[SOLVED] Overclocking RTX 4080 Vram question

[mjbn1977]

I have a couple of questions about overclocking VRAM on the RTX 4080. First and foremost, I do not overclock to improve any gaming performance as the card is plenty fast enough for all my current gaming needs. I am just want to play around a little in benchmarks. Want to see what is possible.

1. what is an expected and save overclock for 4080 Vram that all cards should be able to do? At least on most cards? In what regions it starts being silicone quality dependend?
2. I heard that on the Micron DDR6 vram too much overclock can actually slow the card down, is this true?
3. In Afterburner, when I increase the clock rate for VRAM by 200, is this 200 per vram module or for the whole card? Still getting confused about the VRAM clock rates. Stock is 1400MHz but when I show vram speed in Riva its showing 11200Mhz. When I increase clock in afterburner by 200 the value changes to 11400.

 

[Roland Of Gilead]

I have a couple of questions about overclocking VRAM on the RTX 4080. First and foremost, I do not overclock to improve any gaming performance as the card is plenty fast enough for all my current gaming needs. I am just want to play around a little in benchmarks. Want to see what is possible.

1. what is an expected and save overclock for 4080 Vram that all cards should be able to do? At least on most cards? In what regions it starts being silicone quality dependend?
2. I heard that on the Micron DDR6 vram too much overclock can actually slow the card down, is this true?
3. In Afterburner, when I increase the clock rate for VRAM by 200, is this 200 per vram module or for the whole card? Still getting confused about the VRAM clock rates. Stock is 1400MHz but when I show vram speed in Riva its showing 11200Mhz. When I increase clock in afterburner by 200 the value changes to 11400.

Hey there,

To answer your questions:

1. Each card is different from the next. There isn't one setting for all cards. You have to trial and error it yourself. With that said, 100-150mhz on the core clock, and between 600-1000mhz on the mem is often typical, providing the GPU has decent cooling, and there's good airflow in your case, and you're able to up the power limit. You can use something like Superposition for a bench, and try the OC scanner in MSI Afterburner. See what results you get. Then try it manually. Using increments of 5 or 10 on the core and 50mhz on the mem each time. Then test again. Wash Rinse Repaet untill you get artefacts or temps over 75c on the GPU.

2. Not heard that.

3. No, this is only on the mem. Not on the core clocks.

 

[sizzling]

Regarding #2 I read the same but applying to all gpu’s. From what I recall it had to do with introducing data errors. Hopefully someone who actually knows definitively drops in.

 

[mjbn1977]

Hey there,

To answer your questions:

1. Each card is different from the next. There isn't one setting for all cards. You have to trial and error it yourself. With that said, 100-150mhz on the core clock, and between 600-1000mhz on the mem is often typical, providing the GPU has decent cooling, and there's good airflow in your case, and you're able to up the power limit. You can use something like Superposition for a bench, and try the OC scanner in MSI Afterburner. See what results you get. Then try it manually. Using increments of 5 or 10 on the core and 50mhz on the mem each time. Then test again. Wash Rinse Repaet untill you get artefacts or temps over 75c on the GPU.

2. Not heard that.

3. No, this is only on the mem. Not on the core clocks.

This is what confuses me. steps of 50 Mhz on what? each module or the whole card? So the 4080 stock memory clock is 1400Mhz, but shows as 11200 Mhz in afterburner/River Tuner overlay. So, when I supposed to go up by 50mhz on the memory each time, do you mean 50 Mhz in total....basically going from 11200Mhz to 11250, or 50 Mhz per module, basically going to 11600Mhz total? This is the confusing part about Vram overclocking.

I know how to approach overclocking, but when it comes to VRAM, I don't understand what I use as base....

Do I use 1400Mhz or 11200Mhz as base I am working from....?

 

[Roland Of Gilead]

This is what confuses me. steps of 50 Mhz on what? each module or the whole card? So the 4080 stock memory clock is 1400Mhz, but shows as 11200 Mhz in afterburner/River Tuner overlay. So, when I supposed to go up by 50mhz on the memory each time, do you mean 50 Mhz in total....basically going from 11200Mhz to 11250, or 50 Mhz per module, basically going to 11600Mhz total? This is the confusing part about Vram overclocking.

I know how to approach overclocking, but when it comes to VRAM, I don't understand what I use as base....

Do I use 1400Mhz or 11200Mhz as base I am working from....?

Okay, so there are literally only 3 things you need to change :

https://imgur.com/MVlgUwg

View: https://imgur.com/MVlgUwg


So, you need to take a logical step by step approach.

1. Open up power limits by unlocking voltage control and voltage monitoring on the general tab under the Settings (Cog) button.
2. Put power limits to max (normally about 110%
3. Do core clock OC or Vram OC, not both. Lets start with core clocks
4. Up the slider on the core clocks by increments of 5/10, something manageable. Apply the settings. Test as explained. Keep upping the core clocks until you start seeing artefacts or bugs on screen. Once you hit a limit, then row back a couple of notches for stability.
5. Once that's done, do exactly the same for the vram, but this time in higher increments of 50mhz (because mem OC's better than the cores more often than not.
6. Wash, rinse repeat, and see how far you get.

Now, as mentioned you can also try the OC Scanner built into MSI Afterburner. This takes all of the pain in testing away from the user.

Remember to check out some games with your OC settings. If they work well, then the OC should be stable. If you have any crashes, the OC might be a bit too much.

Some games will not respond to any OC at all, and will simply crash just because of that.

