|
GrizzlyCow posted:Upside: Maybe the prices for lower end AMD cards will finally go down? Probably not. They'll still be attractive to coiners who can't grab Maxwell cards. Maybe the cryptocurrency bubble will burst a couple weeks sooner. Probably not.
|
#
?
Feb 21, 2014 06:06
|
|
|
|
| # ? Apr 23, 2024 16:29 |
|
|
Nephilm posted:Price-wise to what? This is efficiency, and the initial investment doesn't matter because mining bitcoins is pure profit, so better performance per watt is more pure profit. As someone who mined 60 bucks this week for shits and giggles, it's not "pure profit" unless you're stealing someone's electricity and place no value on your own time. I can make more money doing just about anything else than dicking around with these things. And of course the initial investment matters. If you bought an R9 290X at $1000 bucks like it hit over the weekend and mine $50 a month from it you're probably never making any money back at all between electricity bills, depreciation of the card, and the arms-race of mining equipment and difficulty increases. Sir Unimaginative posted:Probably not. They'll still be attractive to coiners who can't grab Maxwell cards. A lot of "serious business" miners are using that BAMT linux live-cd to mine with, and it doesn't have CUDA mining integrated yet anyway.
|
|
#
?
Feb 21, 2014 06:11
|
|
|
Zero VGS posted:As someone who mined 60 bucks this week for shits and giggles, it's not "pure profit" unless you're stealing someone's electricity and place no value on your own time. I can make more money doing just about anything else than dicking around with these things. That's the joke.
|
|
#
?
Feb 21, 2014 07:36
|
|
|
Nephilm posted:That's the joke. Can't be too careful! Anyhoo, I just went on NewEgg and nabbed one of the EVGA 750ti FTW to try out: http://www.newegg.com/Product/Product.aspx?Item=N82E16814487028 I've never overclocked before but this looks like a fun card to start out with. $160 after rebate for a base clock of almost 1200mhz and reviews are saying it will readily top 1400. Edit: Oh, my previous GPU setup was two 770's in SLI to run two instances of Realm Reborn, one on each 1080p monitor for a friend and I. Is there an easy way in Windows to set things up so one instance of a game would run on one 750ti and the other instance of the game would run on the other 750ti card on another monitor? I know they can't do SLI but that would be the next best thing since it's the one time my system is going full blast, and when I'm not using it for two copies of a game I could set it to PhysX so it's not totally useless. Zero VGS fucked around with this message at 08:21 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 08:07
|
|
|
So after having 3 290x's be returned for different problems I've decided to switch to team green and get a 780ti, how ever I'm completely out of my depth when it comes to what model to get, I'm looking at EVGA as they seem to be the go to guys for the green team, how ever as I'm in the magical land of Australia  so not all of the models are available here, So my question is thus, Between this modelhttp://www.pccasegear.com/index.php?main_page=product_info&cPath=193_1577&products_id=25658 -780ti Superclocked and this model http://www.scorptec.com.au/product/Graphics_Cards/NVIDIA/53380-03G-P4-2888-KR -780ti Classified how much of a "guaranty" is there that the classified will be more overclockable then the Superclocked? as I have a 480mm & 360mm radiator system just waiting for a card to cool, the cooler that the card comes with isn't a factor as it will come off so is the extra $100 worth it at all?
|
|
#
?
Feb 21, 2014 09:41
|
|
|
Purchase questions go in the shopping thread. As do attempts to justify video cards that cost as much as most people's computers, even in The Loneliest Bit of the Anglosphere money. (Please tell them that at least there's 1440p involved.) Since you're here now: EVGA bins tight; you'll only get any reliable headroom out of the highest-clocked version of any given GPU, v v v and apparently your retailers of choice don't stock it, and googling "780 Ti Kingpin site:.au" suggests that not even eBay is an option for you. Try SA-Mart? 
dont be mean to me fucked around with this message at 10:18 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 10:10
|
|
|
EVGA bins their GPUs really extensively. The Classified line up is their second highest bin. The KingPin edition is their highest binned GPUs.
|
|
#
?
Feb 21, 2014 10:11
|
|
|
The waterblock they are using on the hydocopper version is garbage as well. No VRM cooling at all. If that was even a remote consideration.
|
|
#
?
