I'm just happy that I'll be able to get my 1070 this Summer. Who's right and wrong is a trivial issue, really. I do think it's funny how some folks seemed near certain about things they clearly didn't know for certain, but whatever. I just want to upgrade my video card!
How do you know you can get by the summer ?
I don't. Based on past release cycles, it was logical to assume that a release would occur sometime this Summer. The evidence we have seems to suggest that is indeed going to be the case. But until I see the card for sale on newegg.com next to the words "in stock" it's still just an assumption. But it's a safe assumption at this point.
and if TSMC was late with 16nm like they've been "slightly" late in the past it would have been a rebrand
The foundry is the god, judge, jury and executioner, everything else comes afterwards.
TSMC has been producing 16 nm chips for quite some time now. Some iPhone 6S chips are built on TSMC 16 nm, and that launched last September. Still, it's rather harder to build huge, high performance chips than small, low performance chips like cell phones have, and the GPU vendors seem to want TSMC's 16 nm FF+ or Samsung's 14 nm LPP process nodes, which are not the first at their respective foundaries.
I don't see a reason why the 1070 wouldn't ship on time. What factors would lead to that conclusion?
All of the paper launch theory is mainly centered around the availability of GDDR5X. Unless there is a die production problem there isn't a huge flag that says the 1070 won't ship on time. In fact I think it's a huge part of the PR behind this launch. The 1070 will ship on time. The 1080 will "ship" but in absurdly limited quantities with apologetic delays and then will actually hit shelves in early fall.
At some point, Nvidia placed a large-scale production order for GP104 wafers at TSMC. It's really just a question of when they did this, and whether anything subsequently goes wrong to delay a launch. It takes several months for wafers to go through the fabs and all of the subsequent steps in assembling and shipping cards to happen. And you can't hard launch the day the first retailer gets the first cards, as those few cards will sell out almost immediately.
You don't place big orders until you know that they're going to work, which generally means you already have working silicon from a very small order. It's not likely that Nvidia started a big production run last year, or else they could have shown real Pascal chips in January instead of showing Maxwell and claiming it was Pascal. But neither is it likely that they still haven't started production.
If they were going to do a hard launch, then the whole pay $70 more for early access bit makes no sense. So I could realistically see wide availability (say, continuously in stock at MSRP or below on New Egg) on the GTX 1070 starting anywhere between July and October, with only a small chance that it won't happen by the end of October due to unlikely events (e.g., TSMC gets hit by a huge earthquake that ruins the first batch of wafers). June 10 could plausibly be anywhere from a strict paper launch to plenty of "founder's edition" cards for anyone who wants one, but nothing at MSRP.
I'm sorry but moving the goalposts to be defines as "broad retail availability" is just moving the goalposts. The question here was whether the cards would be available on the release dates. Not readily available, available. No high end part in the history of computing has had "broad retail availability" on launch date, at least not in recent memory. There is always more demand than production for these parts. The skylake chips were selling for 20-30% higher than MSRP for SIX MONTHS after release.
What quiz was insinuating was that these were in fact paper launches and we wouldn't see the cards until later this year at the earliest. Then it comes out that the paper launch is early may, then after the paper launch we find out that reviewers have had the cards for a while now and the NDA will be lifted on the 17th, so he then starts backpedaling and moving the goalposts and saying "oh well what I really meant was that they wouldn't be "readily available" until this date".
Trying to use the "oh I can't buy one and have it in a couple days means its a paper launch" argument is just ridiculous and is ignoring the realities of the high end PC component market. AMD does that, and DID do the exact same thing with the R9's. Nvidia does it, and has done it, and will continue to do it (as has AMD). Intel does it, etc etc etc. The list goes on. This is not any of our first times at this rodeo.
Now, there is some validity to the whole "founders edition" cards and pre orders, etc, however I think people are grossly exaggerating how long its going to be for mass volume to be in effect. Nvidia is honestly just being good businessmen and recognizing that there is a portion of the market who is willing to pay through the nose for the right to be the first kid on the block with a new toy.
Keep in mind not 3 weeks ago quiz was certain that we wouldn't see pascal until the very end of 2016 and "probably" 2017. Most of us who were reasonable said "Nah, 2-4 months tops, prob July/Aug at the latest..." and guess what. That's precisely what is looking to be the case. Limited availability (like normal, due to high demand) for the first 4-8 weeks after hard launch, and then readily available after that.
