Best GPU that doesn't need a pcie power slot.

[Bryf50]

I didn't mean that all connectorless RX 460 was equal to the 2GB, I simply meant that the connectorless RX 460 was equal to the 2GB version in the graph I linked, since there was also a 4GB version in said graph which does have a 6-pin connector.

But by all means if you know of another site that has tested both the connectorless 950 and a connectorless RX 460 (including the card you own), I would love to see it.

The 4GB Sapphire card in your reviews has a boost clock that is only slightly different than the connectorless 4GB Gigabyte card (1212mhz vs 1250mhz) and the same memory clock. The card I have holds it's boost clocks just fine so there is no reason to assume it should inherently perform significantly worse.

 

[antihelten]

Its semantics and i dont intend to move goal post. The question is what is the fastest but what we want to give advice about is what is the smart and best card to buy.
We live life forward and the gfx we buy will play both old and comming games.
Comming high perf aaa games is where the gfx is stressed. Bm and reviews should reflect that. They seldom do.
I will always recommend eg a standard 1070 instead of a very high clock 980ti. And a 460 to a 950.
Maxwell will eg get nothing that will even resemble asynch capability in games and in one year it will hurt it like we see in Doom. It will only go one way.

Personally i would never bother upgrade for like 50% so i would either wait for 1050 or better take the chance with a 1060 6gb. Or a voltage underclocked ref 470 if budget is tight and more nerd fun is needed.

I don't think the difference between on the one hand saying that GPU A is faster than GPU B and on the other hand trying to predict that GPU A may be faster than GPU B in the future is semantics. It's obviously two completely different claims. If you meant the second claim, then that is what you should have written and I would have had no objection.

Either way I agree that waiting for the 1050 is the prudent thing to do at this stage.

The card I have holds it's boost clocks just fine so there is no reason to assume it should inherently perform significantly worse.

And I'm sure that if this is the case then it will also show up in reviews, but unfortunately the Computerbase.de reviews are still the only ones I know of that have tested both connector less cards.

 

[krumme]

Juat looked at the ot headline and it actually says "best gpu card" and in that case its imo pretty simple the answer is 460. Without going into semantics
But for most part we are actually framing it the wrong way. Like whats is the fastest. Like we can ask what the fastest car is. But actually best card is a better way to do it.
Then comes the problem of anticipating how things is eg in a year or two. But imo its similar to the situation where benchmarking without eg minimums, frametime variance or subjective gameplay because argumenting its either to costly or subjective. We need to cross that line. But the info is spread all over. But its there.

Perhaps some of you guys should make some sort of meta review site collecting and presenting all this info. Make the framework.

Someone reading this will be millionaire

 

[Blue_Max]

It's been my experience that so long as it's a card right on the edge of not needing an extra power connector at all, the SATA-to-6pin works perfectly. My boy is using one now on a superclocked 750TI that requires the 6-pin and I'm using a GTX 960 off a single molex 4pin to 8pin connector (rather than the preferred 2x4pin to 8pin).

I'd say any current GTX 950 (not overclocked) and maybe even a 960 will work just fine.

 

[happy medium]

wait a few weeks for the gtx1050. It will be 10 to 20% faster than the rx 460 and will not need a power connector. It should cost about 139$ for the 4gb card.

 

[antihelten]

Juat looked at the ot headline and it actually says "best gpu card" and in that case its imo pretty simple the answer is 460. Without going into semantics
But for most part we are actually framing it the wrong way. Like whats is the fastest. Like we can ask what the fastest car is. But actually best card is a better way to do it.
Then comes the problem of anticipating how things is eg in a year or two. But imo its similar to the situation where benchmarking without eg minimums, frametime variance or subjective gameplay because argumenting its either to costly or subjective. We need to cross that line. But the info is spread all over. But its there.

Perhaps some of you guys should make some sort of meta review site collecting and presenting all this info. Make the framework.

Someone reading this will be millionaire

I agree that just using the average FPS to judge GPUs is a flawed approach, however I would strongly opine against the value of using minimum FPS values. Minimum FPS is a largely useless metric given that the majority of the time it is just an outlier, and these days no one should really be using them when it is so easy to use metrics like 99th percentile or 95th percentile values instead.

Regarding a meta review site, I've personally always been partial to 3Dcenter who collates a large number of launch review with any new GPU and calculates the average performance. Here's their RX 460 analysis.

 

[krumme]

I agree that just using the average FPS to judge GPUs is a flawed approach, however I would strongly opine against the value of using minimum FPS values. Minimum FPS is a largely useless metric given that the majority of the time it is just an outlier, and these days no one should really be using them when it is so easy to use metrics like 99th percentile or 95th percentile values instead.

Regarding a meta review site, I've personally always been partial to 3Dcenter who collates a large number of launch review with any new GPU and calculates the average performance. Here's their RX 460 analysis.

Thanx for the link.
I perfectly agree minimums is not needed in general when 99th percentile or 95th percentile can be used. They are far better.
But we have to be carefull because it depends on the situation. Take a game like bf4, 64 man server, eg Shanghai. There is some situations where cpu becomes the limiting factor because of the API even for 4ghz intel quad core. Reviews and the 99th percentile should reflect that.
I am sure we will have the same discussion when DX12 gets fixed in bf1. And imo its important to get fps where they are needed most, and thats when the battle becomes most intense, and there is a lot of destruction/physics at the same time.
To be frank i dont really care if my textures is medium or ultra. I simply dont feel it while gaming. It doesnt add to the enjoyment. But dips of fps is imo a pain. Its a personal matter ofcource. But imo reviews dont reflect real world use of how the cards is going to get used.