Feb 21, 2014 10:17
|
|
|
Sir Unimaginative posted:Purchase questions go in the shopping thread. As do attempts to justify video cards that cost as much as most people's computers, even in The Loneliest Bit of the Anglosphere money. (Please tell them that at least there's 1440p involved.) Well I figured this thread would be better suited to my question as its not so much a question of what GPU to get, more so the question of how EVGA bins their chips and to how much of a degree the different models differ in the chance of a awesome overclock (BTW Yes i have a 1440p monitor) SlayVus posted:EVGA bins their GPUs really extensively. The Classified line up is their second highest bin. The KingPin edition is their highest binned GPUs. but how much of a difference is there between them? if it goes SC => Classified => Kingping how much of a gap is there? I know the Kingpins are the bet of the best of the best, but whats the difference between the SC and the classified? I know from other sites that the Kingpin edition is really only for people doing LN2 bench runs or WC'ed setup's and there's a shortage of chips that fit the binning requirements for them (never mind the 0% chance of ever seeing stock in Australia)
|
|
#
?
Feb 21, 2014 10:40
|
|
|
40 MHz per step, give or take, and they test on the boost level. I don't know if EVGA's methods are as reliable as nVidia's or Intel's, or if they drop some batches down a bin because market segmentation, but we can safely say that your methods aren't as reliable or thorough as those of any of the corporations named above. Also GPGPU means that the line for a stable-enough overclock isn't a simple 'rendered image doesn't spew artifacts or anomalies and card, game and OS don't poo poo the bed' anymore. ... I suppose I'll be nostalgic for the days and ways of overclocking for the foreseeable future (disclaimer: posted from a 2500K @ 4.2 GHz), but I understand why it's going away (design efficiency and sheer thermo- and electro-dynamics) and given what even stock can do now it's pretty easy to come to terms with it. EDIT: If you still want to give overclocking a shot, Agreed wrote a pretty decent guide to it back in July (yonks ago in GPU years but directly applicable to the 780 Ti). Just, you know, don't expect much. dont be mean to me fucked around with this message at 12:47 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 11:55
|
|
|
Michymech posted:So after having 3 290x's be returned for different problems I've decided to switch to team green and get a 780ti, how ever I'm completely out of my depth when it comes to what model to get, I'm looking at EVGA as they seem to be the go to guys for the green team, how ever as I'm in the magical land of Australia Whilst we typically recommend against Gigabyte products, when you consider the fact that: 1. Gigabyte typically puts a fair amount of effort into top end parts 2. Nvidia's greenlight program stops vendors from cutting corners 3. Australian Consumer Protection Law renders warranties and manufacturer support irrelevant Gigabyte's version of the 780ti becomes more compelling. It's really well reviewed and they have the highest factory clocked 780ti on the market. The windforce cooler really is amazing, and the shroud and backplate are solid metal. http://www.pccasegear.com/index.php?main_page=product_info&cPath=193_1577&products_id=26525 It's what I plan on getting.
|
|
#
?
Feb 21, 2014 13:04
|
|
|
Sir Unimaginative posted:EDIT: If you still want to give overclocking a shot, Agreed wrote a pretty decent guide to it back in July (yonks ago in GPU years but directly applicable to the 780 Ti). Just, you know, don't expect much. I think it's funny that Factory Factory and I are both kind of in a weird position where we're totally ready for overclocking to go away as engineered efficiency continues to just fully wow everyone, but at the same time we're sort of ... the dudes who talk about that stuff, so, welp. Guys, you really shouldn't overclock. Validation is next to impossible, gains are more synthetic than real in many cases anyway, god knows the toll it takes on the hardware and oh here's my handy e-z guide for it 
|
|
#
?
Feb 21, 2014 14:08
|
|
|
Agreed posted:I think it's funny that Factory Factory and I are both kind of in a weird position where we're totally ready for overclocking to go away as engineered efficiency continues to just fully wow everyone, but at the same time we're sort of ... the dudes who talk about that stuff, so, welp. Guys, you really shouldn't overclock. Validation is next to impossible, gains are more synthetic than real in many cases anyway, god knows the toll it takes on the hardware and oh here's my handy e-z guide for it ZOMG IT'S SO FUN THOUGH!!!11!!1 But seriously, I recognise overclocking will probably be dead in 5 years, but I'm sure as you yourself can attest Agreed, it's so exhilarating to get "free" gains and to overclock. As someone who has only got into the computer game in the last two years, it saddens me to think of a future where overclocking will likely not exist, or, if it does, it will be so irrelevant it won't even matter. I think that the writing has been on the wall the last couple of years with respects to overclocking (and that applies to both CPU's and GPU's) but whatcanyado y'know? Just gotta enjoy it while you can and bitch about how your mate got 26MHz extra on his memory over clock.
|
|
#
?
Feb 21, 2014 15:42
|
|
|
Overclocking is fun, there's no doubt about it. People didn't fit peltier-effect coolers to their CPUs, VapoChill, or assemble their own cascade cooler for a few FPS. They did it because, dammit, it was a fun hobby.
|
|
#
?
