nav[aria-label="Primary Navigation"] { padding: 0; & ul { list-style: none; width: 100%; display: flex; flex-direction: row; justify-content: start; align-items: start; gap: 30px; padding: 0; & li { margin: 0; } & ul li { list-style: none; } } }

Child Item

Nvidia RTX Discussion

K.O.Kiki

A new era in 3D graphics stands before us.

First-gen Raytracing GPUs:

Nvidia "Turing" - Quadro RTX 5000 / 6000 / 8000 (US$ 2'300 / 6'300 / 10'000)

https://www.anandtech.com/show/13214/nvidia-reveals-next-gen-turing-gpu-architecture
https://www.anandtech.com/show/13217/nvidia-announces-turing-powered-quadro-rtx-family

Find more posts tagged with

Quick Links

Comments

Genevieve Disgusting Self-improvement wrote: »

Was the point of that video not that the RMA rate is very low on partner RTX cards?

Nah, the point was that there's nothing pointing to an unusually high failure rate in the RTX series and the media shouldn't be judging these things based purely on forum/reddit posts. It's a valuable point for him to make since he's one of fairly few people with access to real data.

Worthwhile information to add to this:

1080 - sample size of "a lot of" thousand units.
1080 Ti - sample size of "several" thousand units.
2080 - sample size of "very close to" a thousand units.
2080 Ti - sample size of "a bit less than" the 2080/a thousand.

So, newer release dates aside, there's a lot less 20 series in the wild to be basing the numbers on yet.

Unfortunately I don't think the information we have is precise or complete enough to actually do any kind of halfway useful statistical inference with it (I'd like exact numbers and RMAs through time for each card ideally but beggars can't be choosers). All we can really do is weigh their accuracy in order of how big his ballparks were.

Genevieve Disgusting Self-improvement wrote: »

5% of 1080ti's and 7% of 1080 is more alarming albeit larger N at risk for more time, as he specifically says a RMA rate below 3% is fine...

Well, the video says here that the 10 series failure rates are higher because they've been out a lot longer. He also says here that 5-7% is fairly typical for high end components, and 15-20% is when they figure something is wrong. So, I wouldn't call it alarming.

Genevieve Disgusting Self-improvement wrote: »

Dont let the information in the video get in the way of the narrative, ...RTX Bad.... :rolleyes:

I don't think that's what Terror was implying.

Look at the difference in failure rate between the 2080 and 2080 Ti. They're cards with similar release dates, processes and architecture - but a bit over a percent in difference in their RMA rates.

On top of that, compare the 1080 Ti and the 2080 Ti - the 2080 Ti has, in six weeks, accrued a failure rate of a little under a third of what the 1080 Ti accrued in two and a half years.

Either that's low numbers skewing it high, or suggestive of a problem. I'd tend towards low sample size as a first guess.

It'll be very interesting to see how it develops as we see more of them in the wild. Could go down, but chances are good it'll go up a fair bit but not up to "problem child" levels.

TerrorFirmer

That is precisely what I was pointing out. The 2080Ti RMA rate is significantly higher than the 2080, which is interesting.

Also, of course the 10 series in general has way higher rates at this point in time....the cards have been out for years. The discussion is around if there's a potential issue with brand new launch cards.

You really need to stop being so hyper sensitive about the RTX discussion.

btkm8unsl0w5r4

The fact is there is no problem and there is nothing indicating a problem, the interest people have in watching the "developments" is one of wanting, wishing, hoping and praying that RTX crashes and burns.

Debauers numbers are strong enough to draw a valid conclusion. RMA's happen mostly when people receive new items rather than failures over time. I am not sure what the TaR impact is for GPU or what the MTTF but I would suggest that there is a strong drop off in RMA over time owned.

The 1080ti was release 18 months ago BTW, and initial stock levels were very poor. Only really readily available for about 15 months. You would need the mean time before RMA to draw any conclusion. The fact that you stand a 5% chance of having to return a 1080ti within 18 months (or likely less) is deemed normal, but the chance of a defective 2080 after 2 months being 1.4% is something we need to monitor carefully to see where it goes...thats illogical.