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
High end GPUs are scarce at launch? Well, sometimes.
The Radeon HD 5870 was readily available for about a week at launch, then dried up for a couple months. Rumors say it was a gas leak at one of TSMC's fabs that wiped out a couple months of production or so.
The GeForce GTX 480 was a little scarce on launch day, but was continuously in stock from a week or so later until it got discontinued.
The GeForce GTX 580 had a hard launch and remained available until it was discontinued.
The Radeon HD 6970 had its launch date delayed, but did have a hard launch when it did eventually launch. It would remain in stock for several months before a drought blamed on bitcoin miners.
AMD ended the NDA on the Radeon HD 7870 in late December 2011, saying the cards wouldn't be available until early January. They were widely available from January until it got discontinued.
The GeForce GTX 680 did have a paper launch in March 2012, but real availability came when the GeForce GTX 670 launched in May.
The Radeon R9 290X was continuously available for a month or more after launch, then the cryptocurrency miners bought them up in huge quantities, causing prices to spike. The cards were still widely available, but at a price tag of $100 over MSRP or so.
The GeForce GTX 780 Ti had a hard launch and was widely available until it was discontinued.
The GeForce GTX 980 had a hard launch and has been widely available ever since.
The GeForce GTX Titan X was available only on Nvidia's web site for a few weeks after launch, but has been available ever since then.
The Radeon R9 Fury X had a soft launch last June, and was only sporadically available until September. It has been widely available ever since. The Radeon R9 Fury was widely available much sooner than the Fury X.
So out of all of that, only the GTX 680 and the Fury X were hard to find a month after launch--and it's not a coincidence that both of those launches were pushed up because without those chips, the vendor in question was way behind. The GeForce GTX 1070 may or may not join that club, but it's a near lock that the GTX 1080 is going to be scarce for at least three months after launch and possibly much longer. If that's not unprecedented, it's at least extremely rare.
Keep in mind not 3 weeks ago quiz was certain that we wouldn't see pascal until the very end of 2016 and "probably" 2017. Most of us who were reasonable said "Nah, 2-4 months tops, prob July/Aug at the latest..." and guess what. That's precisely what is looking to be the case. Limited availability (like normal, due to high demand) for the first 4-8 weeks after hard launch, and then readily available after that.
What don't you understand about the impossibility of selling cards using memory that doesn't exist? That's the hold-up on GP100 (HBM2), and it's also forcing the delay of the GTX 1080 (GDDR5X). It's very plausible that Vega would have launched much sooner if HBM2 were available sooner, too, but it's also pushed off until next year because of HBM2.
GP100-based GeForce cards aren't going to be widely available until late this year at the earliest, and likely not until next year. That's not a prediction of GeForce cards based on any other Pascal chip.
I bought a R9 380 with the thought of it to hold me over until some of the high end new cards came out. So it looks like an upgrade may take a bit longer as I decide which to get. I could care less about green or red camp I just get the best card for my money at the time. The 1080 sounds great but we shall see. I typically do not spend over $500 for a graphics card. There was not one I wanted at that price point when I bought the 380 so I just went cheap until the new batches started coming out. Now, it seems, the landscape is beginning to show itself a bit.
It's really tough to keep up with technology. My philosophy is buy the best most reliable card available. Should last you about year and half before it's obsolete lol.
The GTX 1080 and Polaris 10 cards are paper launches. They release them to reviewers but have no availability. Don't expect the GTX 1080 or Polaris cards until Fall. Don't expect Vega or the GP100 until March. The only card with availability before fall could possibly be the GTX 1070 since it uses GDDR5. However, I would be nervous buying this if you are eager to play new games. It doesn't look like nVidia changed the architecture much from Maxwell, so it may not do asynchronous computing well. Asynchronous computing will be very important for GPUs moving forward and is the main reason you are seeing AMD post better scores with DX12 and probably Vulkan on the PC.
I dont see how will they perform (relatively) much better in DX12 than Maxwell. There are some minor architectural changes, but bascially they took Maxwell, lowered "IPC" and raised clocks (on same clocks Maxwell would perform better from what info we have now)
GPUs don't have a clear notion of IPC the way that CPUs do. With any recent GPU, you can get the theoretical GFLOPS as a computation of (number of shaders) * (clock speed) * 2, with the 2 because FMA counts as two operations by convention.