Read some stuff about what you want to achieve:

Nvidia RTX 4080 Overclocking and Test Setup - Nvidia GeForce RTX 4080 Review: More Efficient, Still Expensive | Tom's Hardware (tomshardware.com)

How to Use MSI Afterburner OC Scanner to Boost GPU Performance (partitionwizard.com)

In terms of vram, you are only shifting the mem clockspeed. When you do it, it's across all mem chips. You don't need to factor in each individual mem chip. It's one for all approach.

edit: When you move the core or mem slider, you will only see what you are adding. Not what the speeds should be. Take it that at stock with no changes, you are simply adding mhz incrementally.

If you use MSI OSD for in game metrics, you will see that for every increase in 200mhz on the mem clock, the OSD will display at twice that amount.

The 11200mhz you are seeing is the default base for your ram. This translates to 22.4 gbps bandwidth. When you add more mhz to the base, it doubles up in terms of bandwidth.
So you will test at 11250/11300/11350 and so one till you hit a wall.

 

[mjbn1977]

Okay, so there are literally only 3 things you need to change :

https://imgur.com/MVlgUwg

View: https://imgur.com/MVlgUwg


So, you need to take a logical step by step approach.

1. Open up power limits by unlocking voltage control and voltage monitoring on the general tab under the Settings (Cog) button.
2. Put power limits to max (normally about 110%
3. Do core clock OC or Vram OC, not both. Lets start with core clocks
4. Up the slider on the core clocks by increments of 5/10, something manageable. Apply the settings. Test as explained. Keep upping the core clocks until you start seeing artefacts or bugs on screen. Once you hit a limit, then row back a couple of notches for stability.
5. Once that's done, do exactly the same for the vram, but this time in higher increments of 50mhz (because mem OC's better than the cores more often than not.
6. Wash, rinse repeat, and see how far you get.

Now, as mentioned you can also try the OC Scanner built into MSI Afterburner. This takes all of the pain in testing away from the user.

Remember to check out some games with your OC settings. If they work well, then the OC should be stable. If you have any crashes, the OC might be a bit too much.

Some games will not respond to any OC at all, and will simply crash just because of that.

Read some stuff about what you want to achieve:

Nvidia RTX 4080 Overclocking and Test Setup - Nvidia GeForce RTX 4080 Review: More Efficient, Still Expensive | Tom's Hardware (tomshardware.com)

How to Use MSI Afterburner OC Scanner to Boost GPU Performance (partitionwizard.com)

In terms of vram, you are only shifting the mem clockspeed. When you do it, it's across all mem chips. You don't need to factor in each individual mem chip. It's one for all approach.

edit: When you move the core or mem slider, you will only see what you are adding. Not what the speeds should be. Take it that at stock with no changes, you are simply adding mhz incrementally.

If you use MSI OSD for in game metrics, you will see that for every increase in 200mhz on the mem clock, the OSD will display at twice that amount.

The 11200mhz you are seeing is the default base for your ram. This translates to 22.4 gbps bandwidth. When you add more mhz to the base, it doubles up in terms of bandwidth.
So you will test at 11250/11300/11350 and so one till you hit a wall.

Thanks for the detailed explanation. I know how to overclock the core and already figured out the limit. My questions was mainly around the VRAM clocks and by how much to increase due to the confusing different reporting 1400Mhz vs 11200Mhz.

 

[mjbn1977]

Like I said, I am just playing around and want to get familiar with what is possible. My next goal is to undervolt the card by adjusting the curve. Trying to get my stock speeds and slightly less power use.

 

[Phaaze88]

Memory frequency is at the mercy of silicon lottery just like core clock is; the memory controller in the die that needs to put up with it.

As for the 'slow down': https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-founders-edition/39.html (scroll down to Overclocking, but I'll copy-paste the part of interest)
"GDDR6X memory comes with an error-detection mechanism that is active on the GeForce RTX 3080. On previous generations, the "ECC memory" feature was only available for much more expensive, professional cards. AMD has enabled it on several Radeon cards, too. While GDDR6X error-detection isn't identical to ECC, it is similar; the memory controller is able to detect most transmission errors, and Error Detection and Replay (EDR) functionality will keep retrying that memory transfer until it succeeds.

For memory overclocking, this means you can no longer rely on crashes or artifacts as indicators of the maximum memory clock. Rather, you have to observe performance, which will reach a plateau when memory is 100% stable and no transactions have to be replayed. Replaying memory errors reduces overall memory performance, which will be visible in reduced FPS scores."

 

[mjbn1977]

Memory frequency is at the mercy of silicon lottery just like core clock is; the memory controller in the die that needs to put up with it.

As for the 'slow down': https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-founders-edition/39.html (scroll down to Overclocking, but I'll copy-paste the part of interest)
"GDDR6X memory comes with an error-detection mechanism that is active on the GeForce RTX 3080. On previous generations, the "ECC memory" feature was only available for much more expensive, professional cards. AMD has enabled it on several Radeon cards, too. While GDDR6X error-detection isn't identical to ECC, it is similar; the memory controller is able to detect most transmission errors, and Error Detection and Replay (EDR) functionality will keep retrying that memory transfer until it succeeds.

For memory overclocking, this means you can no longer rely on crashes or artifacts as indicators of the maximum memory clock. Rather, you have to observe performance, which will reach a plateau when memory is 100% stable and no transactions have to be replayed. Replaying memory errors reduces overall memory performance, which will be visible in reduced FPS scores."

That was exactly what I was referring to, but I couldn't remember the details.....thank you!!!