Separate names with a comma.
Discussion in 'nVidia Flavor' started by Putz, May 8, 2018.
so are we going to have systems with more VRAM than system ram? seems like major overkill
If it's 256 bus (as rumored) it's either 8GB or 16GB. It's a toss up in my view. Wouldn't be surprised if we get an 8gb and 16gb version (maybe with 16gb coming first).
So true. Nvidia is not launching 1180 at 400 bucks. Heck they wouldn’t even do it if they had competition.
It’ll be 20-25% at least faster than the 1080ti for gaming and use less power and be 40-70% better for my crypto mining compliments of gddr6 (Titan V’s hashrate is > 100% faster than 1080ti)
They can’t keep 1080ti cards in stock now for $700-$800. Why would they release a faster, more energy efficient card for less than the card that is still currently selling so fast it is sold out nearly instantly when available for $700.
It won't be $800. Not sure why people keep saying that. The FE 1180's will be $699. Aftermarket 1180's will probably more, but only because AIB's and resellers will add their own markup, but Nvidia direct will not.
Because Nvidia already has publicly said prices are going up —- and ....primarily because the hot market says they can.
Maybe you are right about 800. What you said is the particular reason they might milk the cards out even longer. I mean I wouldn't be surprised if they push the launch in to late 2018. You are right 1080ti sell out even for 700. I got one for sub 600 here in the forum and I plan to hold on to it until next gen. Either navi or whatever nvidia has coming after this.
it was all over the rumor mongering sites in February, but none of the sites that only publish confirmed stories carried it. That may mean the echo chamber just was echoing itself, or despite a minor leak NVidia is keeping the attempt close to vest.
One thing to remember: the GV104 and associated cards will be cheaper to produce at volume than the GP102 that the 1080Ti uses. Smaller GPU die, fewer memory traces for less complex PCB routing, less power so less complex power delivery all mean that the future 1180's (and 1170's and whatever else) make Nvidia money by taking over the 1080Ti's price/performance slot.
It's also important for them to sell them in volume, which is the big factor in terms of pricing: the Gx104 GPUs used to be relegated to the GTXx60 up until the 600-series. Nvidia sold them for ~$250 and less (see GTX560 and GTX560Ti). I realize that Nvidia may have upsized the Gx104 GPUs a bit, but they're still smaller than their top-end compute Gx100 and gaming Gx102 parts, and they still run the narrower 256bit memory bus. Hell, the TDP is still the same at ~160-180W!
That’s what Nvidia’s done and it works our quiet well. They start with pro cards and have done an epic job making sure cut down gaming version use power according to specs if you know what I mean. I think AMD was doing the other thing. One design fit all, it may be have been cheaper for R&D but we can clearly see that Nvidia approach worked much better as it resulted in more lean and power effecient design.
I think Lisa might go the same route. You don’t need a lot of the features for gaming cards so keep them separate. With more money coming in from Ryzen it might allow them to do just that.
I honestly hope they do- they haven't been able to meet Nvidia in terms of efficiency at high-performance in over half a decade, and that has kept them from top-end competition.
Of course, they also need to catch up on releases. Coming up several years behind for the same power and performance while their competition already has something faster and leaner is not helping!
Perhaps if they diversified their lines as above they'll be able to iterate and improve at a quicker pace. Would certainly incentivize Nvidia to keep their own product lines in line, though I will have to admit that I've been fairly impressed with Nvidia's MSRPs lately. They could certainly have priced higher, even when market demand was lower, but it seems that they've decided to focus on market penetration and volume over unit margins.
GCN was pretty epic early one, but AMD stuck to it far too long. Lisa turned around CPU side pretty quickly, I think she had too much faith in Raja who it seemed already had one leg out the door. I was reading somewhere how she had some of the engineers that helped with ryzen help out RTG to improve efficiency. I sort of glanced over it and hopefully I read that right lol. I think not releasing anything is better than releasing a over hyped product. Mining might have been the saving grace for AMD. It sort of worked perfectly. Hopefully she has them guys working as RTG has been fairly silent lol.
With GDDR6 you can also have 12 GB of vram on a 256 bit bus.
"Today, a GPU with a 256-bit bus can only cleanly support 4GB, 8GB or 16GB. With GDDR6, they will also be able to support 12GB, while still maintaining a full balanced load with identical sized memories connected to each controller."
It won't be 800. Retail price and market value have no correlation.
No doubt. Especially if it's hashrate is better than 1080 ti. In any case, it's power effeciency will almost certainly be better.
I think we may see 1180 going for well over $900.
Link? The only thing I recall JHH ('them') saying in interview last year was that Voltas manufacturing costs were high. No other evidence of "Nvidia already has publicly said prices are going up". And no mention of Turing prices. I could be wrong, but can you provide link or reference to that?
How do you know they are not winding down stocks of 1080/1080ti purposely to make room for new cards to replace them? An 1170/1180 should outperform those cards easily, so why would they encumber themselves with old stocks? I doubt very much they are still making 1080/1080ti and THAT is the reason they are out of stock.
GV100's manufacturing costs are high, not Volta's.
haHaaa! 900$ MSRP for next x80 card. Gotcha.
I can see MSRP being $750 to $800 but not $900. Miners are not buying new cards atm cause the ROI is shit atm. Getting a little better but no where near how it was in December and January.
Exactly, it's a larger chip, more silicon per chip and your costs are going up.
Most important thing: Whens the Ti/TitanX come out?
I will probably do something I have never done before and skip the cutting edge, upgrade my 290s with last gen Ti or V64 in the Fall.
This man knows his stuff scalpers will be out in full force unless nvidia has a million of the cards at launch hopefully not all FE cards. I could easily see Sold out for a good 2-3 months with 1.4k prices.
I don't think you understood my point, GV100 is >800mm². Whatever the x04 chip will be, it won't be any larger than GP102, so what exactly is the problem?
1.4k huh ?
I think you're all full of shit, it's going to be selling for several million dollars within a week of launch.
I honestly think this upcoming generation is ripe for price gouging by nVidia. I'm taking a wait-and-see approach, personally. I may not have a true upgrade path for some time with my GTX 1080 Ti SLI setup. What we need is for AMD to be competitive. I am hopeful that these 7nm Vega rumors are true and that they are able to ramp up the clockspeed and lower the power consumption as a result. But no, I don't expect die shrunk Vega to be competitive with Geforce Volta. Perhaps with Pascal to some extent, which would be nice.
Something like 9-15 months, and 0-6 months later respectively based on recent history.
I really like the ability to handle 1.5x capacity configurations without sacrificing either total bandwidth or having more ram on some controller ports.
That said, I wonder if the driving reason behind this is yield management on the DRAM makers part. If they have a 4 bank internal structure for their GDDR6 chips, a 1.5Gb chip could just be a 2Gb one that had one of its 4 banks be faulty allowing them to salvage it instead of junking it similar to the 3core parts AMD sold some years back.