Also what are the RMA for....coil whine? Faulty fans? Damage in shipping?...none of this data is broken down enough to draw any conclusion....and the reason for this is because there is no specific story here different from any other GPU launch. I am not saying that there is any fault with the beloved, last gen, discontinued, and ageing 1080ti, of which one in 20 need returning

, I only suggest that peoples interest here is in the correctional suffering or schadenfreude of the early adopters.

I am not sensitive, merely persistently not allowing this thread to devolve into another echo chamber where detractors focus on and embellish the negative while they ignore or minimise positive. We are all invested in our own little ways, I recognise mine.

Inquitus

Genevieve Disgusting Self-improvement wrote: »

The fact is there is no problem and there is nothing indicating a problem, the interest people have in watching the "developments" is one of wanting, wishing, hoping and praying that RTX crashes and burns.

Debauers numbers are strong enough to draw a valid conclusion. RMA's happen mostly when people receive new items rather than failures over time. I am not sure what the TaR impact is for GPU or what the MTTF but I would suggest that there is a strong drop off in RMA over time owned.

The 1080ti was release 18 months ago BTW, and initial stock levels were very poor. Only really readily available for about 15 months. You would need the mean time before RMA to draw any conclusion. The fact that you stand a 5% chance of having to return a 1080ti within 18 months (or likely less) is deemed normal, but the chance of a defective 2080 after 2 months being 1.4% is something we need to monitor carefully to see where it goes...thats illogical.

Also what are the RMA for....coil whine? Faulty fans? Damages in shipping?...none of this data is broken down enough to draw any conclusion....and the reason for this is because there is not specific story here different from any other GPU launch. I am not saying that there is any fault with the beloved 1080ti, I only suggest that peoples interest here is in the correctional suffering or schadenfreude of the early adopters.

I am not sensitive, merely persistently not allowing this thread to devolve into another echo chamber where detractors focus on and embellish the negative while they ignore or minimise positive. We are all invested in our own little ways, I recognise mine.

But there are few positives, yes if you want to throw well over a Grand at a marginal gain, you can, and it will be the fastest there is. As it stands that's the long and the short of it.

BloodBath

Yeah I don't think Fitz is being unreasonable at all. I see the same people just looking for reasons to hate the cards.

Totally clutching at straws looking for negative stuff from other negative people on youtube.

Around 5% failure rate for electronics is the norm in the first year. Hence why almost everything on Amazon has at least 5% negative 1 star reviews.

btkm8unsl0w5r4

if by marginal you mean 25% over 1080ti to 2080ti (and the gain depends totally on what you are coming from ), and so long as you ignore any benefits of the video encoding engine, the HDR handling, the memory speed, the compute power (its literally twice as fast as my 1080ti at some of my work compute CAD processing tasks, shaving 2 minutes off my standard STL model processing time) , and any-benefits in the future of AI cores and RT cores...then yes there are very few other positives. Cost is also a factor again with greater or lesser importance depending. Ask yourself, if cost was not a factor what would you get....now imaging a large swath of people for whom cost is not the overriding factor....I know, the capitalist pigs.

The extreme tech article here offers a very moderate and considered opinion on RTX that I find it hard to disagree with....especially as I sit here looking at my 8800 gt, a card admonished at the time but hailed now as a landmark launch and the start of a new architecture. Pascal for all the love was iterative, a good card that will be remembered for nothing except being a bit faster and a bit more expensive than the last card in your machine. I am astonished on this forum that there is so little enthusiasm for this.

Bloodbath...there is another thread where a guy is looking for a 9900k...a totally pointless processor for almost everyone, hard to overclock, hot as hell, delidding is very hard and the die needs sanding, motherboards need the VRM's from a starship and offers nothing except to the cinebench addict aka the ryzen owner

(proper productivity is done on Xeons...and if you want to look at cost to performance dont go there). Does anyone give him guff...not really, cause hey its a good processor if you have the sheckles and want the latest and greatest. More power to him.

https://www.extremetech.com/gaming/278454-nvidia-rtx-2080-and-rtx-2080-ti-review-you-cant-polish-a-turing

6034073

K.O.Kiki

Pretty sure we all told him not to bother lol

tuxy

K.O.Kiki wrote: »

Pretty sure we all told him not to bother lol

A link to the thread would be interesting as the only one I'm aware of people were advising someone not to buy the 9900k.

