Does GPU frequency really matter all that much between GTX and RTX?

Wobitancrawfodi · 05 Jun 2020

I have a GTX 1080 Ti which I've been wanting to upgrade for some time.

When I look at the CPU Clock the RTX is slower in almost all cases (base @ 1350 compared to 1530 for the GTX). Except for a few top end cards, the gaming or OC clock speed is also often less than the GTX.

Does it really matter. Both have 11G GDDR. Does the memory speed, bus size and CUDA cores make up for the lower clock speeds? I guess they must because all RTX 2080 Ti's out perform the GTX 1080 Ti by a good percentage.

I ask because prices are astronomical, and if the clock speed isn't the biggest concern I can save anywhere up to A$700.

EDIT: I'd be looking only at ASUS ROG and MSI cards. Gigabyte are too noisy, and I have no interest in the clones like Galax, Zotac and so forth. Leaning strongly towards the ASUS as it has 2xHDM 2.0b. I have tried dual DP and I can't fix the issue with applications being moved to the primary screen when the monitor sleeps.

Faith · 05 Jun 2020

The core clock speeds are lower because the RTX cards have more processing units and CUDA/RT cores to adhere to the set power threshold. Higher clocks speeds are fine, but if the core count is too low and inefficient for the running application, it doesn't really matter how fast it's going. Think about it like this; clock speed can be seen as revolutions on a fan. If it is a small fan it has to work harder to be efficient. If the fan is bigger it can rotate slower at the same efficiency. Core clock and memory clock are seperate. The RTX cards are using GDDR6 memory which is faster than the GTX GDDR5(x) memory. All of this has something to do with how good the card is, but it's more in the engineering and architecture rather than just raw speed. Hope it helps.

Wobitancrawfodi · 06 Jun 2020

I had a read of this... Unfortunately it only talks about RTX 2080 Ti's generically. But if I read it right, There is no real need to be going for the top end card, given pretty much all RTX 2080 Ti's have 14Gbps (616GB/s), 352 bit bandwidth etc.

If I read you right to limit power consumption to the quoted rating the speeds are lowered, so higher clocked cards could exceed the power ratings? And I might not get much if any benefit from the extra 100 - 150Mhz clock speed?

Faith · 06 Jun 2020

It's more of a balance I think. I'm no engineer so I can't know the specifics, but if something is increased in one area, something else may be decreased to hold the card within its power and thermal limit. Cards can exceed the threshold, and 100 - 150 MHz increase can be noticable depending on what you want out of it, but the silicon is a part in how they want their factory overclocks and how well it can perform without it becoming unstable. I forgot to mention that these cards also have boost clocks, so it can go higher based on a few factors like voltage and temperature. As I wrote, all has something to do with how well the card will perform in the end. Memory clock, core clock, core count, architecture etc. We can't really judge a card based only on frequency, that was my point. If you want to know how well a card is performing, look up benchmarks and reviews that's directed towards your use-case.