• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

GPU Memory Latency ?

Zenoth

Diamond Member
Hi.

I just downloaded ATi Tool, and I noticed an option where I can change / decrease the number of Clocks for my GPU's Memory, in term of "TRCTRD", "TRCDWR", "TRP", "TRAS" and many others.

I know about system Memory latency / over-clocking already, but honestly, it's the first time I actually see it's pretty much the same when it comes to a GPU's Memory.

Logical, when you consider a Graphics Card is pretty much a Motherboard with a Processor, in terms of design.

I am using a Radeon 9800 Pro 128 MB, 256-Bit at the moment. And, by default, the Clocks seem alright, I guess. But there is many other things in that Memory section of ATi Tool I do not understand, so therefore, do not want to mess around with.

I was just wondering...

Do any of you here actually know the benefits, if any, when decreasing the Memory Latency directly on the GPU Memory ?

I mean...that, combined with good system Memory latency can surely have advantages.

It is stated in the Tool section to change any of the setting at your own risk, and that it may damage the GPU.

So...anyone here tried that actually ? Or is it really not recommenced to mess around with those GPU Memory settings ?

Thanks.
 
I don't think it will do that much as the memory latency on your VGA is already pretty low, it's usually quite a fair bit lower than your system memory.
 
I also thought about it, currently using a 9800p, and after seeing the message that messing with the timings might damage the hardware, i decided not to mess with it. Anyway, I'm pretty sure the memory latency on a modern graphics card is much lower than on a motherboard, so I dont see the need to change it. Instead, just oc the gpu as far as you can to get more performence out of it. And it's not like the 9800 is starved by memory anyway, with it's 256-bit bus.
 
Originally posted by: Jeff7181
According to manufacturers, overclocking period might damage the hardware.

Yes, if you don't know what you're doing, and push it too far. That's why I installed a Zalman vf700 on my card to keep it cool, and I'm not worried about overclocking it now.
 
Originally posted by: munky
Originally posted by: Jeff7181
According to manufacturers, overclocking period might damage the hardware.

Yes, if you don't know what you're doing, and push it too far. That's why I installed a Zalman vf700 on my card to keep it cool, and I'm not worried about overclocking it now.

So what makes you think reducing memory latency would be any different?
 
Well, one thing I know, is that my system Memory's TRAS is set at 5 Clocks, that's the fastest I can put it at.

But my GPU Memory's TRAS is set at 8 Clocks, which is, going by a system Memory logic, quite low.

If a GPU Memory's Latency, by default, should be the same or faster than a system's Memory, than mine is not set properly.

And...personally, I never ever over-clocked any of my GPUs Core frequency. Even with better, custom air-cooling. I always only over-clocked the Memory, as it is less risky than doing it with the Core. Well...it has always been like that for me anyway.

Maybe I would try to decrease the number of my GPU's TRAS. Since my system Memory's is lower. But I'm not sure what could be the performance advantages of such a change.

That's why I'm just asking if anyone here actually ever changed any of their GPU's Memory settings. And, I mean settings, not frequency.
 
Back
Top