Cuddlesworth

K.O.Kiki wrote: »

Caseking AiB cards only:
SKU | GTX1080 | GTX1080Ti | RTX2080 | RTX2080Ti
RMA rate | 7.1% | 4.6% | 0.17% | 1.4%

I don't believe caseking sell many 2080/ti with the reference pcb, which by all accounts appears to be the problem. Also, its not exactly a statistically correct comparison. If you have 2 years of 1080 data vs 1 month of 2080 data, you have a much higher chance of a higher overall failure rate for the 1080. It has to be within the same timeframe.

btkm8unsl0w5r4

Cuddlesworth wrote: »

I don't believe caseking sell many 2080/ti with the reference pcb, which by all accounts appears to be the problem. Also, its not exactly a statistically correct comparison. If you have 2 years of 1080 data vs 1 month of 2080 data, you have a much higher chance of a higher overall failure rate for the 1080. It has to be within the same timeframe.

in the vid DeBauer made the point that most of the AIB cards he sells are all on the reference PCB design. As also mentioned the time at Risk (TaR) would be important but also the average time from purchase to RMA (MTTF mean time to failure). I would suggest most RMA;s are done soon after purchase negating the TaR...its a complex correlation that neither works to the benefit nor detriment of either side of the argument.

Had not read the 9900k thread in a few days, I see some of our regular pascal acolytes on this thread actually recommended an RTX card to the chap....strange times indeed, but the fabled 580 euro last gen, discontinued, and ageing 1080ti does get a mention, and only one in 20 need to be returned. :pac:...maybe the incredibly low return rate of the 2080 would push the warranty minded towards the RTX

tuxy

Genevieve Disgusting Self-improvement wrote: »

Had not read the 9900k thread in a few days, I see some of our regular pascal acolytes on this thread actually recommended an RTX card to the chap....strange times indeed, but the fabled 580 euro last gen, discontinued, and ageing 1080ti does get a mention

Well he did make it very obvious that he wanted to spend as much money as possible for small gains. Perhaps RTX was a better fix for that criteria.
Like I said to him, at least he might get a benchmark out of the RTX cards where the 9900k wouldn't give him anything interesting to do.

Metric Tensor

This thread has more salt than the dead sea!

Keep it coming lads!

Revoker88

Metric Tensor wrote: »

This thread has more salt than the dead sea!

Keep it coming lads!

Do you not know its a conspiracy against RTX?

All of us on the forum have decided to band together,form a narrative and trash RTX relentlessly.

btkm8unsl0w5r4

EoinHef wrote: »

Do you not know its a conspiracy against RTX?

All of us on the forum have decided to band together,form a narrative and trash RTX relentlessly.

Lol....thing is the narrative has been fed to you, you only think its your own, thats the way media works these days. You will find now that all the pascals are gone ( and the crappy 104 chips are binned to 1060's to complete with whatever mid tier thing amd bring out) the narrative will slowly change....

Genevieve Disgusting Self-improvement wrote: »

The fact is there is no problem and there is nothing indicating a problem, the interest people have in watching the "developments" is one of wanting, wishing, hoping and praying that RTX crashes and burns.

And here I thought my primary interest in continuing to keep an eye on RTX, despite it looking too pricey for what it is (and having a wasteful design), was the ticking clock of my EVGA step-up (T-18 days and counting).

Damn, thanks for knowing my mind better than me Fitz ... I'll adjust trim and start ****ting on the RTX cards randomly like my secret subconscious motivation clearly tells me to.

Genevieve Disgusting Self-improvement wrote: »

Debauers numbers are strong enough to draw a valid conclusion.

