|
Gwaihir posted:Selling PC things to Mac users at huge markups is after all a tried and true Apple approved method, too I was just bummed I didn't think of it first. He got 2 of my cards for below the price of 1 he sells. Hah I guess that's the cost of possibly getting an Mac Pro that can actually do VR since Apple won't make you one.
|
#
?
Apr 5, 2016 22:54
|
|
|
|
| # ? May 28, 2024 17:01 |
|
|
THE DOG HOUSE posted:From the official launch date of the 970 / 980 there was practically no gap at all. It was a very good launch, and while availability was scarce - it was not a paper launch. If I recall there were rumors mere weeks leading up to it. And it was available like the next day after the announcement. so really no one has any clue
|
|
#
?
Apr 5, 2016 23:12
|
|
|
RME posted:so really And that's the way we like it.
|
|
#
?
Apr 5, 2016 23:15
|
|
|
Anime Schoolgirl posted:I dunno if that's the case, given the history of the Knights series... I mean it's going to be like 6 months before they're available, maybe a little less if you agree to give NVIDIA a handie in your next research paper. But yeah, hopefully this will push some of the GTX Titans, 780 Tis, and 780s out of the computers of some starving professors. And for that matter I'm not gonna lie, I would looooove to have a K20 or K40, regardless of its unsuitability for gaming  At this point I'm basically expecting this generation to play out exactly like Kepler. The big chips go to the compute guys, we get a GP104 and GP106, and at some point the GP100 filters down to us as a Titan, then a x80 Ti.
|
|
#
?
Apr 5, 2016 23:19
|
|
|
I just hope they give one to Anandtech or similar before too long so I can see the insane benchmark stats.
|
|
#
?
Apr 5, 2016 23:33
|
|
|
Paul MaudDib posted:I mean it's going to be like 6 months before they're available, maybe a little less if you agree to give NVIDIA a handie in your next research paper. The only wild card is whether AMD cocks up the drivers for Polaris and Vega enough that they don't look competitive, although they really seem to trying to fix that with Crimson. What's the chance they're timing their yearly overhaul with a Vega launch (paper or not)?
|
|
#
?
Apr 6, 2016 00:11
|
|
|
Since things move pretty slowly on the low end, is the GTX 950 a solid choice for medium detail 1080p gaming now that it's less than $140? I was hoping for a 970 firesale but that doesn't seem that likely.
|
|
#
?
Apr 6, 2016 00:23
|
|
|
Naffer posted:Since things move pretty slowly on the low end, is the GTX 950 a solid choice for medium detail 1080p gaming now that it's less than $140? I was hoping for a 970 firesale but that doesn't seem that likely. You can get a 950 for $120 AR. Shortie single-fan card, dunno if that helps or hinders you. Or, you can get a 960 for $150 AR with ROTR. If that's the performance level you need then those cards are fine for the cost. They're not as futureproof as trying to track down a 780 Ti or 290/290x for <$200 with 3 GB or 4 GB of VRAM but it's fine if that's what you want to do with it. Paul MaudDib fucked around with this message at 00:33 on Apr 6, 2016 |
|
#
?
Apr 6, 2016 00:30
|
|
|
Paul MaudDib posted:You can get a 950 for $120 AR. Shortie single-fan card, dunno if that helps or hinders you. That's helpful. I checked ebay and it looks like many of the 290's are selling for ~$240 and the 780Ti's closer to $275, which is close enough to new 970 territory. Outside of SA-Mart, it doesn't seem like moderately-aged used cards actually sell at much of a discount.
|
|
#
?
Apr 6, 2016 01:03
|
|
|
The 290's a real solid card, the 950 doesn't have too much room to depreciate, both are likely to be reasonable options.
|
|
#
?
Apr 6, 2016 01:13
|
|
|
I can't wait for Pascal to hit the market soon, I think the release date is around Q3.
|
|
#
?
Apr 6, 2016 01:45
|
|
|
I predict exactly June 16th my uncle works for nintendo
|
|
#
?
