Nvidia Announces GeForce GTX 1080 Ti; $700, Coming Next Week

Page 5 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

bit_user

Titan
Ambassador
They are not tiny PCs. Processors have been inside HDDs for decades, although the details change.

This is irrelevant to the subject of GPUs, because the processors in SSDs are tiny, by comparison. Ultimately, what makes a GPU fast is what requires the large die area, with which price scales non-linearly. See above.

They got cheap because the functionality stayed relatively fixed. So, they were able to use smaller and smaller dies, as silicon fabrication improved. Smaller dies are cheaper, and facilitate more integration, which saves more cost. At the end, you have a board with a relatively inexpensive chip on it, and not much else. That's why it got cheap.

GPUs don't work like this, because they keep getting bigger and requiring more memory.


R & D costs vary widely, as does the amount you can recoup, when you go to do the next generation. Because things like sound card chips are relatively simple, they can use a simple ASIC process and reuse much of the design for the next generation.

Not if you just take advantage of the smaller lithography to pack more compute, cache, etc. on the chip. See above. This is the point you're missing.
 


I can give you an easy example that goes against your logic, with your most common example: SSDs

2 or 3 years ago, the flagship consumer SSDs were worth about $500, maybe even $800 or $1000. Now we have SSDs that are worth close to $2000 (the new 2tb flagships).
The price has gone UP, and a lot, because they offer more stuff (higher capacity). But it's still the flagship, and according to your logic they should be cheaper.
 


Gigabyte Does
13-128-964-V04.jpg


Phanteks Does
articleview.cfm


We do ... :)
2398399

2391222

2415094

2391221





 

ledhead11

Reputable
Oct 10, 2014
585
0
5,160


Not normally a fan of orange(even though I use a lot of Gigabyte), but d**n those are some nice looking setups.
 


Hmmmmm, that board interests me. Though, for the case I think I'd like something black/orange. I'll admit, a small fantasy of mine is to get a case that replicates the computers from Aperture Science (Portal). I love how those chairs look too, but I know the ergonomics would leave a lot to be desired xD
 

ledhead11

Reputable
Oct 10, 2014
585
0
5,160
Just a heads up. . . .

I haven't seen any pre-order links other than NV but I've seen articles for EVGA, ZOTAC, ASUS, and INNO3D for their variants. Most saying to be available by the end of the month and the OC editions shortly after. Saw 'em on Guru3d and Overclock3d.

Update: Just went to most major manufacturer sites and most already have listings for their FE versions. Some have the standard NV cooling while a few have brand specific cooling but all with the same clocks. Some have started listings for OC editions but no specs yet.

Oh, well. Now we just have to wait for the tidal wave of reviews.
 

ledhead11

Reputable
Oct 10, 2014
585
0
5,160


I was reading on Guru3d that the review embargo ends on the 10th when purchases become available. As of yesterday a number of sites were allowed, however, to simply show thier NV cards open from packaging as well as disassembling them to show the newer cooling solution. There are some significant changes.
 

bit_user

Titan
Ambassador
Huh?

No, that's only necessary when the reviews are going to contain some disappointments. With the similarity of this to Titan XP, I doubt there'll be any surprises. The launch inventory will probably sell out within hours.
 

ledhead11

Reputable
Oct 10, 2014
585
0
5,160


I fully agree. I honestly expect this generation of TI, including the likely OC custom versions later, to set a record for GPU sales in 2017.

My 1080SLI setup usually averages 20-40% over a single 1080 and pretty much rocks 4k/1440p gaming. The expected FE's are estimated 30-35% over a single 1080 and performance should be very similar sans the need for SLI in 4k/1440p gaming.

This means people with or building a mid, even possibly low, range PC can save their money for the GPU and really enjoy 4k or 1440p with little or no compromise now that the displays are becoming more affordable.

 

Geezer760

Distinguished
Aug 29, 2009
219
108
18,870
I Think $500 and up are just ridiculous prices for video cards,just to play a $60 game when that game will fall in price in a few months, Video cards are just way overpriced these days, and next year Hey guess what another NEWER better card there goes another $500,$600,$700. on and on an on. squeezing your wallets out of your cash again.
 
One has to remember that for a long time, in the mind of consumers the best card is the one from the best card's maker - at a time when you had a couple tier, it could be defended - owning a 3dfx was owning the best 3D card; owning a Geforce3 was too; etc.
Thus, the reason why Nvidia and AMD still duke it out on the high end with Nvidia throwing its cash around to build teh l33t35T chip ev4r (GP102, GP105). Had Intel decided to go forth with Larrabee, we may have had an upset, as it was powerful, and expensive - unreasonably so. Still, in the mind of many, gaming on PC today means Nvidia - because the best chip is from Nvidia.
But, if you want to game with proper performance without having to sell a kidney, things get muddier - usually in favour of AMD, if the successes of Radeon 9500 Pro, RadeonHD4850 (recommended at Tom's for more than 2 years), RadeonHD7770 (took up the crown at the $100 level and stayed there for almost as long) and now confirmed worthiness of the RX470/RX480 are any indication.
THAT's where the core market is at: reasonably priced, reasonably powerful GPUs... And Nvidia knows it too. However, the tiers are moving to compensate both for inflation, and also for the raising baseline that require more pure horsepower: while it was fine for a long while to sell DX7 compatible entry level cards and leave DX8/DX9 level features for the high end, starting with DX9 ALL chips from the low end to the high end had to support the same feature set - making creating a GPU a both more complicated and simpler matter: compute units must be able to do everything, and power is now a matter of parallelizing and increasing clock speed. Since we're getting ever closer from then end of Moore's law, chips are getting bigger and bigger and as such more sensitive to defects and as such more power means bigger chips that cost more to make - while at the same time consumers expect power to increase all the time.
I mean come on! 10 years ago playing at 1280x800 was the kicks and FullHD was golden, today 1920x1080 is standard - that's twice as many pixels to push without counting stuff like HDR, tessellation, fatter textures, better physics etc. and pure gamers are asking for VR (90fps minimum), 4K (4 times the pixels)... So yes, the top end is costing much, much more than it used to and mid range is evolving along with it - but since there's now a limit to the amount of power one can get out of the silicon, prices scale logarithmically with power.
 
Status
Not open for further replies.