GPU's - Isn't it about time they shipped bare?

[Agent]

So I've sorted out a few GPUs tonight. Some of them are a few years old, but have huge heatsinks on them. Some are very high quality.....but I can't do a thing with them.
Why? Because the offsets of the holes, memory and core GPU often change between generations. In some cases, even the same generation can differ between manufacturers. You can forget ever swapping between Nv/AMD.

Why is this still going on? The changes from my 260 GTX to my 760 GTX are not huge. A bit of forward thinking on the layout of the core, memory and VRMs could mean the same heat-sinks could be used for generations. The only thing you need to worry about would be the ability of the HSF to handle to heat output - which could easily be handled through basic specifications of the HSF.

They could use a heatspeader on the GPU (like CPU's) and then standoffs if really needed - again like they do with CPU heat-sinks that work on Intel, AMD and across tons of sockets. There are loads of ways to do this, none are particularly complex, and they've been used in many other areas of tech for years. Incorrect fitting / burnout can be easily solved by the thermal cutoffs that are embedded in the chips, so ensure no damage is caused.

I suspect the answer is that so the GPU manufacturers can differentiate themselves from each other. But really, this is getting silly. It's such a waste. I dread to think how many of them end up in landfills after only a few years of use.

Am I the only one who wants to buy the GPU and HSF for it separately? Not only for easier choice of the quality of the HSF, but so it can be used on my next card...and the one after?

As for the different ranges - well, you simply have 3 types of HSF. Ones for low power cards like the 750Ti, medium for 760GTX, highend for anything more. The sizes have stayed around the same for many years now - and they can't get much / any bigger without screwing up ATX fitting.

Am I mad?

[AETAaAS]

I wonder now, do manufacturers who do their own custom PCBs tweak the design so that their cooler works across many cards? Things like Gigabyte's Windforce, Asus' DirectCUII and MSI's TwinFrozr whose coolers look basically the same no matter the GPU; they may save some time, effort and money not having to make a custom cooler for each and every card. Watercooling guys probably know better than I since waterblocks are a bit more pernickety when it comes to the holes. I just leave the air cooler on and slap it in the slot. Slap... slap... slap dat slot.

[Agent]

I wonder now, do manufacturers who do their own custom PCBs tweak the design so that their cooler works across many cards? Things like Gigabyte's Windforce, Asus' DirectCUII and MSI's TwinFrozr whose coolers look basically the same no matter the GPU; they may save some time, effort and money not having to make a custom cooler for each and every card.

Yep they do. I've transplanted a Frozer II a couple of times.

Now if only they'd all agree on a standard....

[KeyboardDemon]

Am I mad?

Quite possibly but you do have a great point.

When I read what you had written, the first thing that I thought was all of those people that want to water cool their GPUs are paying extra for a cooler that's been fitted to their GPU only to then have to remove and store that cooler so that they can reapply it to said GPU when they sell it. If there was a market for bare GPUs then they could buy the GPU at the lower price and sell it on to someone that is looking to buy a used GPU from that same market place.

[Saracen]

Hmm. The cynic in me surfaces, here. Is it about warranty invalidation?

[Agent]

Hmm. The cynic in me surfaces, here. Is it about warranty invalidation?

Quite possibly, but it's not an issue for CPU's or Motherboards (with their 1000+ pins), which are much easier to damage than a bare GPU. Heck, CPU's used to have the HSF placed directly on the core. While a simple spreader could solve this if needed (some GPU generations have done this anyway), I can't help feeling that the warranty issue isn't really an issue. It seems more of a self imposed limitation by the manufacturer. I mean, what are users going to do exactly if they can see the PCB? :\

[watercooled]

I've been thinking along those lines too - aside from being wasteful, being lumbered with fixed coolers is negative from a performance standpoint when comparing to CPUs.

It just seems backwards, that we have a huge amount of very efficient coolers designed to cool that 95W CPU, while we have cheap OEM blowers strapped to the 300W GPU.

Something like a GPU socket on a motherboard could prove very interesting, but unfortunately it would probably take an awful lot of effort to make that change. We have Intel trying to lock out competing GPUs already, and Nvidia competing against both the major CPU producers. Maybe AMD is the one in a position to do something - it would probably be a good selling point for their platform, especially if they didn't try to lock in AMD GPUs. Then again, even when access is opened up to Nvidia they historically just like to make life harder for everyone by releasing another unnecessary closed, proprietary 'standard'.

[Saracen]

Quite possibly, but it's not an issue for CPU's or Motherboards (with their 1000+ pins), which are much easier to damage than a bare GPU. Heck, CPU's used to have the HSF placed directly on the core. While a simple spreader could solve this if needed (some GPU generations have done this anyway), I can't help feeling that the warranty issue isn't really an issue. It seems more of a self imposed limitation by the manufacturer. I mean, what are users going to do exactly if they can see the PCB? :\

Well, it is an issue for bare CPUs, but dealt with in the T&Cs. It gives quite a bit of lattitude for rejecting warranty claims where overheating was a problem. That's not to say GPUs couldn't use the same escape route, though. Just that it may be the concern.