There is the issue of how well you can exhaust that computational capability. In simple synthetics, you can get something like 99% of theoretical peak performance on Maxwell or the various GCN cards, though not on Kepler. Nvidia does have more clever scheduling than AMD, so in real code that is doing complicated things (such as graphics), they tend to be able to come closer to exhausting theoretical peak performance than AMD. AMD generally counters by having higher theoretical peak performance available; I suspect that this is made possible by space and power savings allowed by simpler scheduling. This has been the case every single generation in the entire unified shader era.
I'm personally expecting Pascal to be about as clever about scheduling instructions as Maxwell, and Polaris about as clever as GCN, which would mean that Nvidia maintains their scheduling advantage into the next generation. That advantage is real, but it's much smaller than it was before GCN.
I just used "IPC" as general term to describe it. More important is that Maxwell core works faster than Pascal core on same frequency.
And AMD has definitely donr biggest revision to GCN so far, removing most possible "bottlenecks" they have identified since they started using it in 2011.
I'm sorry but moving the goalposts to be defines as "broad retail availability" is just moving the goalposts. The question here was whether the cards would be available on the release dates. Not readily available, available. No high end part in the history of computing has had "broad retail availability" on launch date, at least not in recent memory. There is always more demand than production for these parts. The skylake chips were selling for 20-30% higher than MSRP for SIX MONTHS after release.
What quiz was insinuating was that these were in fact paper launches and we wouldn't see the cards until later this year at the earliest. Then it comes out that the paper launch is early may, then after the paper launch we find out that reviewers have had the cards for a while now and the NDA will be lifted on the 17th, so he then starts backpedaling and moving the goalposts and saying "oh well what I really meant was that they wouldn't be "readily available" until this date".
Trying to use the "oh I can't buy one and have it in a couple days means its a paper launch" argument is just ridiculous and is ignoring the realities of the high end PC component market. AMD does that, and DID do the exact same thing with the R9's. Nvidia does it, and has done it, and will continue to do it (as has AMD). Intel does it, etc etc etc. The list goes on. This is not any of our first times at this rodeo.
Now, there is some validity to the whole "founders edition" cards and pre orders, etc, however I think people are grossly exaggerating how long its going to be for mass volume to be in effect. Nvidia is honestly just being good businessmen and recognizing that there is a portion of the market who is willing to pay through the nose for the right to be the first kid on the block with a new toy.
Keep in mind not 3 weeks ago quiz was certain that we wouldn't see pascal until the very end of 2016 and "probably" 2017. Most of us who were reasonable said "Nah, 2-4 months tops, prob July/Aug at the latest..." and guess what. That's precisely what is looking to be the case. Limited availability (like normal, due to high demand) for the first 4-8 weeks after hard launch, and then readily available after that.
Skylake chips werent selling, first shipments always have increased price, but bad reviews influenced bad sales. People looked at benchamrks and saw that theres almost no difference to Haswell and Haswell chips were 20% cheaper (+cost of new DDR4 RAM). Skylake chips were NEVER EVER out of stock you could have bought it at any time after launch with no problems whatsoever.
Paper launch DOES include no/super low availability.
High end GPUs are scarce at launch? Well, sometimes.
The Radeon HD 5870 was readily available for about a week at launch, then dried up for a couple months. Rumors say it was a gas leak at one of TSMC's fabs that wiped out a couple months of production or so.
The GeForce GTX 480 was a little scarce on launch day, but was continuously in stock from a week or so later until it got discontinued.
The GeForce GTX 580 had a hard launch and remained available until it was discontinued.
The Radeon HD 6970 had its launch date delayed, but did have a hard launch when it did eventually launch. It would remain in stock for several months before a drought blamed on bitcoin miners.
AMD ended the NDA on the Radeon HD 7870 in late December 2011, saying the cards wouldn't be available until early January. They were widely available from January until it got discontinued.
The GeForce GTX 680 did have a paper launch in March 2012, but real availability came when the GeForce GTX 670 launched in May.
The Radeon R9 290X was continuously available for a month or more after launch, then the cryptocurrency miners bought them up in huge quantities, causing prices to spike. The cards were still widely available, but at a price tag of $100 over MSRP or so.
The GeForce GTX 780 Ti had a hard launch and was widely available until it was discontinued.
