Opinions on Nvidia's new GPU binning system?

[3DVagabond]

Intel "resort" to these solutions too, and they have much more binning.
To me it seems like a sensible move.
The only people who can really complain are overclockers, but then
1) When you get the chip you have a better idea what it's going to be like because you have an accurate voltage
2) These chips are going to be factory overclocked so you could get one of them instead.

Really the only difference is possibly a higher chance of a "dud" (one that requires high volts for stock), but then that's quite likely anyway since partners will be selecting the best chips for their overclocker models, and overclocking is always YMMV anyway.

My comment was directed at blaming TSMC for nVidia going to this. I think it's just something nVidia has to do because their chips aren't binning very well.

 

[Ben90]

EX: lets say you have 3 chips, chip A requires 0.9v to run at stock speed, chip B requires 1.0v, and chip C requires 1.1v.
used to be, nvidia would set the voltage on chip A and B to 1.0v, stock speed, and sell them to you. they consumed the same amount of power, and if you lucked out you got chip A.
Chip C would be clocked lower, and sold for less money as a lower end card.

Or they could just run chips A and B at 1.1v and sell all three as whatever they want, which is probably what they did already. I mean why wouldn't they?

Don't see much difference to before, only thing is, if you're lucky you'll end up with a lower voltage chip that's a bit cooler and needs less power. But it's not as if all chips OCed equally good beforehand.

Voo is exactly right, without dynamic voltage binning, Nvidia is just going to set the voltage to the lowest common denominator. Nothing is changing for the worse. In your example Taltamir, under the old system, all three cards would be sold at 1.1v. Lets introduce card D that needs 1.2v to run at clock speeds, for arguments sake lets say 1.1v is the highest Nvidia will use. Under either system card D is going to be thrown out or binned lower so its not like this is a horrible horrible thing.

I'm supervised at the responses here actually, you guys should know that power consumption varies even within the same voltage. And even more so is that VID != overclockability. I cannot state this enough, actually here:
VID != overclockability
VID != overclockability
VID != overclockability
Some of the best overclocking chips from Intel and AMD can barely pass for the lowest speed bin without using too much voltage, however when their TDP is ignored, these chips shoot to the moon using better cooling.

The ONLY possible way this new binning process could be bad is if Nvidia sold chips that it knew were running at a dangerous voltage to begin with, but that has nothing to do with the binning, and this business decision could have been made quite easily without binning.

 

[Phynaz]

Which CPUs have different voltages for the same model number? I've never heard of that.

That i7 you're running for one.

All current generation CPU's have their voltage set at the time the chip is tested.

 

[Phynaz]

V
VID != overclockability
VID != overclockability
VID != overclockability

:thumbsup:

 

[happy medium]

My comment was directed at blaming TSMC for nVidia going to this. I think it's just something nVidia has to do because their chips aren't binning very well.

And mabe they are not binning very well because they are using A1 silicon for the gf104 chips?

"As you would expect the new GF104 GPU is quite a bit smaller than the original GF100 - and it is an odd rectangular shape based on the dimensions of the heat spreader resting over it. Also note that we are testing A1 silicon; a notable achievement for anyone familiar with processor design. "


http://www.pcper.com/article.php?aid=954&type=expert&pid=1

 

[DaveSimmons]

I dislike it slightly since it adds a little uncertainty to the power use, heat generation and noise of the card at stock speed. Some cards are better than others at stock, it's now the luck of the draw.

But based on the AT review the choice of fan is much more important, and it should be OK as long as you run away from Zotac.

 

[RussianSensation]

But based on the AT review the choice of fan is much more important, and it should be OK as long as you run away from Zotac.

It's not Zotac necessariliy. It relates to the external exhaust cooler layout similar to 5850. For example, EVGA with an external exhaust:

http://www.evga.com/PRODUCTS/enlarge...P3-1361-TR&I=7

 

[tviceman]

None of this is any different than buying a CPU. If you buy the exact same CPU three times chances are each one will be operating at slightly different stock voltages and their max overclocks will be different and will require different voltage tweaks.

 

[GaiaHunter]

I'm supervised at the responses here actually, you guys should know that power consumption varies even within the same voltage. And even more so is that VID != overclockability. I cannot state this enough, actually here:
VID != overclockability
VID != overclockability
VID != overclockability
Some of the best overclocking chips from Intel and AMD can barely pass for the lowest speed bin without using too much voltage, however when their TDP is ignored, these chips shoot to the moon using better cooling.