 

[Arachnotronic]

I will always recommend eg a standard 1070 instead of a very high clock 980ti. And a 460 to a 950.
Maxwell will eg get nothing that will even resemble asynch capability in games and in one year it will hurt it like we see in Doom. It will only go one way.

Personally i would never bother upgrade for like 50% so i would either wait for 1050 or better take the chance with a 1060 6gb. Or a voltage underclocked ref 470 if budget is tight and more nerd fun is needed.

+1

 

[antihelten]

Thanx for the link.
I perfectly agree minimums is not needed in general when 99th percentile or 95th percentile can be used. They are far better.
But we have to be carefull because it depends on the situation. Take a game like bf4, 64 man server, eg Shanghai. There is some situations where cpu becomes the limiting factor because of the API even for 4ghz intel quad core. Reviews and the 99th percentile should reflect that.
I am sure we will have the same discussion when DX12 gets fixed in bf1. And imo its important to get fps where they are needed most, and thats when the battle becomes most intense, and there is a lot of destruction/physics at the same time.
To be frank i dont really care if my textures is medium or ultra. I simply dont feel it while gaming. It doesnt add to the enjoyment. But dips of fps is imo a pain. Its a personal matter ofcource. But imo reviews dont reflect real world use of how the cards is going to get used.

The problem with minimum FPS is that while it can often give a hint about dips that can ruin gameplay even though the average FPS looks fine, it can just as often be from a dip that has absolutely zero influence on gameplay.

For instance if you're playing BF4 for 2-3 hours at a smooth 60 FPS average, and during these 2-3 hours you experience a singular dip down to 15 FPS whilst re-spawning, it is safe to say that this would have zero influence on gameplay. On the other hand if you are playing and you have multiple dips down to 15FPS during firefights, then this would have a clearly detrimental effect on gameplay. With minimum FPS measurements, you won't be able to tell the difference between these two scenarios, whereas with 95/99 percentile you will.

So I would never say that review sites should go without some sort of measurement of dips/smoothness (another option is simply including frame graphs), merely that minimum FPS is quite possibly the worst possible measurement to use for this.

 

[krumme]

The problem with minimum FPS is that while it can often give a hint about dips that can ruin gameplay even though the average FPS looks fine, it can just as often be from a dip that has absolutely zero influence on gameplay.

For instance if you're playing BF4 for 2-3 hours at a smooth 60 FPS average, and during these 2-3 hours you experience a singular dip down to 15 FPS whilst re-spawning, it is safe to say that this would have zero influence on gameplay. On the other hand if you are playing and you have multiple dips down to 15FPS during firefights, then this would have a clearly detrimental effect on gameplay. With minimum FPS measurements, you won't be able to tell the difference between these two scenarios, whereas with 95/99 percentile you will.

So I would never say that review sites should go without some sort of measurement of dips/smoothness (another option is simply including frame graphs), merely that minimum FPS is quite possibly the worst possible measurement to use for this.

A perfectly agree on this. I would prefer if more site did like H. That is to take a spin in the game. Perhaps not as a prime evaluator as H do and for a few games, but more just a short spin to confirm if the bm is representative.

 

[Bacon1]

The problem with minimum FPS is that while it can often give a hint about dips that can ruin gameplay even though the average FPS looks fine, it can just as often be from a dip that has absolutely zero influence on gameplay.

For instance if you're playing BF4 for 2-3 hours at a smooth 60 FPS average, and during these 2-3 hours you experience a singular dip down to 15 FPS whilst re-spawning, it is safe to say that this would have zero influence on gameplay. On the other hand if you are playing and you have multiple dips down to 15FPS during firefights, then this would have a clearly detrimental effect on gameplay. With minimum FPS measurements, you won't be able to tell the difference between these two scenarios, whereas with 95/99 percentile you will.

100% agree, It's why I like line graphs over bar charts for FPS. You can see where dips occur.

The first time I noticed how bad this was, was the ROTTR benchmark. Using DX11 it would sometimes give minimums in the single digits, but it never happened during the actual benchmark "gameplay", so I'm assuming it must have happened during the loading between scenes as watching it with afterburner / fraps / presentmon didn't show those dips either during gameplay. However when you get the final FPS report at the end, it would show these minimums that never showed up in the actual benchmark "gameplay" part.

 

[Headfoot]

I'd wait for the 1050

 

[moneer]

I'd wait for the 1050

wait a few weeks for the gtx1050. It will be 10 to 20% faster than the rx 460 and will not need a power connector. It should cost about 139$ for the 4gb card.

I don't think the difference between on the one hand saying that GPU A is faster than GPU B and on the other hand trying to predict that GPU A may be faster than GPU B in the future is semantics. It's obviously two completely different claims. If you meant the second claim, then that is what you should have written and I would have had no objection.

Either way I agree that waiting for the 1050 is the prudent thing to do at this stage.



And I'm sure that if this is the case then it will also show up in reviews, but unfortunately the Computerbase.de reviews are still the only ones I know of that have tested both connector less cards.

Is the 1050 ti still considered the way to go? Is it worth the premium over the 1050 or 460?

 

[happy medium]

Is the 1050 ti still considered the way to go? Is it worth the premium over the 1050 or 460?

If you want the fastest card with no power connector it is.

 

[Headfoot]

Before you conclude you don't have a power connector available - do you have any molex 4-pin connectors available? You mention you have SATA.

 

[moneer]

Before you conclude you don't have a power connector available - do you have any molex 4-pin connectors available? You mention you have SATA.

Nope. No Molex. Sadly.

 

[Blue_Max]

You can get SATA to 6-pin adapters, provided you do the research to make sure power draw isn't excessive.
https://www.amazon.com/Branded-8inch-15pin-Express-Power/dp/B005NJXY7O