The GeForce GTX 980 had a hard launch and has been widely available ever since.
The GeForce GTX Titan X was available only on Nvidia's web site for a few weeks after launch, but has been available ever since then.
The Radeon R9 Fury X had a soft launch last June, and was only sporadically available until September. It has been widely available ever since. The Radeon R9 Fury was widely available much sooner than the Fury X.
So out of all of that, only the GTX 680 and the Fury X were hard to find a month after launch--and it's not a coincidence that both of those launches were pushed up because without those chips, the vendor in question was way behind. The GeForce GTX 1070 may or may not join that club, but it's a near lock that the GTX 1080 is going to be scarce for at least three months after launch and possibly much longer. If that's not unprecedented, it's at least extremely rare.
Theres also a matter of how much has the actuall chip tech changed. 5xx and 7xx series were just respins of previous series, Maxwell was readily available but was also 28nm completely established process.
The GTX 1080 and Polaris 10 cards are paper
launches. They release them to reviewers but have no availability. Don't
expect the GTX 1080 or Polaris cards until Fall. Don't expect Vega or
the GP100 until March. The only card with availability before fall could
possibly be the GTX 1070 since it uses GDDR5. However, I would be
nervous buying this if you are eager to play new games. It doesn't look
like nVidia changed the architecture much from Maxwell, so it may not do
asynchronous computing well. Asynchronous computing will be very
important for GPUs moving forward and is the main reason you are seeing
AMD post better scores with DX12 and probably Vulkan on the PC.
Polaris will definitely be available soon, they showed fully working GPU in December, so i dont really know why it shouldnt be available. If the need arises they can manufacutre it in both GloFO and Samsungs foundries. In fact they said in their financial call its 2Q2016 product 2 weeks ago.
Cool! I won't buy one but it will likely mean they'll lower the prices on other graphics cards. I may buy one of them for my aging computer.
The 9xx generation can't do proper Dx12/Vulkan. I'd be wary with such a purchase. Why do you think they showoff with 1080 but there's deafening silence about 980's Dx12 capabilities?
Async Compute does not work properly on 9 series GTX. It's software emulated through obscene use of context switching which defeats the purpose of it. nVidia is doing damage control and working with developers to mitigate this (Ashes of Singularity) but lets face it ... these cards are going nowhere.
I dont see how will they perform (relatively) much better in DX12 than Maxwell. There are some minor architectural changes, but bascially they took Maxwell, lowered "IPC" and raised clocks (on same clocks Maxwell would perform better from what info we have now)
GPUs don't have a clear notion of IPC the way that CPUs do. With any recent GPU, you can get the theoretical GFLOPS as a computation of (number of shaders) * (clock speed) * 2, with the 2 because FMA counts as two operations by convention.
There is the issue of how well you can exhaust that computational capability. In simple synthetics, you can get something like 99% of theoretical peak performance on Maxwell or the various GCN cards, though not on Kepler. Nvidia does have more clever scheduling than AMD, so in real code that is doing complicated things (such as graphics), they tend to be able to come closer to exhausting theoretical peak performance than AMD. AMD generally counters by having higher theoretical peak performance available; I suspect that this is made possible by space and power savings allowed by simpler scheduling. This has been the case every single generation in the entire unified shader era.
I'm personally expecting Pascal to be about as clever about scheduling instructions as Maxwell, and Polaris about as clever as GCN, which would mean that Nvidia maintains their scheduling advantage into the next generation. That advantage is real, but it's much smaller than it was before GCN.
I just used "IPC" as general term to describe it. More important is that Maxwell core works faster than Pascal core on same frequency.
And AMD has definitely donr biggest revision to GCN so far, removing most possible "bottlenecks" they have identified since they started using it in 2011.
Maxwell cards have a higher ratio of global memory bandwidth to compute power than the GTX 1070 or 1080. There will be many situations where the latter is waiting on memory while Maxwell has the bandwidth and is waiting on things that happen on die. If you disabled all but one memory controller on a Titan X, you'd slow it way down in many things, but IPC is the wrong way to describe it a memory bandwidth bottleneck.