Der8auer's numbers *could* be strong enough to draw a valid conclusion - if we knew what they actually were. As is, all we know about his 2080 Ti numbers is it's more than zero and a good bit less than a thousand. That covers a multitude of sins.

Genevieve Disgusting Self-improvement wrote: »

RMA's happen mostly when people receive new items rather than failures over time. I am not sure what the TaR impact is for GPU or what the MTTF but I would suggest that there is a strong drop off in RMA over time owned.

While that's true (though cards do fail over time as well), there's not that many 2080 Tis out in the wild to be failing yet. At a 1.4% failure rate, you're looking at about 700 sold 2080 Ti's needed to reach double digit failures (which I'm not certain der8auer's ballpark sales numbers are reaching). When we're down at the level where the RMA of a single card shifts the failure rate, it's hard to have much confidence in it.

On that note as well, that 2080 figure seems really odd - since we know RMAs can't be fractional, that would put the only valid sales and RMA number in as a ballpark of 1 RMA and 600 sales (which doesn't fit his description as being close to a thousand). I'm wondering did he drop a power of 10 (which would put us back in the range of the RTX 2080 Ti @1.7% and give him ~950 sales/16 RMAs).

Genevieve Disgusting Self-improvement wrote: »

As also mentioned the time at Risk (TaR) would be important but also the average time from purchase to RMA (MTTF mean time to failure).

(Pulled this in for context from a different post) TaR's also the acronym for threshold autoregression btw. You'd me seriously confused there for a minute (it'd also make sense in context but it'd be a bit of a strange choice).

Genevieve Disgusting Self-improvement wrote: »

The 1080ti was release 18 months ago BTW, and initial stock levels were very poor. Only really readily available for about 15 months. You would need the mean time before RMA to draw any conclusion. The fact that you stand a 5% chance of having to return a 1080ti within 18 months (or likely less) is deemed normal, but the chance of a defective 2080 after 2 months being 1.4% is something we need to monitor carefully to see where it goes...thats illogical.

Bah, that's what I get for changing horse midstream - I'd started with the 1080 before realising the 1080 Ti was a better comparison. Point still stands well enough at 20 months (with lets say 17 for your point there). I'd note the 2080 Ti has limited availability as well at the moment.

A lot can change between 6 weeks and 20 months - especially when we're discussing a chip where a good third of it is sitting completely unused at the moment. The interesting bit for me is the big difference between the 2080 Ti and the (extremely low) 2080 RMAs. I'd expect them to look a lot more similar (a little higher maybe), and that makes me think

Genevieve Disgusting Self-improvement wrote: »

Also what are the RMA for....coil whine? Faulty fans? Damage in shipping?...none of this data is broken down enough to draw any conclusion....and the reason for this is because there is no specific story here different from any other GPU launch.

I'm interested to see how it pans out - you're right that we don't have an adequate breakdown on the RMAs. Those figures do look kind of funky though - I'd love it if we had some equivalent numbers for the 10 series RMAs at 6 weeks out to compare them against. As is, we've really just got gut to go on.

Genevieve Disgusting Self-improvement wrote: »

I am not saying that there is any fault with the beloved, last gen, discontinued, and ageing 1080ti, of which one in 20 need returning , I only suggest that peoples interest here is in the correctional suffering or schadenfreude of the early adopters.

I am not sensitive, merely persistently not allowing this thread to devolve into another echo chamber where detractors focus on and embellish the negative while they ignore or minimise positive. We are all invested in our own little ways, I recognise mine.

I think most of us would love to see the 2080 Ti do something amazing Fitz. People are disappointed, not looking to enjoy schadenfreude at your expense.

The issue is at the moment the Emperor needs to pull on some pants - it's overpriced, oversized, and underperforming (compared to expectations for a 12 nm card). It's a strong card, but only because there's nothing else new at the top end.