Feb 21, 2014 16:09
|
|
|
HalloKitty posted:Overclocking is fun, there's no doubt about it. Yeah, it's just fun. Really reminds me getting more boost out of my cars except cheaper! and cleaner! really sad about that mining chart. I'm just going to buy an overpriced 660 ti, sli and call it a long day
|
|
#
?
Feb 21, 2014 16:30
|
|
|
Michymech posted:So after having 3 290x's be returned for different problems I've decided to switch to team green and get a 780ti, how ever I'm completely out of my depth when it comes to what model to get, I'm looking at EVGA as they seem to be the go to guys for the green team, how ever as I'm in the magical land of Australia Go for the Classified. It has better power circuitry than the SC and there are waterblocks from EK for it. The kingpin doesn't have any waterblocks made for it.
|
|
#
?
Feb 21, 2014 16:57
|
|
|
Until every single chip that is made, works at the very best potential speed, overclocking will exist. Overclocking gets so out of focus with the internet and instant gratification of community benchmarks. It isn't that exciting, because the only people that have to constantly make changes to their machines, don't know what they are doing or only care about pushing it as far as you can for some benchmark. Set it up and test over the first few weeks, then let it go. A good overclock lasts for years. Don't mistake careful marketing and binned processes for actual processor capabilities, they are all unique.
|
|
#
?
Feb 21, 2014 17:47
|
|
|
Phuzun posted:Until every single chip that is made, works at the very best potential speed, overclocking will exist. Overclocking gets so out of focus with the internet and instant gratification of community benchmarks. It isn't that exciting, because the only people that have to constantly make changes to their machines, don't know what they are doing or only care about pushing it as far as you can for some benchmark. Set it up and test over the first few weeks, then let it go. A good overclock lasts for years. Don't mistake careful marketing and binned processes for actual processor capabilities, they are all unique. I think it's exciting. I'm delidding my i5-4670k next week for no reason other than to overclock it more. I seriously doubt I'll see any real gains if I can push it to say 4.6 4.7, but I might, and that alone is worth it. Although I don't do benchmarks, in fact I never have. The only numbers I care about is how fast is it going and how hot is it getting. (shes not a dyno queen, its all about sotp)
|
|
#
?
Feb 21, 2014 17:58
|
|
|
veedubfreak posted:Go for the Classified. It has better power circuitry than the SC and there are waterblocks from EK for it. The kingpin doesn't have any waterblocks made for it. The classified EK block fits the kingpin card. If you're just water cooling though the kingpin is a waste.
|
|
#
?
Feb 21, 2014 19:00
|
|
|
I'm going to start a venture capital fund to buy up all the Maxwell cards on the market then resell them for a 50% markup. Who's in? 
|
|
#
?
Feb 21, 2014 19:01
|
|
|
Well, it's an easier way to make money from altcoins than actually mining the things.
|
|
#
?
Feb 21, 2014 19:06
|
|
|
Nostrum posted:The classified EK block fits the kingpin card. If you're just water cooling though the kingpin is a waste. Ah, it looked like the kingpin card was a different board from the Classy.
|
|
#
?
Feb 21, 2014 19:09
|
|
|
Phuzun posted:Until every single chip that is made, works at the very best potential speed, overclocking will exist. Overclocking gets so out of focus with the internet and instant gratification of community benchmarks. It isn't that exciting, because the only people that have to constantly make changes to their machines, don't know what they are doing or only care about pushing it as far as you can for some benchmark. Set it up and test over the first few weeks, then let it go. A good overclock lasts for years. Don't mistake careful marketing and binned processes for actual processor capabilities, they are all unique. The discussion here is an effort/benefit balance. As the efficiency of "auto overclock" on CPUs/GPUs improves and thermal constraints get narrower, for the end user the cost entry bar gets higher while the potential gains lower. Pretty soon it's just going to be "get good cooling and let Boost do its thing" (and the definition of good cooling is also changing as parts need less heat dissipation), with little sense in getting costlier unlocked components or BIOS tweaking except for, as you mentioned, push-it-to-the-limit enthusiasts.
|
|
#
?
Feb 21, 2014 19:21
|
|
|
Nephilm posted:The discussion here is an effort/benefit balance. As the efficiency of "auto overclock" on CPUs/GPUs improves and thermal constraints get narrower, for the end user the cost entry bar gets higher while the potential gains lower. Pretty soon it's just going to be "get good cooling and let Boost do its thing" (and the definition of good cooling is also changing as parts need less heat dissipation), with little sense in getting costlier unlocked components or BIOS tweaking except for, as you mentioned, push-it-to-the-limit enthusiasts. No, this is just the mainstream being confused about power saving circuitry. Running something below a known max speed, then dynamically increasing to that under load, is not overclocking.
|
|
#
?