The GTX 1080 and Polaris 10 cards are paper
launches. They release them to reviewers but have no availability. Don't
expect the GTX 1080 or Polaris cards until Fall. Don't expect Vega or
the GP100 until March. The only card with availability before fall could
possibly be the GTX 1070 since it uses GDDR5. However, I would be
nervous buying this if you are eager to play new games. It doesn't look
like nVidia changed the architecture much from Maxwell, so it may not do
asynchronous computing well. Asynchronous computing will be very
important for GPUs moving forward and is the main reason you are seeing
AMD post better scores with DX12 and probably Vulkan on the PC.
Polaris will definitely be available soon, they showed fully working GPU in December, so i dont really know why it shouldnt be available. If the need arises they can manufacutre it in both GloFO and Samsungs foundries. In fact they said in their financial call its 2Q2016 product 2 weeks ago.
GDDR5x that both the 1080 and Polaris use won't be widely available until Fall. You can get it in limited quantity, but not in the millions of chips needed for retail.
The GTX 1080 and Polaris 10 cards are paper
launches. They release them to reviewers but have no availability. Don't
expect the GTX 1080 or Polaris cards until Fall. Don't expect Vega or
the GP100 until March. The only card with availability before fall could
possibly be the GTX 1070 since it uses GDDR5. However, I would be
nervous buying this if you are eager to play new games. It doesn't look
like nVidia changed the architecture much from Maxwell, so it may not do
asynchronous computing well. Asynchronous computing will be very
important for GPUs moving forward and is the main reason you are seeing
AMD post better scores with DX12 and probably Vulkan on the PC.
Polaris will definitely be available soon, they showed fully working GPU in December, so i dont really know why it shouldnt be available. If the need arises they can manufacutre it in both GloFO and Samsungs foundries. In fact they said in their financial call its 2Q2016 product 2 weeks ago.
GDDR5x that both the 1080 and Polaris use won't be widely available until Fall. You can get it in limited quantity, but not in the millions of chips needed for retail.
Polaris is using ordinary GDDR5, not GDDR5X. AMD showed off working Polaris 10 and 11 cards in January, more than two months before GDDR5X started sampling, so working cards with GDDR5X couldn't have existed. Most likely, the reason Polaris doesn't use GDDR5X is precisely because AMD wanted to launch it before GDDR5X would be available.
How much bandwidth you need is a ratio of bandwidth to compute, not an absolute number. The Titan X is a 6.6 TFLOPS card, while the GTX 1080 is a 9 TFLOPS card. To be comparably memory limited, you'd expect the GTX 1080 to need about 30% more bandwidth than the Titan X. Instead, it has less. It may have better texture compression or more L2 cache or some such, but it's probably going to be more memory limited than the Titan X.
For comparison, the Tesla P100 has 720 GB/s of bandwidth, massively more than the GTX 1080, in spite of only having slightly more (single precision) compute power.
How much bandwidth you need is a ratio of bandwidth to compute, not an
absolute number. The Titan X is a 6.6 TFLOPS card, while the GTX 1080
is a 9 TFLOPS card. To be comparably memory limited, you'd expect the
GTX 1080 to need about 30% more bandwidth than the Titan X. Instead, it
has less. It may have better texture compression or more L2 cache or
some such, but it's probably going to be more memory limited than the
Titan X.
For comparison, the Tesla P100 has 720 GB/s of
bandwidth, massively more than the GTX 1080, in spite of only having
slightly more (single precision) compute power.
These tests werent done on some insane clock speeds, its all in the ballpark. I guess we will know for sure soon, but OCing memory didnt have almost any effect on Maxwell so i would presume bandwidth on Maxwell was adequate (960 has 128bit bus)
Is a GTX970 now consider the 'standard' for a gaming card ?
As NVidia showed thesleves 80% of NVidia users still have GPU weaker than PS4.
not really what I asked.
My mother who barely knows how to use a computer qualifies as a 'Nvidia user' so I ask again but somehwhat differently.
Among gamers there is usually a card that is considered the baseline for a good gaming system. I have seen specifically the GTX970 come up in many game specs and many benchmark tests so. Is this card the 'standard' and if not which card is?
Please do not respond to me, even if I ask you a question, its rhetorical.
Some things are changing so you need to qualify what sort of baseline system that is. There is 1080p monitor systems, 4k, and VR. The baseline for each is different if you want a smooth gaming experience.
The baseline for graphics in a 1080p gaming system is probably an R9 370 or a GTX 950. That's about the minimum current card in a system that can still be called a gaming rig. That doesn't account for all the previous generations of hardware. That also doesn't account for 4k or VR.