Apr 6, 2016 02:02
|
|
|
Naffer posted:That's helpful. I checked ebay and it looks like many of the 290's are selling for ~$240 and the 780Ti's closer to $275, which is close enough to new 970 territory. Outside of SA-Mart, it doesn't seem like moderately-aged used cards actually sell at much of a discount. Check out Reddit's hardwareswap section too, it's the same kind of market as SA-mart. Also, check EVGA's b-stock for 780 Tis, they carry them for $180 + shipping. You will absolutely have to set a site monitor with the keyword "780 Ti" because they go in like less than ten minutes. Typically they get posted around 6:00 to 6:30 (on no particular day of the week). Also, watch ebay carefully. You can pick up a custom 290X for $250 or less, and a 290 for slightly less. Note that you should prefer the custom cards, because the reference blower cooler sucks and will significantly throttle performance. Sometimes they come around really cheap (<$200) and at that price it could be worth it to slap a HG10 bracket and a AIO cooler on it, you can get both of those for around $60 total sometimes. It's a tough market, if there's a discount someone will certainly snap it up quick which can inflate the apparent price.
|
|
#
?
Apr 6, 2016 02:24
|
|
|
Well I hopped on the 970 train, pretty late but I don't think I'll regret it. I've been running a 7870/270x for a couple years and it's starting to crash a lot and give big black boxes and artifacts in games and it's loud as heck.
|
|
#
?
Apr 6, 2016 02:41
|
|
|
Interesting how many people are excited for Pascal Tesla for AI work, considering that all the stuff that makes it expensive (ECC, FP64) is totally useless for deep learning. At the typical pricing for Tesla cards you could build an entire quad-titan X system for the cost of one card. Being 3.5x faster (75% base on top of 2x from FP16 support) doesn't make up for being 5x the price. I guess the power consumption is way better, but that's pretty irrelevant when the card costs more than a decade of 24/7 electricity. The consumer cards will be kickin rad though, whenever they come out.
|
|
#
?
Apr 6, 2016 04:52
|
|
|
Pity the Radeon Pro Duo and its FirePro equivalent are 4 GB and have ecosystem weaknesses, that card absolutely screams in single precision.
|
|
#
?
Apr 6, 2016 05:10
|
|
|
Eyes Only posted:Being 3.5x faster (75% base on top of 2x from FP16 support) doesn't make up for being 5x the price. For some customers, 3.5x perf in exchange for 5x price is fantastic. Pretty sure we'll be lining up for it. Similarly, the value of power efficiency isn't just that it saves opex, but that increasingly data centers are limited by how much power they can get fed from the grid, irrespective of price. These cards allow more computing power to be applied to problems in some scenarios where that wasn't an option at any price before.
|
|
#
?
Apr 6, 2016 05:23
|
|
|
beejay posted:There's no way AMD launches major new products in a wide variety of price ranges (not more rebrands) this spring/summer and Nvidia doesn't... right? xthetenth posted:I think the last node shrink NV was first to was 65 nm with the 8800 GT. Curiously, it feels like the situations are flipped between Nvidia and AMD. Nvidia wants to come out with their latest and greatest but can't because HBM2 ain't cooked yet, whereas AMD feels ready to snare anything and everything they can with silicon that's ready to go well in advance of anything Nvidia might come up with. (God knows we should have had 20nm GCN.) Subjunctive posted:And that's the way we like it. Punditry is fun!
|
|
#
?
Apr 6, 2016 05:38
|
|
|
Subjunctive posted:For some customers, 3.5x perf in exchange for 5x price is fantastic. Pretty sure we'll be lining up for it. Fair, but the consumer flagship Pascal will perform the same at 1/5 the price. Who's to say it won't be on shelves shortly after you finally get your waitlisted Teslas? I guess my point is there's no real reason to pay a premium for enterprise hardware unless your use case actually makes use of the enterprise features. It makes sense if you have an aged system that needs to be upgraded during the Tesla -> Geforce gap (however long that may be) but ML just got excellent hardware with the Titan X last year. It's not like FP64 compute where they're two gens behind and itching to upgrade pronto. Good point on the power hookup limits though. I guess data center sqft and HVAC utilization are important as well.
|
|
#
?
Apr 6, 2016 06:13
|
|
|
Was really hoping for news of a ship date for the new cards. Getting ready to build a new PC to replace the one I'm using which runs a 660gtx and if it could be 8+ months I wonder if I'd be better off just building the PC and getting a mid-range card since I can't convince myself that ~600 USD for a 980TI is worth it at this point, though being able to game in 1440p as an upgrade from a 1920x1200 monitor would be nice. Come on Nvidia. 
|
|
#
?
Apr 6, 2016 07:10
|
|
|
Evil Fluffy posted:Was really hoping for news of a ship date for the new cards. Getting ready to build a new PC to replace the one I'm using which runs a 660gtx and if it could be 8+ months I wonder if I'd be better off just building the PC and getting a mid-range card since I can't convince myself that ~600 USD for a 980TI is worth it at this point, though being able to game in 1440p as an upgrade from a 1920x1200 monitor would be nice. Come on Nvidia. That's where I'm at right now. My 660 is showing its age and I don't know if I wanna wait another 6+ months.
|
|
#
?