Genevieve Disgusting Self-improvement wrote: »

Had not read the 9900k thread in a few days, I see some of our regular pascal acolytes on this thread actually recommended an RTX card to the chap....strange times indeed,

It's almost like we're not some moustache twirling hivemind out to tie RTX to the railtracks and cackle maniacally Fitz, but a board of enthusiasts trying to give the best advice for assorted situations ...

Genevieve Disgusting Self-improvement wrote: »

maybe the incredibly low return rate of the 2080 would push the warranty minded towards the RTX

It pushes me towards wondering about his maths to tell you the truth (see above).

So, something a little different (and a little more practical). As I mentioned above, I've 18 days left on my EVGA step up so I need to start making decisions on whether to go for an RTX or stay with my 1080 Ti.

At the moment, these are my options and figures:

Card| |Price|cost to step up
EVGA GeForce GTX 1080 Ti FTW3 GAMING, 11G-P4-6696-KR| |€635| -
EVGA GeForce RTX 2080 XC BLACK EDITION GAMING, 08G-P4-2082-KR| |€874|€239
EVGA GeForce RTX 2080 Ti XC BLACK EDITION GAMING, 11G-P4-2282-KR| |€1,271|€635

There'll be shipping costs as well, so maybe another €15-30 on top of the cost of these.

The 2080 is low binned, and the 2080 Ti seems a big chunk of change for a small enough upgrade.

I'm heavily leaning towards not stepping up, but I'm open to being convinced. What do folk reckon?

Xenoronin

L wrote: »

I'm heavily leaning towards not stepping up, but I'm open to being convinced. What do folk reckon?

Feels like a pretty hefty cost tbh. I don't think you will get >€200 worth of value out of the stepup. Not now anyway... In a few years maybe, but by then, depending on how big an enthusiast you are, the itch might be back to upgrade anyway.

btkm8unsl0w5r4

L wrote: »

So, something a little different (and a little more practical). A..........?

It makes no sense to step up unless.

1. You want RTX features
2. You want the latest and greatest.
3 You need to game at 60fps 4k in all games.
4. You know in your heart that you will end up upgrading in the near future and now is the time to maximize value in your 1080ti.

Its a discretionary purchase and you have to want to do it

Personally I would if you have the oppertunity, its not that much more in the grand scheme and you already have such an expensive card. If you want to upgrade in a while you will suffer more loss as the 1080ti will be second hand value only. The RTX 2080ti is the only one that makes sense, cause you will always feel you have a lower end card with the 2080 despite its flagship price. If your perfectly happy with your 1080ti then stick.

Metric Tensor

Genevieve Disgusting Self-improvement wrote: »

its literally twice as fast as my 1080ti at some of my work compute CAD processing tasks, shaving 2 minutes off my standard STL model processing time

I'd love to hear more about this Genevieve Disgusting Self-improvement.

I do low level 2D CAD myself and duck and dodge to avoid having to interact with BIM models but even those wouldn't have much in the way of optimised rendering and or textures. I have recently started using 3d textured drone models though for topographical surveys.

What type of stuff are you doing and what software is making best use of the 2080ti?

btkm8unsl0w5r4

I use a specialized CAD package that triangulates 2d pictures into 3d models, it is heavily compute and cuda accelerated, and usually a 4000 image set takes 5 minutes to process. With RTX cards its more like 3 minutes, all CPU cores are pinned too, it really kicks the crap out of any PC. Seems the memory speed is a great help. You mileage may vary. I use another package meshmixer a simple modeling tool and it sees no benefit.

Metric Tensor

Triangulating 2D pictures onto 3d models is what happens for the drone models but most suppliers seem to operate on a cloud based system where you upload the raw data and 24/48 hours later they send you back a point cloud, model, texture file, etc.

The one desktop based one I know is Pix4D and now that I google it - they do seem to have their software optimised for nVidia cards. Might look into it more!

Thanks.

btkm8unsl0w5r4

Metric Tensor wrote: »

Triangulating 2D pictures onto 3d models is what happens for the drone models but most suppliers seem to operate on a cloud based system where you upload the raw data and 24/48 hours later they send you back a point cloud, model, texture file, etc.