Feb 21, 2014 19:47
|
|
|
Phuzun posted:No, this is just the mainstream being confused about power saving circuitry. Running something below a known max speed, then dynamically increasing to that under load, is not overclocking. ...correct me if I'm completely wrong, but for example Nvidia's GPU Boost, it isn't a "throttling down from a known max speed", but boosting clocks up until hitting temperature or TDP constraints - the former being a variable factor based upon cooling solutions implemented (which is what I mentioned), and the latter a manufacturer-specified limit that, as has been brought up by other people, has an ever narrowing fringe between the spec and how much a user can push beyond it without causing actual hardware damage.
|
|
#
?
Feb 21, 2014 20:03
|
|
|
Phuzun posted:No, this is just the mainstream being confused about power saving circuitry. Running something below a known max speed, then dynamically increasing to that under load, is not overclocking. I feel like you're casually misrepresenting or hand-waving a pretty important sea change that has occurred in how clock rate, power, and temperature is all handled. Efficiency is what every market segment is focusing on at this point except for the lunatic fringe, and even then, you have to fight the hardware to get it to discard the focus on efficiency. Cards, CPUs, motherboard components and controlled cooling - efficiency has become the ultimate goal. You're right that people will still overclock even after it doesn't make much sense, from earlier, but this point here is off the mark in my opinion.
|
|
#
?
Feb 21, 2014 20:12
|
|
|
Nephilm posted:...correct me if I'm completely wrong, but for example Nvidia's GPU Boost, it isn't a "throttling down from a known max speed", but boosting clocks up until hitting temperature or TDP constraints - the former being a variable factor based upon cooling solutions implemented (which is what I mentioned), and the latter a manufacturer-specified limit that, as has been brought up by other people, has an ever narrowing fringe between the spec and how much a user can push beyond it without causing actual hardware damage. Boost clocks are set by the manufacturer and are binned specs, not individually tested and set. Overclocking is pushing it further than the manufacturers specifications and each chip is unique in capabilities. Not trying to handwave anything. The technology in running these cards efficiently is great. But overclocking has a definition and it isn't the same as boost. Phuzun fucked around with this message at 20:20 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 20:15
|
|
|
Just keep moving the slider to the right until the computer crashes, then go a little less to the right the next time. You'll be ok.
|
|
#
?
Feb 21, 2014 20:24
|
|
|
Phuzun posted:Boost clocks are set by the manufacturer and are binned specs, not individually tested and set. Overclocking is pushing it further than the manufacturers specifications and each chip is unique in capabilities. Did I forget to put the quotes on my post? Nephilm posted:...As the efficiency of "auto overclock" on CPUs/GPUs improves and thermal constraints get narrower... Oh, no I didn't, it's all good. Every single point brought up by everyone but you has been correct. Stop being pointlessly anal about semantics.
|
|
#
?
Feb 21, 2014 20:43
|
|
|
Does anyone know what specifically are the actual limitations of a chip are for clock speed besides heat and voltage? Assuming heat and available power aren't a problem, what's the next wall that you hit? I imagine it's a fundamental quality of the chip itself but I have no idea what it is.
|
|
#
?
Feb 21, 2014 20:56
|
|
|
Ignoarints posted:Does anyone know what specifically are the actual limitations of a chip are for clock speed besides heat and voltage? Assuming heat and available power aren't a problem, what's the next wall that you hit? I imagine it's a fundamental quality of the chip itself but I have no idea what it is. I don't think there is one. If you had a GPU with no voltage limit, and you could magically keep the card cooled to Absolute Zero, you can jam as many electrons through it as you want. But in real life, heating up chips makes them less efficient, making them push harder and heat further in a vicious cycle until the poo poo melts itself down.
|
|
#
?
Feb 21, 2014 21:01
|
|
|
Zero VGS posted:I don't think there is one. If you had a GPU with no voltage limit, and you could magically keep the card cooled to Absolute Zero, you can jam as many electrons through it as you want. So it's really just a heat limit then. I figured at some point the ... uh, pipes would clog. For the extreme lack of a better phrase.
|
|
#
?
Feb 21, 2014 21:03
|
|
|
Some quick Googling suggests that electrons generally move at 90% of the speed of light, so nothing's getting clogged in that.
|
|
#
?