You're asking a question with no simple answer and it makes me wonder if you have an underlying question you're trying to answer but asking the wrong question to find that out.
then just tell us what the baseline cards are.
simple as that
part two of my question 'if not then what card is'
yes you are allowed to have multiple dimensions to your answer if it makes you feel better I give you permission to do so
Please do not respond to me, even if I ask you a question, its rhetorical.
I did for 1080p gaming. This is really something you could make your own thread about. It's not really about the 1080 or Pascal. Those aren't going to be baseline cards.
ah my bad ok.
GTX950 fair enough.
I havent seen many games with that as a suggested requirement or benchmarks using that card lately but maybe I have been paying attention.
Please do not respond to me, even if I ask you a question, its rhetorical.
Agreed, 950 and 370 (and previous gens equivalents) as entry level gaming cards. But then, when you see how many players LoL and CS:GO and such have and they only really need an APU....what is really "gaming". AAA games only? Latest games only?
Agreed, 950 and 370 (and previous gen equivalents) as entry level gaming cards. But then, when you see how many players LoL and CS:GO and such have and they only really need an APU....what is really "gaming". AAA games only? Latest games only?
to be clear I am not really saying 'entry level'
so in video cards there is a 'sweet spot' its generally the price point in which you are getting the best performance for your dollar. It also often represents the highest volume of sales AMONG the 'gaming card sector' are you familar with this general 'ruleism'? to this is what I am refering
its usually in the middle between the most extreeme in high performance and the 'entry level for gaming'
Please do not respond to me, even if I ask you a question, its rhetorical.
I do not change my computer all that often and when I do I search for the best Nvdia card I can get . I know I am boring but been using Nvidia for a long time so I trust them.
Comments
You don't place big orders until you know that they're going to work, which generally means you already have working silicon from a very small order. It's not likely that Nvidia started a big production run last year, or else they could have shown real Pascal chips in January instead of showing Maxwell and claiming it was Pascal. But neither is it likely that they still haven't started production.
If they were going to do a hard launch, then the whole pay $70 more for early access bit makes no sense. So I could realistically see wide availability (say, continuously in stock at MSRP or below on New Egg) on the GTX 1070 starting anywhere between July and October, with only a small chance that it won't happen by the end of October due to unlikely events (e.g., TSMC gets hit by a huge earthquake that ruins the first batch of wafers). June 10 could plausibly be anywhere from a strict paper launch to plenty of "founder's edition" cards for anyone who wants one, but nothing at MSRP.
I'm sorry but moving the goalposts to be defines as "broad retail availability" is just moving the goalposts. The question here was whether the cards would be available on the release dates. Not readily available, available. No high end part in the history of computing has had "broad retail availability" on launch date, at least not in recent memory. There is always more demand than production for these parts. The skylake chips were selling for 20-30% higher than MSRP for SIX MONTHS after release.
What quiz was insinuating was that these were in fact paper launches and we wouldn't see the cards until later this year at the earliest. Then it comes out that the paper launch is early may, then after the paper launch we find out that reviewers have had the cards for a while now and the NDA will be lifted on the 17th, so he then starts backpedaling and moving the goalposts and saying "oh well what I really meant was that they wouldn't be "readily available" until this date".
Trying to use the "oh I can't buy one and have it in a couple days means its a paper launch" argument is just ridiculous and is ignoring the realities of the high end PC component market. AMD does that, and DID do the exact same thing with the R9's. Nvidia does it, and has done it, and will continue to do it (as has AMD). Intel does it, etc etc etc. The list goes on. This is not any of our first times at this rodeo.
Now, there is some validity to the whole "founders edition" cards and pre orders, etc, however I think people are grossly exaggerating how long its going to be for mass volume to be in effect. Nvidia is honestly just being good businessmen and recognizing that there is a portion of the market who is willing to pay through the nose for the right to be the first kid on the block with a new toy.
Keep in mind not 3 weeks ago quiz was certain that we wouldn't see pascal until the very end of 2016 and "probably" 2017. Most of us who were reasonable said "Nah, 2-4 months tops, prob July/Aug at the latest..." and guess what. That's precisely what is looking to be the case. Limited availability (like normal, due to high demand) for the first 4-8 weeks after hard launch, and then readily available after that.
"The surest way to corrupt a youth is to instruct him to hold in higher esteem those who think alike than those who think differently."
