Question Factory overclocked versus non-overclocked graphics cards ?

[Pimpom]

How much performance gain do factory overclocked graphics cards get over cards running at stock clock? (This is not about an end-user doing their own overclocking).

I know the question is vague. I expect that the result will depend on the particular GPU, how much it is overclocked and also on the games and benchmarking software used to test them. So I'm only hoping for some ballpark figures.

As an example, take a recent mid-range GPU like an RTX xx60 Ti and a card manufacturer releases two versions - one overclocked and the other not, everything else being equal. How much of a difference in performance should one normally expect?

 

[Lutfij]

I know the question is vague.
You hit the nail on the head. Reviews of said cards will tell you if the factory overclock and the added price tag(from said overclock) is worth it.

I think I've stated this on your other thread that Nvidia have locked down how much you can actually overclock on your cards, even from the factory(or even as an end user).

So I'm only hoping for some ballpark figures.
Not happening, unless you narrow down a GPU and the camp from which the GPU comes from.

How much of a difference in performance should one normally expect?
Depends on the overclock actually.

The advice we use to give is to pick up a reference card and overclock it yourself, perhaps even undervolt it but you're doing all that yourself and not paying a premium that factory overclocked cards would carry.

 

[geofelt]

Factory overclocked cards have better binned chips that warrant higher performance.
You probably get fair value from them.
I would avoid the versions with the very highest level of overclocking.
If you need more performance, buy a higher class card in the first place.

 

[Eximo]

Another thing to consider is that some factory overclock cards allow for higher power settings than stock. Still within Nvidia's restrictions, but some cards might let you do 110% power, others might go up to 115%, etc.

Even more it depends on the very exact GPU/Memory you get. Not the model, or the factory overclock. Just which exact chips you get. Even after binning doesn't mean that two cards from the same batch will perform the same. The binning just means a set or tray of chips/memory meets the criteria to ship them all with the same factory settings. Some will be on the bottom end of that criteria, some will perform beyond that.

Maybe your GPU is binned well enough, but the memory is exceptional and can be taken well above stock. Professional overclockers either get specially selected cards from vendors or order several to find a good one for their purposes.

You also have to be careful how you measure success. A standard benchmark is one thing. bigger number better. Sometimes an overclock can look good, but actually introduces more stuttering and results in poorer gaming feel.

I would say for most people that want to overclock a late model GPU. Keeping the power output down is better than going for max performance. Undervolting will keep the card a little cooler, and allow for higher average boost clocks. Though you can get quite a bit of free performance from just maxing out the power settings and trying to increase frequencies. A mix of both is usually the ideal, but it takes a lot of fine tuning to get it optimized.

 

[Roland Of Gilead]

You're question is a bit tricky. On one hand you want to get some information, but on the other hand, you stipulate or dismiss certain aspects that would lend some accuracy to your questions, rather than being vague as you've eluded to.

There are many factors outside of some that you've mentioned or other have too; The most simplistic being that each GPU is different from the next, despite binning, in terms of silicon. They all run a little differently.

In ballpark terms, we could say that you might get a 5-10% increase, or sometimes a fraction more, dependant on the game or bench with an SC or SSC/OC 3rd party version, But running multiple repeatable tests will show variances from run to run. With single digit performances, run to run variance can have an impact.

The fact that, if you had some reasonable knowledge about GPU's and how they work, you can get a bog standard non OC version and OC it to within a percent of one that's binned and called SSC or whatever and about 70$ more expensive.

Anyway, just a little bit of context.

 

[Pimpom]

In ballpark terms, we could say that you might get a 5-10% increase, or sometimes a fraction more, dependant on the game or bench with an SC or SSC/OC 3rd party version, But running multiple repeatable tests will show variances from run to run. With single digit performances, run to run variance can have an impact.

That's the kind of answer I was looking for. I don't mean to be rude but sometimes I think the experts here tend to overcomplicate things.

 

[Roland Of Gilead]

That's the kind of answer I was looking for. I don't mean to be rude but sometimes I think the experts here tend to overcomplicate things.

Lol 😆

Not rude at all. I kinda wanted to gi e general answer to a general question because I could jabber on all day going deeper into it.

Quality of components, number of fans, voltage tolerances all play a further part.

Anywho, glad to offer an answer, of sorts 😆

 

[Pimpom]

Lol 😆

Not rude at all. I kinda wanted to gi e general answer to a general question because I could jabber on all day going deeper into it.

Glad you took it that way. I have no doubt you could go much deeper into the subject but sometimes that kind of answer is all that someone needs or expects.

Quality of components, number of fans, voltage tolerances all play a further part.

I'm all too aware of that. I've been in electronics for >50 years, designed and built things for various fields of application, even invented a few new circuits.

Been using computers for over 30 years, Windows PCs for >25 of those. But computers are not my specialty. I don't game, edit videos or do any other power stuff. My favourite CAD program is so old it could run smoothly on an AMD K5. (I have much more advanced software but I love this one).

OTOH, I'm often placed in a situation where I have to help others make purchasing decisions. When that happens, I don't want to brush them off with some throwaway suggestions. If I'm out of my depth, I ask in forums like TH, particularly when the object is to gain a broad overview of something and the alternative is to spend hours or days gathering information from the internet.

Anywho, glad to offer an answer, of sorts 😆

Thanks.

 

[PiersPlowman]

Another thing to consider is that some factory overclock cards allow for higher power settings than stock.

Good point. On my factory over locked RX 480, the power limit for the core is upped to 145 Watts from 110. In practice this means that the card can sustain stock boost clocks as if they were the base clock.

 

[VizzieTheViz]

I’d just see how much higher the gpu is clocked compared to the stock card.

The percentage difference in clockspeed is the very most an overclocked card can be faster compared to stock in an absolutely perfect scenario.

When actually playing games the difference will be (far) less because the gpu clockspeed is seldom the only thing holding performance back.

Generally speaking get a non-stock gpu for the better / more silent cooling if you care about that. That’s the thing you you might actually notice.

Performance differences can be measured but I highly doubt you’ll notice them while gaming. Get a higher tier GPU is you need extra performance, not a slightly overclocked card with the same gpu.