Been using an EVGA GTX295 (@ 650 core / 1401 shader / 1100 memory) since January last year and not even seen a hint of microstutter in that time.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Is there going to be a dual fermi card?
- Thread starter 3t3P
- Start date
Been using an EVGA GTX295 (@ 650 core / 1401 shader / 1100 memory) since January last year and not even seen a hint of microstutter in that time.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
THe problem is, the GTX275, well, they had dozens of gtx275's left in warehouses, and by dozens, I really mean, 10's or 100's of thousands.
This is the problem they had numbers to throw at projects like these, and they could sell those classified editions to a few suckers for huge prices.
There just isn't the chips, theres no warehouse full of GT200b's after a 5870 was launched with badly priced no where near the performance chips looking for new odd projects to be thrown into. Theres maybe 2000-3000 480gtx's likely to come from this week through the next few months, but they've "presold" 10,000's of 480gtx's to AIB's who have promised them to distributors and stores.
Using up a few chips to eventually sell a card in the thousands at a profit, vs the current situation, which would be using up a significant percentage of their total supply, to design a new product that would make several hundred pounds in losses if it could ever sell, is madness.
Unless they plan on harnessing the power of sun to power the dual gpu version, which we should expect to be on the cards from a historical point of view, there will not be one, they need to go through a die shrink cycle or something to lower the heat-output per square cm of die, without some concession of power demand and raw heat output, two dies on one card isnt happening this generation, at least not for a very long time
Last edited:
Permabanned
- Joined
- 22 Feb 2010
- Posts
- 1,629
And i heard they are makeing a mini LHC that we can all make mini black holes with in our Pc's!
Oh wait thats what a 480GTX is as it just Sucks power
Oh wait thats what a 480GTX is as it just Sucks power
Nvidia's power solution for the 2x480, based on 2x285 stuck together and rebranded as a 495, running 10% faster than 480 and 25% hotter.
makes sense?
http://www.lhc.ac.uk/
makes sense?
http://www.lhc.ac.uk/
It's all being going wrong for nVidia since they took thier eye off ATI.
My guess is there won't be a GTX495. It's technically and physically impossible, if nVidia had done some decent research on 40nm and not started testing it on low end cards months to late the story would be hugely different.
My guess is there won't be a GTX495. It's technically and physically impossible, if nVidia had done some decent research on 40nm and not started testing it on low end cards months to late the story would be hugely different.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
Its not even close to physically impossible, board specs are meant to be broken, just because a slot "shouldn't" use more than 300W, doesn't mean it can't. THe 5970 only uses less than 300W stock, overclocked it draws quite a lot more than that, who really cares.
They can indeed make a dual gpu version, however if they keep it as dual slot, or even pushed to triple slot, it will require "proper" cooling, like the proper fans on the newer 5970 4gb/proper 5870 clocks editions.
Keep in mind I just went from a 5850 with stock cooling, overclocked and overvolted pushing 80-85C in a hot room keeping the fan pretty quiet, to the prolimatech cooler, running at 35C load. Theres PLENTY of room for better cooling on these cards, dumping all that hot air in my case increase cpu temps, about 2C max and didn't effect my cpu overclock at all.
Cooling, heat, they aren't a problem, power's only a technicality in that someone somewhere decided it has to run under 300W, why, can anyone tell me why it matters if it goes over? IT doesn't matter that a 5970 goes over 300W overclocked, does it matter if it goes over by 100W, or 300W?
THe problem is cores, as I said, if you've got, maybe in total even for the biggest partners, 500 480gtx's, why would you waste 20 on designs and testing, and only have another 50 to sell the cards? These products take a few months to test and qualify for release, we'll have 6xxx series cards.
Lastly, it also comes down to profitability, dozens of guys working on pcb's, power, cooling, putting in very small orders for very expensive custom coolers, its very expensive to get a new card out, with basically zero chance of making any of that money back, let alone turning a profit.
They already have the ability to say "look we have the fastest card" and when other people mention the 5970 it rather proves their point and lets them say "yeah but look how crap AMD are, they need a 2nd gpu just to beat out top end card".
Its all spin, a dual gpu card to be honest, clocked down further than the 5970 had to clock will likely not beat it. Remember 2 FULL 480gtx's in sli trades blows with a 5970, drop the clocks on the 480gtx's by 25-30%, they likely actually can't beat a 5970, right now they can spin its only faster because its dual gpu, if their 495gtx can't beat it, they can't spin that as anything but a failure.
As I've said though, Nvidia are screwed a respin won't help things either, they quite literally can't fit more shaders/core logic in at 40nm, AMD have 100W of power headroom, and 50% of the core size to add into the 6xxx series if they want to. Nvidia have no where to go at 40nm, AMD can grow around 50% and still be cheaper/higher yields/better power than the 480gtx.
They can indeed make a dual gpu version, however if they keep it as dual slot, or even pushed to triple slot, it will require "proper" cooling, like the proper fans on the newer 5970 4gb/proper 5870 clocks editions.
Keep in mind I just went from a 5850 with stock cooling, overclocked and overvolted pushing 80-85C in a hot room keeping the fan pretty quiet, to the prolimatech cooler, running at 35C load. Theres PLENTY of room for better cooling on these cards, dumping all that hot air in my case increase cpu temps, about 2C max and didn't effect my cpu overclock at all.