Apr 6, 2016 07:51
|
|
|
pctD posted:That's where I'm at right now. My 660 is showing its age and I don't know if I wanna wait another 6+ months. You are all not alone. I should be getting the Rift this month, but I'm currently running on a 2x 275 GTXs, which don't have the power nor the ability to run Direct X 10. I'm between getting a 970 as a holdout until the new cards, or just saying "gently caress it" and get a 980ti.
|
|
#
?
Apr 6, 2016 16:03
|
|
|
Well, Pascal consumer probably won't be announced until June by the looks of things. Even then, we won't be getting big laval. I personally wanted a big Pascal right off the bat for my system, but oh well.
|
|
#
?
Apr 6, 2016 16:22
|
|
|
Yeah, I'm happy I've got a good card already in the 290, I'll be interested in seeing if big Polaris is better by enough to make it worth getting while I wait for Vega. Also, expecting consumer big Pascal off the bat was seriously naive. The hope that'd be worth having is that they release second biggest Pascal like they did second biggest Maxwell and the GTX 970 equivalent is really good.
|
|
#
?
Apr 6, 2016 16:34
|
|
|
If you are going VR, you and I wan't the next Pascale Ti/Titan (if money is no object). The 970 may have been good, but no VR good and the next gen 1070/1080's sound like GDDR5X is going to be used which is fast, but not HBM2 fast like the Titan/Ti appear to be getting along with GP104vsGP100 Currently the 980Ti is a freaking beast, but it can't even keep up with maxed detail for games like Elite or Project CARS in VR. The Rift's ATW works magic to keep things from kicking you in the gut even at 50FPS, but you don't want to stay there forever. Here's hoping that Pascale's Ti can be the one to deliver the performance we need to stay at that sweet 90FPS+ mark in the games the 980Ti can't today.
|
|
#
?
Apr 6, 2016 17:38
|
|
|
SlayVus posted:Well, Pascal consumer probably won't be announced until June by the looks of things. Even then, we won't be getting big laval. I personally wanted a big Pascal right off the bat for my system, but oh well. Isn't it possible that even if fully-functional big Pascal chips are all being reserved for Teslas, there will be enough flawed chips that need to be cut down that the hypothetical 1080 Ti will be released in volume before the quoted Q1 2017 date? The 980 Ti is just a cut-down Titan X so we're expecting a similar pattern here, right?
|
|
#
?
Apr 6, 2016 20:25
|
|
|
Eletriarnation posted:Isn't it possible that even if fully-functional big Pascal chips are all being reserved for Teslas, there will be enough flawed chips that need to be cut down that the hypothetical 1080 Ti will be released in volume before the quoted Q1 2017 date? The 980 Ti is just a cut-down Titan X so we're expecting a similar pattern here, right? The P100 is already a crop chip. Expect Kepler style crop chip Titan after Q1 and the compute customers have had their fill followed by later Ti that might be a full chip.
|
|
#
?
Apr 6, 2016 20:29
|
|
|
Eletriarnation posted:Isn't it possible that even if fully-functional big Pascal chips are all being reserved for Teslas, there will be enough flawed chips that need to be cut down that the hypothetical 1080 Ti will be released in volume before the quoted Q1 2017 date? The flawed chips are reserved for Teslas too - GP100 has 60 SMs but the Tesla P100 only has 56 of them enabled. e;fb
|
|
#
?
Apr 6, 2016 20:30
|
|
|
repiv posted:The flawed chips are reserved for Teslas too - GP100 has 60 SMs but the Tesla P100 only has 56 of them enabled. I just saw that today too... It sucks, but at the same time is exciting if they do it Kepler like with the Ti super beast, vs the Maxwell Titan X king.. We shall see.
|
|
#
?
Apr 6, 2016 20:32
|
|
|
A theoretical Pascal GP100 with the FP64 replaced with FP32 and all 60SMs enabled would be capable of 17TFLOPS of FP32, 3 times faster than a 980Ti  . Not a product that will actually exist but still a crazy stat nonetheless.
|
|
#
?