So basically you are saying if you give a chip enough voltage, enough power and then have good cooling to remove the extra heat created it can actually reach the speeds of chips that consume much less power and produce much less heat and so require much less cooling.

Ok.

The ONLY possible way this new binning process could be bad is if Nvidia sold chips that it knew were running at a dangerous voltage to begin with, but that has nothing to do with the binning, and this business decision could have been made quite easily without binning.

And why is it good?

If all the cards in a same model use similar voltages, all have of them will have similar power consumptions and produce similar heat.

Basically the old process was based on binning for speed at a certain power consumption plateau.

The new binning is simply built for speed.

You call that an improvement?

I don't.

 

[Voo]

So basically you are saying if you give a chip enough voltage, enough power and then have good cooling to remove the extra heat created it can actually reach the speeds of chips that consume much less power and produce much less heat and so require much less cooling.

It's a well known fact that chips with high leak voltage OC better than normal chips, as long as you can cool them adequately (well at least that's what I picked up from our CPU OC gurus so I assume the same holds for GPUs). Though for a moderate OC I'd also take the lower voltage chips.

But the thing is: If Nvidia thinks it can sell all chips <= 1.1V as GTX Y then they'd normally bin all chips for that voltage and sell them as such. Now they're selling chips that run fine @0.9V at .9V instead of 1.1.
Why should that be such a bad thing? You're still getting exactly the same chips, it's just that the default values are more sensible. If you don't want it you can always assume the worst case and base your decisions on that.

 

[GaiaHunter]

The bit I see as a con is that before it was "YMMV regarding OC". Now is "YMMV regarding OC and stock performance".

 

[Lonyo]

The bit I see as a con is that before it was "YMMV regarding OC". Now is "YMMV regarding OC and stock performance".

Stock "performance" only changes with respect to power consumption, and most people would probably assume the TDP = power consumption, and TDP is probably based on the highest voltage, so you're not losing per-se.
It's still exactly the same as we've had with CPUs for many years though, so I don't see why anyone is kicking up a particular fuss over it. Higher voltage doesn't mean lower stock performance, just higher (relative) power consumption, but still less than the TDP (or it should be...)
Power is a consideration, but not a performance consideration, and they should all be 150/160w or below.

 

[Voo]

The bit I see as a con is that before it was "YMMV regarding OC". Now is "YMMV regarding OC and stock performance".

Yeah but since Nvidia still wants to sell as many cards as before (which I think is safe to assume), the difference is that before you always got the worst case and had to downvolt yourself (if that was possible at all).

Now, the worst case is still exactly the same as before, but if you're lucky you'll get a lower voltage card, which is for most people a nice thing. So where's the disadvantage to that method?
Earlier you'd always get a 1.1V card, now you may get in 10&#37; of all cases a 0.9V card (imaginary numbers) - if you don't want to stress your luck and want to be one the safe side, just assume that you'll get the worst possible voltage and nothing has changed at all.


That all assumes that Nvidia does not start selling cards they wouldn't sell before, but since I'd think they tried to bin cards as high as possible already that should stay the same.. if not, well they could do the same thing with only one voltage class as well.

 

[pmv]

I suppose it depends whether its a case of some lucky cards getting set to a lower voltage than they previously would have, or of some cards that would otherwise not have been used at all being made to work at the expensive of running very hot.

Which way round is it?

 

[exar333]

This has worked (well) for CPUs for a long time.....

/thread?

 

[Voo]

I suppose it depends whether its a case of some lucky cards getting set to a lower voltage than they previously would have, or of some cards that would otherwise not have been used at all being made to work at the expensive of running very hot.

Which way round is it?

We can't know for sure, but from a logical standpoint I'd say it's #1. After all if Nvidia is fine selling 10% or whatever cards at xV, they shouldn't have a problem selling every card at that voltage. And I don't think Nvidia would initially bin cards lower than they thought they could (that says obviously nothing about artifical binning for market reasons), so #1 is the only one that makes sense imho

 

[GaiaHunter]

We can't know for sure, but from a logical standpoint I'd say it's #1. After all if Nvidia is fine selling 10% or whatever cards at xV, they shouldn't have a problem selling every card at that voltage. And I don't think Nvidia would initially bin cards lower than they thought they could (that says obviously nothing about artifical binning for market reasons), so #1 is the only one that makes sense imho

Actually most likely is number #2. The best chips of the wafer might as well have the 384 cuda cores functional.