The one desktop based one I know is Pix4D and now that I google it - they do seem to have their software optimised for nVidia cards. Might look into it more!

Thanks.

This model of cloud is a good one, and the AI algo's that Nvidia have developed would seem a good fit, problem for me is that its a time critical task. (Its for dental prosthesis manufacture from Intraoral scans and radiology.). I need the scans processed in order to move on with the next step of the process.

If I can save 30 minutes a day, that pays for the RTX card in a week. TBH I am wondering if it scales with NVlink as usually cuda work is SLI agnostic however memory bandwidth is an issue, however NVlink does share memory rather than pool it. Really I should look into professional cards but I like to game and benchmark in my spare time.

For me its amazing to see consumer cards have this sort of potential, and a wide install base will push development of solutions that people are not even thinking of. Its like pure research, very expensive, very little practical application at the start, but technological innovation for its own sake is always worthwhile. Its a real chicken and egg situation, problem is a lot of people seem to only want to eat eggs and not raise the chickens.

Genevieve Disgusting Self-improvement wrote: »

This model of cloud is a good one, and the AI algo's that Nvidia have developed would seem a good fit, problem for me is that its a time critical task. (Its for dental prosthesis manufacture from Intraoral scans and radiology.). I need the scans processed in order to move on with the next step of the process.

If I can save 30 minutes a day, that pays for the RTX card in a week. TBH I am wondering if it scales with NVlink as usually cuda work is SLI agnostic however memory bandwidth is an issue, however NVlink does share memory rather than pool it. Really I should look into professional cards but I like to game and benchmark in my spare time.

That's pretty interesting - that sounds like it's scaling pretty much straight with core count and memory bandwidth. Makes me wish FLOPS were more stringently defined as looking at that alone I wouldn't have guessed you'd get that performance boost from switching the two cards.

I'd be hesitant to recommend professional cards. I know traditionally most researchers I know have preferred using high end consumer cards because the cost differential is so huge. I also remember Nvidia having to ban GTX cards from being used in datacenters as they were getting used en masse instead of Quadros due to price/performance.

Weird as it is to say, a second 2080 Ti might actually be the cost effective way to get that extra performance. That or quad 1080 Tis (you shouldn't have to worry about quad SLI being 'officially unsupported' if it's for cuda use, and you might well make up the lost memory bandwidth with the jump in total core count). It depends a bit on whether or not it'd be a vatless business expense or the like either.

Genevieve Disgusting Self-improvement wrote: »

For me its amazing to see consumer cards have this sort of potential, and a wide install base will push development of solutions that people are not even thinking of. Its like pure research, very expensive, very little practical application at the start, but technological innovation for its own sake is always worthwhile. Its a real chicken and egg situation, problem is a lot of people seem to only want to eat eggs and not raise the chickens.

Consumer technology isn't pure research though. It needs to have utility and be delivered at a compelling price in order to gain mass adoption.

Diffusion's quite a heavily studied area, and there's been quite a bit of modelling/analysis done on how innovations spread through society. It's strongly tied to risk tolerance - about 10% are open to early adoption of technology, a good 70% or so will adopt when they see the tech as functional/low risk, and the rest will lag way behind. You're kind of our canary in the coal mine.

Besides, don't make me buy a chicken when I just want to make an omelette. :P

btkm8unsl0w5r4

L wrote: »

Consumer technology isn't pure research though. It needs to have utility and be delivered at a compelling price in order to gain mass adoption.

Besides, don't make me buy a chicken when I just want to make an omelette. :P

First gen comsumer stuff is always a diminishing return, and it does have utility...RTX does everything, and it does it pretty quickly, albeit at a price.

I have always found that the single card solution is generally the best from a compatability POV espically when using software from small developers that might not be so well optomised. I ran SLI for gaming for years. A tripple SLI 680 setup which looked cool but ran like muck...god the stutter. Dual 1080 came after than...looked cool but so few games scaled well for the power it was drawing, and SLI runs hot if your not running custom water cooling. I decided a while back to leave that alone and concentrate on single GPU solutions.