As might damage by coolers fitted by ham-fisted, all-thumbs muppets that blunder their way into fitting coolers without reading instructions, 'cos RTFM is for wimps, and somebody else'.

I wonder what proportion of people buying video cards upgrade the cooler? I mean, here, on a tech site, sure, lots will. But what about the gazzillions of cards sold to Joe Public by OEMs, small system integrators or the high-street custom builder? My bet is that of overall annual card sales, a pretty small proportion of general users (and a pretty large chunk of enthusiasts) upgrade the cooler.

[Saracen]

As an aside, though, Agent, I do agree with your logic.

A few years ago, I had an argument with the saleman at my local car dealer, because the model I was after now had air-con as standard. He told me it was "free". I laughed. I pointed out that since the previous model (which I owned) they'd added "free" air-con, and a couple of other things, and stuck five grand on the price. "Free", my .... erm .... botty. What kind of an idiot did he take me for?

Sure, it wasn't the salesman's decision to include it and jack the price, but his choice of explanation was, frankly, insulting. And cost him the sale.

I had a similar discussion on another occasion about car stereo. But the "premium" system is "good value" insisted the salesman. Not when I have a very high end custom system waiting to go in it isn't, because whatever you put in, I'll have ripped out immediately. He didn't seem able to understand why I didn't want his over-priced tat that he called "premium", when it'd never be used.

So, I assure you, I know where you're coming from.

[heinsmit]

I think it would be nice to have a separate bare SKU and the current SKUs. Could save a (little) bit of money for those who want to go aftermarket.

[AndyG]

Genius, OP.

[sqwerty]

Would love this, hardest part doing aftermarket is usually removing stock cooler.

[Askew]

A bare SKU would be nice, but let's face it a reference card from the likes of EVGA can require the removal of as little as 4 screws to just pull off the refrence cooler which can then be kept aside for such times as it's needed (perhaps to sell the card to someone who has nothing to put on a bare card).

If the option was available, I'd still order reference cards complete with the cooler.

[Agent]

I think you missed the main point about wastage....

[kalniel]

This is the kind of thing the EU loves to meddle in surely? Standardised, one-size fits all heatsink interfaces will be on their way as soon as brussels installs a gamer into office

[Askew]

I think you missed the main point about wastage....

I can see why you'd think that, perhaps my post wasn't worded very well.

I think it would be nice to have a bare SKU simply because some people will buy a card with the express desire to remove it and apply a waterblock/aftermarket coolers from OEMs like Zalman and Arctic Cooling.

The reason I don't think it's so important is because for me personally I always sell cards when I'm done with them and I think I have a higher chance of making a sale if I have a cooler with each card that can be offered with it so that a potential user who does not have a cooling solution to hand can purchase it.
If you are selling a 2 or more generation old card chances are that the next guy may not be watercooling as those types of users in my experience tend to opt for the latest and greatest cards with sexy waterblocks from the likes of EK and when they're done most likely sell them with the blocks after only 1 generation, those are the users who I'd expect to purchase a bare SKU from the manufacturer.

It is for the reason of onward transfer that I don't think it matters to me whether the coolers are interchangeable between cards or not, I want the cooler regardless so if it only fits that one SKU then so be it.

I do see your point, I didn't miss it, you want to buy cards and fit your own cooling solution without having to pay for a cooler because you could have kept the cooler from your previous card purchases. However as stated in OP there would need to be tiered sizes, small medium and large to accomadate different PCB lengths.
Let's say for instance for Fermi 2 generation you had a GTX 570, and then when Kepler 1 GK104 came along you decided to upgrade to the 670, well there you're going to come unstuck because Nvidia saw in their infinite wisdom to sell that SKU on a tiny PCB, you'd either have to hunt down one of the versions released on the 680PCB (which cost more and could negate your savings in buying a 670/keeping your cooler) or opt for a 670 with a cooler on it because your cooler from last gen simply would not fit the card.

There's a couple more reasons why I can't see such standardization being a solid idea, but the ones above are the biggest, I think it would be required for people to own one of each size and that's more than I've ever owned.

To finish though, I support the idea of OEM's making bare SKUs available, I believe there is a market for it even if coolers are not standardized, if EVGA can consistently sell people Hydro Copper model cards then I'm sure that retailers can encroach on that market by offering said bare SKUs in combo offers with waterblocks and other aftermarket cooling solutions such as closed loop coolers and the brackets to fit them. (actually that solution is a potential way of re-using a cooler, some of the brackets available now can fit multiple cards).