mayflower Objective 2; Monitor: Asus VG27AQ

Link to post
Share on other sites

General rule of thumb is 125 bit per 1gb.

That rule of thumb makes no sense...

 

OP, the end-result is memory bandwidth. If you have not enough memory bandwidth to fill an overly large buffer, you have a problem.

 

Technically, I would not put more than 2GB of vRAM on a 128-bit memory bus, as you'd need a very fast memory clock to make good use of more than that. 256-bit memory buses can deal with 4GB and over if their memory clocks are fast enough. If your card has a 192-bit memory bus (or higher) and 5000MHz memory clock (or higher), then you probably won't be bottlenecked at all.

 

It's very difficult to determine an actual "bottleneck" for vRAM really... but the full story is only told by memory bandwidth in total, not clock speed or memory bus alone.

 

To calculate memory bandwidth, multiply the effective clock speed by the memory bus then divide that by 8. The number should be in Megabytes per second.

Example: 6000MHz memory clock on GTX 680 * 256-bit memory bus / 8 (cuz 8 bits = 1 byte) = 192,000. Then divide by 1000 for Gigabytes per second, so 192GB/s. If you check nVidia's specs for this card, you'll see that's what they list the bandwidth as.

 

If you need more info, you can click the vRAM information guide link in my signature.

I have finally moved to a desktop. Also my guides are outdated as hell.

 

THE INFORMATION GUIDES: SLI INFORMATION || vRAM INFORMATION || MOBILE i7 CPU INFORMATION || Maybe more someday

Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×