Cooling, heat, they aren't a problem, power's only a technicality in that someone somewhere decided it has to run under 300W, why, can anyone tell me why it matters if it goes over? IT doesn't matter that a 5970 goes over 300W overclocked, does it matter if it goes over by 100W, or 300W?
THe problem is cores, as I said, if you've got, maybe in total even for the biggest partners, 500 480gtx's, why would you waste 20 on designs and testing, and only have another 50 to sell the cards? These products take a few months to test and qualify for release, we'll have 6xxx series cards.
Lastly, it also comes down to profitability, dozens of guys working on pcb's, power, cooling, putting in very small orders for very expensive custom coolers, its very expensive to get a new card out, with basically zero chance of making any of that money back, let alone turning a profit.
They already have the ability to say "look we have the fastest card" and when other people mention the 5970 it rather proves their point and lets them say "yeah but look how crap AMD are, they need a 2nd gpu just to beat out top end card".
Its all spin, a dual gpu card to be honest, clocked down further than the 5970 had to clock will likely not beat it. Remember 2 FULL 480gtx's in sli trades blows with a 5970, drop the clocks on the 480gtx's by 25-30%, they likely actually can't beat a 5970, right now they can spin its only faster because its dual gpu, if their 495gtx can't beat it, they can't spin that as anything but a failure.
As I've said though, Nvidia are screwed a respin won't help things either, they quite literally can't fit more shaders/core logic in at 40nm, AMD have 100W of power headroom, and 50% of the core size to add into the 6xxx series if they want to. Nvidia have no where to go at 40nm, AMD can grow around 50% and still be cheaper/higher yields/better power than the 480gtx.
Power matters because Nvidia will start to break things that don't belong to them - some of the reviewers were already reportinging that the PCI-E power cabling was getting hot to the touch and the plastic sheething was getting soft. It would be embarrasing to say the least if Nvidia were told that they couldn't say the cards were PCI-E compatible due to grossly exceeding the power specification.
They've got nowhere else to go as regards power consumption.
They've got nowhere else to go as regards power consumption.
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 33,188
TO be honest, that review was just being pathetic, cards running at 80C plus isn't new, hot air means plastic on some cables will get a bit more flexible, it wasn't a short in the cable causing a failure.
Does anyone have a problem witha 5970 hitting 400W? I don't, almost no ones mentioned it at all, CPU's aren't rated to draw more than 140W on an AMD board, anyone that overclocks a quad is almost most certainly doing so.
It really doesn't matter in any which way if a gpu inside the case draws extra power from an extra cable from a PSU more than capable of supplying the power.
Theres people out there who put LN2 in their systems, we overclock past specs all the time, OCUK sell pre-overclocked chips and bundles that exceed the specs.
Personally I would never run a dual, or a single Fermi, why would I want to get 5% more performance for 50% more cost at 60% more power and more heat issues to deal with in my case and more noise. I don't want to, but I could should I want to.
As said, my 5850 overclocked way beyond spec, runs plenty hot, I stick a decent 3rd party cooler on it, its running 36C full load, theres PLENTY of headroom for them to add a great cooler that would drop load temps to 50C, or below, if its going out of the pci-e spec power wise, do you think its really a problem going to a triple slot cooler, especially as their competition has already done so.
I haven't actually checked, aren't the 4gb higher clocked 5970's above a 300W power draw now anyway even at stock?
Basically, its easy to make, theres no magic, its a 295gtx with a different gpu and pcb layout, nothing more or less, it needs cooling, they can make it, it needs power, theres dozens of PSU's that are rated high enough, it needs a case with decent airflow, theres hundreds of them.
The only thing it doesn't have, is Fermi cores to put on the things, and thats where it all falls down.
I think you're missing the point, yes they can build it, power it, and cool it. But it's reaching the point that they will no longer being able to claim it's PCI-E compatible because it is miles away from the PCI-E specification, the PCI-E Foundation or whatever they're called could turn round and tell Nvidia that they can't use the PCI-E logo or claim it is PCI-E card.
The PCI-e spec has a thermal/electrical advisory for desktop systems but it doesn't limit the entire card to 300watt - only the amount you can draw from each of the official power connectors, 75watt, 150watt, etc. at no point does it prevent a developer from using another power source - only the above advisory as to points they should adhere to.
A dual GPU fermi card is certainly possible as DM says, even if its slightly impractical.
With a revised PCB, some decent 3rd party cooling rather than your average reference jobbie it should be possible to pull power useage down considerably and cooling efficency much increased. A dual 416SP card would work pretty well and be very competitive.
A dual GPU fermi card is certainly possible as DM says, even if its slightly impractical.
With a revised PCB, some decent 3rd party cooling rather than your average reference jobbie it should be possible to pull power useage down considerably and cooling efficency much increased. A dual 416SP card would work pretty well and be very competitive.
I doubt that would happen, for ratification you only need to prove that you've taken the thermal/electrical guide into consideration in most cases.
Associate
- Joined
- 10 Nov 2009
- Posts
- 1,099
- Location
- London
When they are released there may only be two of them 
Hope you can wait a long time dude. For your own saftey, please don't hold your breath
I think DM has sumed it all up quite nicely, forget any technical or financial reason it can or can't be done, they just don't matter. It's simple numbers of working cores that's the issue.
Is there any news of Nvidia doing another production run?