News - NVIDIA GeForce GTX 680 Specifications Revealed?

[CAT-THE-FIFTH]

And Mooseman - the right image in a thread can be funny, but lets not overdo it with old images that have been around for yonks

You PM box is full!!

[Agent]

Here is the best image for this thread



Driverssmivers.

While pie charts like that are pretty and fan the flames nicely in any ATI/nVidia war, they actually show very little.

The article: http://arstechnica.com/hardware/news...ia-drivers.ars

The data points in the table cover an unspecified period in 2007, and Microsoft makes no attempt to break the aggregate data down into which device drivers, specifically, returned the highest number of crashes. If the number of failures were split by month and then graphed, we'd presumably see the number of NVIDIA driver failures per month decreasing as the company slowly brought its driver issues under control.

The data clearly indicates that NVIDIA had a driver problem, but it's impossible to quantify the scope of that problem given the numbers above. NVIDIA holds a greater percentage of the market than ATI, which means that there will inevitably be a higher percentage of NVIDIA driver crashes than ATI driver crashes; however, the degree to which such market share considerations have affected the results above is hard to determine in the absence of more data. There's also the matter of data collection; Microsoft's charts do not clarify if multiple crashes from a single system each counted as separate events. In theory, NVIDIA's proportion of total driver crashes could be inflated by a relatively small handful of systems with severe driver issues.

So:
* 5 year old data
* Higher market percentage = more crashes logged for that manufacturer
* No information on how the crashes were actually logged in regards to multiple (Same system) crashes
* No information on if the crash was due to hardware failure (or overclocking)
* No information on what a 'crash' actually is. A application bombing out? A BSOD / System lockup? A driver restart? and so on....

I'm the last person on earth when it comes to fanboy nonsense, I'll buy what's best price: performance at the time, but seriously, charts like that cause more trouble than they help solve.

[Jimbo75]

While pie charts like that are pretty and fan the flames nicely in any ATI/nVidia war, they actually show very little.

The article: http://arstechnica.com/hardware/news...ia-drivers.ars



So:
* 5 year old data
* Higher market percentage = more crashes logged for that manufacturer
* No information on how the crashes were actually logged in regards to multiple (Same system) crashes
* No information on if the crash was due to hardware failure (or overclocking)
* No information on what a 'crash' actually is. A application bombing out? A BSOD / System lockup? A driver restart? and so on....

I'm the last person on earth when it comes to fanboy nonsense, I'll buy what's best price: performance at the time, but seriously, charts like that cause more trouble than they help solve.

Nevertheless they do show that Nvidia hardly has a sterling history of driver stability.

I mean I don't mind linking the other stuff they had much more recently, ie..

From 2 years ago - http://www.zdnet.com/blog/hardware/w...hics-card/7551

From last year - http://www.bit-tech.net/news/hardwar...-can-explode/1

Just sayin.

[Terbinator]

I'm more concerned that we're using Vista as a viable starting ground to begin with.

[kalniel]

I'm more concerned that we're using Vista as a viable starting ground to begin with.

Well it's far more stable than XP, if that was your concern?

[Jimbo75]

Nvidia toiled over Vista drivers for a long time before Microsoft was forced to prove it -

http://arstechnica.com/old/content/2007/02/8784.ars

I know it was a long time ago but it's still relevent in combination with the two more recent mini-disasters they've had with card killing drivers. Nvidia drivers are scary, I wouldn't touch new ones until at least 2 weeks after release.

[Agent]

Nevertheless they do show that Nvidia hardly has a sterling history of driver stability.

No, it doesn't for the reasons I've listed. You can't really concede the points about the data and methodology and then say 'they do show...' because they don't

For all we know, there could have been one machine with a hardware error that was causing a driver crash to get reported to Microsoft about the nVidia driver every few seconds of the day for the entire sample length. The same goes for people using BETA drivers which might be known to be unstable.

You just can't use that pie chart as evidence of driver stability, period. On the face of that chart it looks like nVidia crashed about 3 times as much as ATIs drivers. Now what about if there were only 1000 ATI machines, but 300,000 nVidia machines in the sample? You'd expect more crashes with all the different hardware configurations.

You need to know what the sample size of the manufacturer is as well as the crash percentage to draw anything meaningful from it (Heck, you need everything I've listed really to start drawing real comparisons!)

I mean I don't mind linking the other stuff they had much more recently, ie..

From 2 years ago - http://www.zdnet.com/blog/hardware/w...hics-card/7551

Yeah, that was horrid although thankfully they pulled them fast.

