gpu memory vs gpu clock speed

S

snail

Expert Member
Joined
Feb 5, 2008
Messages
1,395
Reaction score
5
Location
Toti
The radeon 5770 memory / clock speeds sit at 4800 / 850 respectfully.

Now the overclocked edition sits at 4900 / 875 respectfully.

This works out in the ratio of roughly 5.6 / 1 respectfully.

Is it safe to assume that working with the aforementioned ratio that I could increase safely in the given ratio, for example 5100 / 910 respectfully.
 
Usually you would actually want to increase both the exact same percentage.
As soon as you're starting to increase them at different percentages, you might actually decrease your performance.

Also note that the memory is using a multiplier of 4, so it's actually running at 1.2GHz and not 4.8GHz.
 
Then that means that the over clocked version is increased by 2.08 percent / 2.94 percent respectfully which is almost an entire percentage difference. Now based on your theory that means that ati are making a stuff up of things.

Don't get me wrong, not flacking anyone, just doing the maths and find out which is the safest method to rely on.
 
Btw: it's "respectively" and not "respectfully" :)

Changing both speeds using the same percentage might give you the best results, but an increase in the GPU clock speed is usually more effective than changing the memory speed.
I'm not very good at overclocking either - I've only recently done this with my 2x MSI Cyclone GTX 460's.

If you have time, you're welcome to play with the different settings. If you ask me, that 4900MHz memory speed is very close to its limit, where as the GPU clock might even reach 900MHz.
3% overclock is fairly pointless. I mean, I've overclocked my GTX 460 by 20% :) but then again, it is known to be a good overclocker.
 
I read on other forums the guys got 5600 running at full throttle without any issues
 
Increasing your memory will just increase how much data can be tossed around in the card itself. So remember, if the gpu clock isn't up to scratch, it won't be able to fill the bus lanes of the memory, which makes the memory overclock less effective.

This is why people usually push their ram along with their core because you want to unlock a few extra Gb/s in the bus lane data transfer, while you increasing the GPU's ability to gobble up that data. You can get away with not overclocking the ram very much, depends on the GPU. For example, the 8600gt's gpu really didn't use much data, yes, overclocking the ram unlocked it, but it wasn't held back very much by the slow ram. GPU overclocks were more effective.

Just be careful, overclocking the ram can lead to artifacts. I pushed my old ATI 1600XT to the limits, and I mean the limits. It then stopped rendering things properly, ran close to the temperature of fire, but I was getting awesome frames per second :p.

I suggest you just read up about overclocking. I haven't been in the game for a while, so just running off prior knowledge of my more 'rebellious' gaming years ;).

Ps, don't use the overclocks guys on forums get. They may very well be legit, buuut, they usually using expensive cooling, have a card from a 'good batch', and also have been overclocking and burning out pc components for years.
 
Last edited:
I also used to overclock along time ago with spare parts, but haven't done much with gpu's.

Not too worried about pushing it, just trying to get best results when I do overclock slightly.

I want to find out which way is better as many people have different opinions.
 
snail said:
I also used to overclock along time ago with spare parts, but haven't done much with gpu's.

Not too worried about pushing it, just trying to get best results when I do overclock slightly.

I want to find out which way is better as many people have different opinions.
Click to expand...

In my opinion, get to know the card in question. Like if there's power constraints (say is draws 70 watts, and is only drawing power from the PCI-E lanes, you can't overclock it much, but it'll be better to push the GPU and leave the ram).

Also, it'll be a good idea is to fiddle and see what works best. Today, cards vary so much in their architecture that it'll be impossible to have a linear ratio. Maybe 5 years back, the architectures weren't so different, but now... yo.

I also think AMD cards like having more bandwidth, or they respond nicely to a decent bandwidth, where as Nvidia cards will see just better AA because those GPU's are more brute force, where as AMD went for something more stream lined (each has benefits and withdraws... so don't bitch at me).

IIRC, those 5770's still have 128bit-bus widths. You'll see a massive improvement with AA performance, as well as boost your fps at higher resolutions, maybe even AF will improve. Give the GPU clock 10% improvement, and the RAM 15~20% improvement, I think that should do you best. but Fiddle and correspond to us :)
 
snail said:
I also used to overclock along time ago with spare parts, but haven't done much with gpu's.

Not too worried about pushing it, just trying to get best results when I do overclock slightly.

I want to find out which way is better as many people have different opinions.
Click to expand...

Depends on the card. The 6800 series cards for example see more of an improvement from memory overclock than gpu overclock.
 
as per asus, if you use their utility to overclock the card, then it will not void the warranty, i think the program is called smart doctor.
 
snail said:
as per asus, if you use their utility to overclock the card, then it will not void the warranty, i think the program is called smart doctor.
Click to expand...

that's because it will limit the overclock, like a 10% overclock will shorten the life time of the card by a minuscule amount. But it's when people don't know what they doing, and they push the living crap out the card when things go horribly wrong :p.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X