AMD, for example, basically just grabs the chips that require lower voltage and use them as 5970.

A policy like this only means they have chips all over the place.

In CPUs basically you just downclock the chips that require more voltage and sell them as lower SKUs.

Additionally there appears to be some misunderstanding in here.

This isn't a big issue - people that pointed the downsides of this simply were responding at the OP, which thought this was a better policy.

Clearly this can be seen from many points of view.

 

[Voo]

Actually most likely is number #2. The best chips of the wafer might as well have the 384 cuda cores functional.

AMD, for example, basically just grabs the chips that require lower voltage and use them as 5970.

Well my point is: If Nvidia thought they couldn't sell chips that needed, lets say 1.2V, for whatever reasons, why should they do that now? And if they decided that 1.2V chips are fine as well, why would it matter if they sold 30% or 100% of all chips with that voltage?


But yeah if they'd do #2 then it'd be something completely different, I agree.

 

[sandorski]

This has worked (well) for CPUs for a long time.....

/thread?

Has it? This seems fundamentally different, as CPU Models with different Power Requirements are clearly Marketed as such. My 1055T is a 125Watt version, a 95Watt version is also available(I believe it is now), but I knew what I was getting when I Ordered it.

This does appear to be what Nvidia is doing here.

 

[SHAQ]

This has already been the norm for CPU's(different VID's). It shouldn't be that surprising that GPU's now do this.

 

[Ben90]

So basically you are saying if you give a chip enough voltage, enough power and then have good cooling to remove the extra heat created it can actually reach the speeds of chips that consume much less power and produce much less heat and so require much less cooling.

Ok.
Talk to Rubycon or head over to XS where people go through tens and tens of chips looking for that gem. Basically someone who handles a lot of processors will tell you that VID doesn't dictate what a chip will do. Some chips with a low VID will use an insanely low amount of power, but won't over clock past a certain limit despite how much voltage you give it. Others will look like horrible candidates at stock due to the high VID but will shoot to the moon once you start pumping voltage. Believe it or not, for extreme overclocking the latter group usually produces best results. Leaky chips in general scale well once you start pushing voltage.


And why is it good?

It is good because now every chip runs at the lowest power needed instead of using an artificially high voltage because of gimp chips.

If all the cards in a same model use similar voltages, all have of them will have similar power consumptions and produce similar heat.

I don't understand why you think cards being purposely overvolted for the sake of being similar is a good thing. If you do your research who gives a shit if Anand's ES uses 10 watts less than yours at load?

The only-only-only downside to this system is if you have to be within 10 watts because of PSU requirements, but if your cutting it that close you should probably be looking at a new PSU.

Anyways, even using the same voltage there will be variances. See a i7 975 vs 920

Basically the old process was based on binning for speed at a certain power consumption plateau.

The new binning is simply built for speed.

You call that an improvement?

I don't.

My responses in bold. If your upset about not having your card overvolted at stock, just make sure you get an Asus and pump voltage to it. Problem Solved.

 

[Ben90]

Has it? This seems fundamentally different, as CPU Models with different Power Requirements are clearly Marketed as such. My 1055T is a 125Watt version, a 95Watt version is also available(I believe it is now), but I knew what I was getting when I Ordered it.

This does appear to be what Nvidia is doing here.

While you are correct there are low powered specially binned processors just for that task. There are processors that are of the same batch using different VIDs. Some i7 920s use 1.225 stock, some use 1.2, some use 1.25.

 

[sandorski]

While you are correct there are low powered specially binned processors just for that task. There are processors that are of the same batch using different VIDs. Some i7 920s use 1.225 stock, some use 1.2, some use 1.25.

Interesting, I did not know this. Is it just Intel then? Only certain Models?

 

[Ben90]

To be honest I couldn't tell you if any other processors do that. I just know Nehalem does it because I couldn't figure out for the life of me what my stock Vcore was supposed to be until I found out that little fun fact.

Wow I just got a bloody nose, the fourth one in like 4 days. Damn this new high altitude town I live in.

 

[Wreckage]

99.9&#37; of end users probably don't know or care what voltage their chip is. Nor should they.

Simply another mountain out of a mole hill.