Drivers to one side though, the two big questions I have with that were:
*Why the heck wasn't there a thermal shut-off, or did 3rd parties remove this?
*Why were the drivers WHQL certified?

From last year - http://www.bit-tech.net/news/hardwar...-can-explode/1

Just sayin.

Woah, hold your horses with that one

Even if we ignore the sensationalist headline "explode - lol!", there is still a few important things about the issue:

Nvidia’s recently released GeForce GTX 590 3GB - a card that costs around £580 inc VAT - can overheat and even blow up if overclocked with an old driver.

The problem was captured on video by the guys at SweClockers, and is apparently a result of the card drawing more power when overclocked or overvolted than its power circuitry can handle.

However, the problem can only occur if you overclock or overvolt the card, and only if you use an older driver. The GeForce 267.52 driver that was originally given out to the press, and which might be available from some websites, does not activate a safety mechanism that prevents damage to the card’s power components. However, Nvidia's latest driver does feature overcurrent protection.

You need to take the card outside of it's rated limits for any issues to happen. This is hardly nVidias fault if the user does this. While it's nice to have driver protection, if you start overvolting something and overclocking it, don't moan when it goes pop.
Further, the driver was only given to the press for testing. Any user who brought it as normal wouldn't have experienced this issue as it was supplied with a latter one.

I fear this could go on forever though and I have no interest in defending nVidia any more than I have an interest defending ATI .
The point I do want to make though is that statistics is complex at the best of times. When you start talking about sample sizes in the millions, you must have the correct methodology and data to support any claims derived from it.

I know it was a long time ago...

The thing is, 5 years in the computing industry is ages. It's several generations of CPUs, GPUs, Games engines, iterations of programming languages and so on. To assume that the staff at nVidia are even the same people as 5 years ago is pushing the boat out on its own, never mind any links to the actual code being written.

Fool me once and all that I know, but if things never moved on no one would buy a Intel chip due to the crappy Pentium 4 netburst era or Hitachi/IBM drives due to the deathstar fiasco.

[CAT-THE-FIFTH]

That and NVidia drivers are so much better across the board.

Sadly not in my experience

You are doing it wrong

Here is the best image for this thread



Driverssmivers.

There is something not right going on here.

[Jimbo75]

Not saying I disagree with you Agent but there could have been more ATI machines causing more errors than normal too. That however is unlikely and with stuff like this you have to take it at face value instead of thinking about ways to excuse the issues.

I get your point on there being more Nvidia machines - undoubtedly there were more however not to those sort of figures. I'd use the intel numbers as a good reason why it's far more likely that the problem was on Nvidia's end - those 8.8% of intel issues probably all came from GMA or intel extreme (or whatever their integrated graphics were back then), and we know based on market share that there were far more of those than both AMD and Nvidia combined. If you then consider than Nvidia has had over 3 times the failures logged compared to intel it paints a fairly clear picture.

edit woops I linked from 2009 - here is a good one from 2007



http://arstechnica.com/hardware/news...-amd-intel.ars

So actually Nvidia had a lot of market share back then but still behind intel and not even twice that of AMD.

[Agent]

Aye, don't disagree with anything you've said there either mate - but even if we solved the sample size prob we still have the others

[Terbinator]

Well it's far more stable than XP, if that was your concern?

In 2007? Not from my experience, especially in terms of how it played with drivers.

SP1 Vista however was solid for me.

[Queelis]

*snip*

Edit: Where's that delete button?

[tribaljet]

In 2007? Not from my experience, especially in terms of how it played with drivers.

SP1 Vista however was solid for me.

^This. SP1 made wonders for Vista, and virtually all BSODs in Vista were driver related since there was such contrasts between driver architecture on Vista and on legacy OSes.

[brasco]

Back to the 680GTX, the latest rumours that it has a "Speed Boost" tech much like Turbo on Intel chips. So maybe that clock is the low end? Rumours mixed with rumours mind

cheers
brasc

[CAT-THE-FIFTH]

This is more or less the thread where most of us are posting any rumours about Kepler:

http://forums.hexus.net/graphics-car...-released.html

[flearider]

amd pretty much have the full range out now (7670 is a rebrand of the 6670)

Pity AMD has pricing strategy that matches that of monopoly. No wonder everyone is waiting for NVidia.

ahh but only a few yrs ago nvidia did the same ...rebranding lower gpu's with higher numbers ...but hay you make money while you can .. it's business