Advertisement
Help Keep Boards Alive. Support us by going ad free today. See here: https://subscriptions.boards.ie/.
If we do not hit our goal we will be forced to close the site.
Current status: https://keepboardsalive.com/
Annual subs are best for most impact. If you are still undecided on going Ad Free - you can also donate using the Paypal Donate option. All contribution helps. Thank you.
If we do not hit our goal we will be forced to close the site.
Current status: https://keepboardsalive.com/
Annual subs are best for most impact. If you are still undecided on going Ad Free - you can also donate using the Paypal Donate option. All contribution helps. Thank you.
https://www.boards.ie/group/1878-subscribers-forum
Private Group for paid up members of Boards.ie. Join the club.
Private Group for paid up members of Boards.ie. Join the club.
Nvidia RTX Discussion
Comments
-
-
-
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.0 -
-
-
Advertisement
-
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-turing0 -
-
-
-
-
Advertisement
-
-
-
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.0 -
-
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.
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.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).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).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 thinkAlso 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.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.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 ...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). 0 -
-
-
-
-
-
Advertisement
-
-
-
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.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. :P0 -
-
-
Wendel is late to the party but does address some of the point on the thread.....
0 -
-
-
-
Advertisement
-
https://www.youtube.com/watch?v=-2vDhwGHWB4Advertisement