Feb 21, 2014 21:08
|
|
|
Ignoarints posted:Does anyone know what specifically are the actual limitations of a chip are for clock speed besides heat and voltage? Assuming heat and available power aren't a problem, what's the next wall that you hit? I imagine it's a fundamental quality of the chip itself but I have no idea what it is. Well, those are the two. Particularly at the scale we're looking at with modern microtransistors, chips just start doing wonky poo poo when they get too hot or run too much current - you affect the conductivity of the materials and gates start flipping when they shouldn't. If you mean the hypothetical "what if we had room-temperature superconductors to work with", then that's some rather esoteric poo poo right there, but to answer the question then there wouldn't be overclocking per se, instead our limit would become the operational speed of the minimum logic components and how fast they can talk to each other. It'd all be design and software efficiency.
|
|
#
?
Feb 21, 2014 21:08
|
|
|
Nephilm posted:Well, those are the two. Particularly at the scale we're looking at with modern microtransistors, chips just start doing wonky poo poo when they get too hot or run too much current - you affect the conductivity of the materials and gates start flipping when they shouldn't. Hmm. So my 660 TI has been bios modded at 1.21 V but is getting unstable at around 1300 mhz. The temperature hits 66* and stays there because thats when the fans really start kicking in, but they only get to 40-50% duty cycle if I recall. So it seems like I still have some temperature room to work with, is there any other good reason why I couldn't just keep bumping the voltage until the card couldn't keep it at 66 degrees anymore? I guess another thing I haven't thought about is the other components of the card as well. edit: Also by too much current do you mean there is simply a limit to how much power a chip can take before it can no longer function correctly? I never considered this and this would be a new limitation to me. Ignoarints fucked around with this message at 21:18 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 21:15
|
|
|
Nephilm posted:Did I forget to put the quotes on my post? Oh poo poo, I missed the quotes. I didn't realize I was arguing against a generalization, guess I'm just to loving anal. Thanks.
|
|
#
?
Feb 21, 2014 21:22
|
|
|
Ignoarints posted:Hmm. So my 660 TI has been bios modded at 1.21 V but is getting unstable at around 1300 mhz. The temperature hits 66* and stays there because thats when the fans really start kicking in, but they only get to 40-50% duty cycle if I recall. So it seems like I still have some temperature room to work with, is there any other good reason why I couldn't just keep bumping the voltage until the card couldn't keep it at 66 degrees anymore? I never got deeper than the FPGA level, but IIRC once you get to the very small scales chips operate at, you have to look for quantum bullshit when you run electrons through pathways, and there's capacitance to take into account too. I'm sure an actual engineer involved with IC design can explain it better. It's still mostly temperature, though. Heat transfer isn't instant, and running current generates heat, which in turns affects conductivity, so even if you're supercooling components there's a point where you just can't cool enough for the little bits deep inside the circuitry to keep working properly, and in all likelihood with your attempts you're throwing out-of-whack the voodoo magic the designers implemented to make the thing work in the first place. Phuzun posted:Oh poo poo, I missed the quotes. I didn't realize I was arguing against a generalization, guess I'm just to loving anal. Thanks. We're goons in a loving nerd tech thread here, I think mild autism is a prereq; I was just momentarily confused what you were even arguing, it's all good. Nephilm fucked around with this message at 21:40 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 21:34
|
|
|
Nephilm posted:I never got deeper than the FPGA level, but IIRC once you get to the very small scales chips operate at, you have to look for quantum bullshit when you run electrons through pathways, and there's capacitance to take into account too. I'm sure an actual engineer involved with IC design can explain it better. I've been reading up a bit about it and the simplest explanation I can find is that heat increases the atomic vibrations which increase the amount of actual collisions the electrons have with the lattice structure itself. This decreases the overall drift velocity of the electrons, which at a certain tipping point I imagine this would cause a processor to fail that depends on electrons being in the right places by a certain time. Coolness. I was always under the impression that heat was essentially vibrations, it almost seems like an obvious consequence of heat now. But back to reality, im just going to bump the voltage and hopefully it doesnt explode edit: ah nm, it seems like these are usually hard limited to 1.212 v Ignoarints fucked around with this message at 22:19 on Feb 21, 2014 |
|
#
?
Feb 21, 2014 21:45
|
|
|
Heat IS atomic vibrations. I believe there is also a maximum switching speed of a transistor. Not to mention clock generators.
|
|
#
?
Feb 21, 2014 22:27
|
|
|
|
| # ? Apr 23, 2024 16:29 |
|
|
Jago posted:Heat IS atomic vibrations. One of the few things I remember from high school that always stuck to me for whatever reason. That and why things feel colder than other things at the same temperature.
|
|
#
?
Feb 21, 2014 22:31
|
|