Apr 7, 2016 02:25
|
|
|
EdEddnEddy posted:
What I'm trying to figure out at this point is the subjective difference between the 970 and 980Ti since as you say the Rift has nice magic to smooth out the lows. I haven't bought a new graphics card since the GTX 460 which has honestly been fine for most of my gaming, but I have a rift on preorder. I can easily buy a 980Ti, I feel a bit foolish for spending $600 on a graphics card, but it's not going to break the bank or anything. But if Pascal is going to be a HUGE jump at the same pricepoint next summer and the 970 isn't a subpar experience with the Rift, I'm thinking it would be better to go that route and save the $300 in the near term. I have some time still, when I preordered my Rift wasn't supposed to ship until May and that's possibly been pushed back now. However, I sure as poo poo don't want that thing showing up and having to wait for a GFX card to ship before I can use it.
|
|
#
?
Apr 7, 2016 02:35
|
|
|
bull3964 posted:What I'm trying to figure out at this point is the subjective difference between the 970 and 980Ti since as you say the Rift has nice magic to smooth out the lows. We might be getting the 1070 or 1080 at Computex in June, but nothing until then and it's anyone's guess as to how papery that launch will be. A full GP104 will probably be 10-15% faster tops at ~$500-600 so if you want performance now then just look for a good sale on a high-end 980 Ti. Last week you could get MSI Golden Edition 980 Tis for $545, that's where I'd pull the trigger if I was looking. The only other wild card is how quick off the line and aggressively priced the crop-GP104 (eg 970 replacement) will be, since that will probably be in the 980 Ti's ballpark in terms of performance and might be $400-ish. I don't think you will lose more than $100 on a high-end 980 Ti card in the near future either way. There's always a better card a summer down the road. When the consumer GP100 (eg Titan) launches it's going to be a $1200 card at first, and then gradually drift down in the x80 Ti crop variant and so on. By then Volta will be on the horizon and so on. If all you want is 970 performance then just buy a 970 though. Or a 780 Ti. Though recommended-spec or no, I don't think the 970 is going to be enough performance for the Rift long-term for serious games on actual engines. Who knows though, maybe this is the kick in the butt to finally optimize game engines a little bit. Paul MaudDib fucked around with this message at 03:13 on Apr 7, 2016 |
|
#
?
Apr 7, 2016 03:08
|
|
|
Paul MaudDib posted:We might be getting the 1070 or 1080 at Computex in June, but nothing until then and it's anyone's guess as to how papery that launch will be. A full GP104 will probably be 10-15% faster tops at ~$500-600 so if you want performance now then just look for a good sale on a high-end 980 Ti. Last week you could get MSI Golden Edition 980 Tis for $545, that's where I'd pull the trigger if I was looking. Honestly, for myself (and likely bull3964), I have a hard time justifying paying a premium price for a GPU when in less than 90 days, it will likely no longer be the best of the best. It's one thing to buy tech when it first comes out, but it's completely another issue when it's about to be overtaken. That's what's really stressing me out here, is that I have no problem paying $600 for a GPU, but I DO have a problem paying for it if it's at the end of it's peak lifecycle. It's like buying a new iPhone a month before the next one is released. Sadly with the timing of the Oculus and Vive coming out, it puts us in an odd position of having to upgrade to soon to be outdated hardware. Obviously if we are willing to buy in to VR so soon, we have no problem being early adopters, we just don't want to be late adopters. What is someone who wants the best bang for their buck for VR supposed to do in this situation? It seems like the best solution is to sit on your non-working VR for a couple of months until Pascal hits, or live with a lower end card until that same point. That's not really pleasant but it certainly saves a minimum of $200 in depreciation costs over buying the very soon to be sunset 980ti.
|
|
#
?
Apr 7, 2016 04:21
|
|
|
Solvency posted:Honestly, for myself (and likely bull3964), I have a hard time justifying paying a premium price for a GPU when in less than 90 days, it will likely no longer be the best of the best. It's one thing to buy tech when it first comes out, but it's completely another issue when it's about to be overtaken. That's what's really stressing me out here, is that I have no problem paying $600 for a GPU, but I DO have a problem paying for it if it's at the end of it's peak lifecycle. It's like buying a new iPhone a month before the next one is released. Sadly with the timing of the Oculus and Vive coming out, it puts us in an odd position of having to upgrade to soon to be outdated hardware. The thing to remember is that there's always a price to pay for having the latest and greatest. When the full GP104 comes out you will pay an extra $100-150 for it relative to the performance it justifies against the 980 Ti, because it'll be the newest and bestest and most power efficient. And stock will be scarce even at retail prices, retailers will scalp like crazy. The flip side of a new card coming out is that you have the chance to purchase the old latest-and-greatest at a major discount. Back when the R9 390 was coming out at $400 a pop for the 390X you could pick up a 295x2 for $550 a pop and get two of the exact same chip under a liquid cooler for $150 more, custom-cooled 290Xs for $250, refurb 290s for $175, etc. B-stock 780 Tis sell for $180 even though the 970 is $250+, etc. May-to-June is probably literally the best time to pick up a 980 Ti. I've already seen top-tier 980 Tis under $550, once NVIDIA pops the announcement I fully expect high-end 980 Tis to start hitting $550 regularly and $475 to $500 on super-sales, with the vanilla cards under $500 regularly and $450 on super-sales. It all depends on whether you literally must have the latest-and-greatest, or whether the best card of the last generation would also be OK. Because the older card is often a bargain relative to the newer one. Paul MaudDib fucked around with this message at 05:13 on Apr 7, 2016 |
|
#
?