- Friedrich Nietzsche
The Radeon HD 5870 was readily available for about a week at launch, then dried up for a couple months. Rumors say it was a gas leak at one of TSMC's fabs that wiped out a couple months of production or so.
The GeForce GTX 480 was a little scarce on launch day, but was continuously in stock from a week or so later until it got discontinued.
The GeForce GTX 580 had a hard launch and remained available until it was discontinued.
The Radeon HD 6970 had its launch date delayed, but did have a hard launch when it did eventually launch. It would remain in stock for several months before a drought blamed on bitcoin miners.
AMD ended the NDA on the Radeon HD 7870 in late December 2011, saying the cards wouldn't be available until early January. They were widely available from January until it got discontinued.
The GeForce GTX 680 did have a paper launch in March 2012, but real availability came when the GeForce GTX 670 launched in May.
The Radeon R9 290X was continuously available for a month or more after launch, then the cryptocurrency miners bought them up in huge quantities, causing prices to spike. The cards were still widely available, but at a price tag of $100 over MSRP or so.
The GeForce GTX 780 Ti had a hard launch and was widely available until it was discontinued.
The GeForce GTX 980 had a hard launch and has been widely available ever since.
The GeForce GTX Titan X was available only on Nvidia's web site for a few weeks after launch, but has been available ever since then.
The Radeon R9 Fury X had a soft launch last June, and was only sporadically available until September. It has been widely available ever since. The Radeon R9 Fury was widely available much sooner than the Fury X.
So out of all of that, only the GTX 680 and the Fury X were hard to find a month after launch--and it's not a coincidence that both of those launches were pushed up because without those chips, the vendor in question was way behind. The GeForce GTX 1070 may or may not join that club, but it's a near lock that the GTX 1080 is going to be scarce for at least three months after launch and possibly much longer. If that's not unprecedented, it's at least extremely rare.
GP100-based GeForce cards aren't going to be widely available until late this year at the earliest, and likely not until next year. That's not a prediction of GeForce cards based on any other Pascal chip.
거북이는 목을 내밀 때 안 움직입니다
And AMD has definitely donr biggest revision to GCN so far, removing most possible "bottlenecks" they have identified since they started using it in 2011.
Paper launch DOES include no/super low availability.
Polaris will definitely be available soon, they showed fully working GPU in December, so i dont really know why it shouldnt be available. If the need arises they can manufacutre it in both GloFO and Samsungs foundries. In fact they said in their financial call its 2Q2016 product 2 weeks ago.
Async Compute does not work properly on 9 series GTX. It's software emulated through obscene use of context switching which defeats the purpose of it. nVidia is doing damage control and working with developers to mitigate this (Ashes of Singularity) but lets face it ... these cards are going nowhere.
Titan X - 336,5 GB/s
980ti - 336,5 GB/s
1080 - 320 GB/s
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me
For comparison, the Tesla P100 has 720 GB/s of bandwidth, massively more than the GTX 1080, in spite of only having slightly more (single precision) compute power.
GTX970/R9 290 are entry level for VR. AMD said that there are only 7,5 million "VR ready" machines.
These tests werent done on some insane clock speeds, its all in the ballpark. I guess we will know for sure soon, but OCing memory didnt have almost any effect on Maxwell so i would presume bandwidth on Maxwell was adequate (960 has 128bit bus)
My mother who barely knows how to use a computer qualifies as a 'Nvidia user' so I ask again but somehwhat differently.
Among gamers there is usually a card that is considered the baseline for a good gaming system. I have seen specifically the GTX970 come up in many game specs and many benchmark tests so. Is this card the 'standard' and if not which card is?
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me
simple
as
that
part two of my question
'if not then what card is'
yes you are allowed to have multiple dimensions to your answer if it makes you feel better I give you permission to do so
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me
GTX950 fair enough.
I havent seen many games with that as a suggested requirement or benchmarks using that card lately but maybe I have been paying attention.
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me
so in video cards there is a 'sweet spot' its generally the price point in which you are getting the best performance for your dollar. It also often represents the highest volume of sales AMONG the 'gaming card sector' are you familar with this general 'ruleism'?
to this is what I am refering
its usually in the middle between the most extreeme in high performance and the 'entry level for gaming'
Please do not respond to me, even if I ask you a question, its rhetorical.
Please do not respond to me