You dont need to buy a chicken, just wait for the omelette. Just dont give out that chickens are expensive, smelly, dirty, hungry, some die, dont do anything except potentially lay eggs that you have yet to see and that the eggs have no more calories than the potatoes you already have been eating for years....,this analogy is getting scrambled at this stage, thats no eggaggeration, its beyond a yolk. An Oeuf is an Oeuf.

smacl

Metric Tensor wrote: »

The one desktop based one I know is Pix4D and now that I google it - they do seem to have their software optimised for nVidia cards. Might look into it more!

There are a few desktop based photogrammetry packages out there for drones including Pix4D, AgiSoft, ContextCapture and RealityCapture. One of my UK clients does a lot of this work and for the very high resolution stuff the processing can take days from pressing the start button even on a top end workstation, some more discussion here. CUDA tends to be used by specialist packages because there aren't many portable options. OpenCL isn't well supported on nVidia for obvious enough reasons. I've started doing more general purpose GPU programming and use DirectCompute/HLSL which is Microsoft centric but GPU vendor agnostic. Rayracing in hardware isn't of any real benefit here as it is too specific a task, you want as many compute cores and as much on board memory as your money will buy.

btkm8unsl0w5r4

Wendel is late to the party but does address some of the point on the thread.....

https://www.youtube.com/watch?v=-2vDhwGHWB4

Genevieve Disgusting Self-improvement wrote: »

I have always found that the single card solution is generally the best from a compatability POV espically when using software from small developers that might not be so well optomised. I ran SLI for gaming for years. A tripple SLI 680 setup which looked cool but ran like muck...god the stutter. Dual 1080 came after than...looked cool but so few games scaled well for the power it was drawing, and SLI runs hot if your not running custom water cooling. I decided a while back to leave that alone and concentrate on single GPU solutions.

For gaming I definitely agree with you - SLI is a bit of a white elephant. For pure computation though it might be worth a look (supposedly it's actually quite good with a proper cooling solution). You'd know better than I do how well it'd work with your CAD package though - and I guess it'd be sensible if you are trying it to try with a less costly set of cards initially to see how it scales (if you've any old pairs lying around for example for a proof of concept).

Genevieve Disgusting Self-improvement wrote: »

You dont need to buy a chicken, just wait for the omelette. Just dont give out that chickens are expensive, smelly, dirty, hungry, some die, dont do anything except potentially lay eggs that you have yet to see and that the eggs have no more calories than the potatoes you already have been eating for years....,this analogy is getting scrambled at this stage, thats no eggaggeration, its beyond a yolk. An Oeuf is an Oeuf.

That was plain oeuf-ul.

I can still wish for my nice clean pleasingly engineered chicken though. :P

Revoker88

Genevieve Disgusting Self-improvement wrote: »

Lol....thing is the narrative has been fed to you, you only think its your own, thats the way media works these days. You will find now that all the pascals are gone ( and the crappy 104 chips are binned to 1060's to complete with whatever mid tier thing amd bring out) the narrative will slowly change....

Ha,so im just thinking what im told to think?

Gimme a break,your patronising arrogance is getting boring now.

btkm8unsl0w5r4

EoinHef wrote: »

Ha,so im just thinking what im told to think?

Gimme a break,your patronising arrogance is getting boring now.

Have we not been warned not to fight....come here and give me a hug ya big lug........I am beginning to think you dont like me

So surprise surprise on examination of the failing cards a lot of the BSOD are due to monitor issues, and is likely a driver fix, there are some that are hardware issues but thats less common. Steve on gamer nexus is doing a series on them, https://www.youtube.com/watch?v=A4g5CZCaWXo&t=393s, I am sure those watching the situation develop will be interested in this.

I saw that this morning - seems like the drivers are a real mess. You'd think if you were guaranteed a good experience, it'd be with Nvidia's own licensed tech - G-sync related crashes are just plain embarrassing.

Pretty impressive he got people to send in their failed cards to be tested - I'm hoping he'll do up some numbers on the various types of failures he's seeing. Could be very interesting.