Apr 7, 2016 05:07
|
|
|
Maxwell may age poorly though. Pascal looks a good bit more like GCN in some important ways, so development converging on it and GCN might leave Maxwell behind. On the other hand, a lot of transistors is going to be pretty good performance. I don't regret my 290 one bit.
|
|
#
?
Apr 7, 2016 05:23
|
|
|
MaxxBot posted:A theoretical Pascal GP100 with the FP64 replaced with FP32 and all 60SMs enabled would be capable of 17TFLOPS of FP32, 3 times faster than a 980Ti Is that how it works? I don't know anything about FP circuit design. I imagine since FP16 runs at twice the rate as FP32 that they share at least some circuitry, SIMD style. Otherwise Pascal would need waaay more transistors for FP16 to work at that speed. Does the same not apply to FP64?
|
|
#
?
Apr 7, 2016 05:35
|
|
|
Paul MaudDib posted:
That's the $500 million dollar question right now. I think some of us are fumbling because we have no frame of reference. I'm not a hardcore gamer. I get on kicks with certain games and play them all the way through, but it may be months before I pick up another. Traditionally, it's been pretty straightforward. I either tolerate some frame drops or I turn down detail. VR is just a whole other thing. You have to ask yourself if performance will lower the level of immersion or introduce motion sickness. It's no longer about just tolerating some sub-par framerate during intense scenes, it can ruin the entire experience. Why pickup VR if not a hardcore gamer? The long and the short of it is I'm in my mid 30s and I've been hearing about VR all my life. Last time VR was hyped up, I was in my late teens in the 90s and I've been waiting for that experience to live up to promises. Because of that, I'm going to be like a 5 year old at Christmas when that box shows up, I don't think I could tolerate sitting it in the corner for a month or two until the next card releases show up. But, as Solvency noted, I'm not to keen on buying the top end at the end of it's lifecycle without some deep discounts. The discounts will come eventually, it's just a matter of when and what will torture me more (paying too much for a card or letting my Rift sit until the new hardware comes.) Mostly just venting, I don't expect anyone to have any huge words of wisdom since no one has any more information right now. Timing is just bad right now all the way around. It seems like a miss by Nvidia and AMD. If they would have had newer high end options ready for the launch of these headsets, I think they would have moved a LOT of high margin product. Right now I guess it's going to be watch and wait.
|
|
#
?
Apr 7, 2016 05:40
|
|
|
I use to care pretty deeply about the price/performance curve, and it's still interesting to talk about. I'm looking to upgrade, and I'm coming from a 680Ti. Checking my e-mail, I bought it in Sept. 2012. If my next upgrade lasts that long, that's a whole $2/month per $100 extra I spend. It's really hard not to really shoot near the top end of the range when cards last so long these days; which makes buying at the end of a product cycle extra agonizing.
|
|
#
?
Apr 7, 2016 06:51
|
|
|
This issue for me with Pascal is that I'm typing this on a Dell 2412M monitor. By that right, a 970 should 'do' for 1200p for a long time to come, so long as I don't go insane on post-processing, and/or run a few games in a 1680x1050 window. Upgrading to Pascal would invariably involve me buying an Ultrawide display to feed it, to say nothing of the fact that *eventually* I'm going to look into upgrading the rest of my computer. If nVidia put out a 'GTX 975' at the 2 1/2-3 year mark, right when my Gigabyte's warranty will be expiring, I think I'd probably snap it up and wait for Kaby Lake, continuing to use my 2500K until the blue smoke bled out.
|
|
#
?
Apr 7, 2016 06:52
|
|
|
|
| # ? May 28, 2024 17:01 |
|
|
Pascal could let you do things like run DSR to get almost-like-4K image quality out of your games. I'm a fan of waiting these situations, though, so I'm biased.
|
|
#
?
Apr 7, 2016 06:55
|
|
