NVIDIA GeForce GTX 880 Details Surface

Published by

Click here to post a comment for NVIDIA GeForce GTX 880 Details Surface on our message forum
https://forums.guru3d.com/data/avatars/m/223/223793.jpg
well that surely sounds far more interesting than both titan-z and 295x2.
https://forums.guru3d.com/data/avatars/m/230/230424.jpg
256-bit???
https://forums.guru3d.com/data/avatars/m/154/154498.jpg
256-bit???
My thoughts exactly... 7400 MHz GDDR5 to push through that narrow bus width?
data/avatar/default/avatar21.webp
I don't understand the whole 256bit, 384bit and 512mb bit thing, just a number to me that is meaningless and has no impact on anything, it doesn't seem to matter in games at all. It is like ram latency and speed, all meaningless to me, none of it impacts game performance, I've tried loads of different types of ram and the FPS is always the same. Same with bit rate, has no impact on games.
https://forums.guru3d.com/data/avatars/m/142/142454.jpg
I don't understand the whole 256bit, 384bit and 512mb bit thing, just a number to me that is meaningless and has no impact on anything, it doesn't seem to matter in games at all. It is like ram latency and speed, all meaningless to me, none of it impacts game performance, I've tried loads of different types of ram and the FPS is always the same. Same with bit rate, has no impact on games.
If you imagine the amount of data that your gfx card is processing as a motorway, the speed of the traffic is the memory speed and the bus size is the number of lanes. Bandwidth = bus size x memory speed. If you have enough bandwidth then great and they don't build gfx cards that are crippled in that aread but trust me, there is a point where if the memory bandwidth is too low, your gfx card performance will suffer and bottleneck the rest of the GPU. Awesome that 880 rumours are starting. It looks like another 680 (not the full die) Is this a response to the leak of Volcanic Islands being expected this year? What happened to 2015 for 20nm? Exciting times ahead! Yay for competition.
https://forums.guru3d.com/data/avatars/m/230/230424.jpg
agreed on the system ram part. From what ive read and tried myself, system ram and game performance seems to be based on how much you have more often than if its 'performance' ram or not. I didnt want to be the first to post the - another 680 - comment. Glad im not alone on that thought. As for the 256-bit, Its almost the same as slapping stupid amounts of ram on low to mid-range cards. Why!
https://forums.guru3d.com/data/avatars/m/239/239932.jpg
Boring. I'll wait for a 980Ti something instead.
https://forums.guru3d.com/data/avatars/m/142/142454.jpg
As for the 256-bit, Its almost the same as slapping stupid amounts of ram on low to mid-range cards. Why!
I suppose because 256-bit is sufficient and wont cause a bottleneck.
https://forums.guru3d.com/data/avatars/m/230/230424.jpg
I suppose because 256-bit is sufficient and wont cause a bottleneck.
But you can load the vram as much as you want, but it the bus is still only got x amount of lanes, then its still going to take the same amount of time. Sure theres the 'its more effective at higher res's' argument, but then youd be limited by the performance of the actual gpu itself. if these indeed turn out to be the final product, then i still go to the 880 as the cost of a new highend is the same as the cost of the previous gens highend, the prices dont drop in NZ, they just wait until stock has sold out and sell the new release highend and last highend side by side with a $40-60 price diff.
https://forums.guru3d.com/data/avatars/m/202/202673.jpg
What's this...nVidia turning 20nm wafers into scratched records?
https://forums.guru3d.com/data/avatars/m/230/230424.jpg
What's this...nVidia turning 20nm wafers into scratched records?
holy fu.. hahaha.
https://forums.guru3d.com/data/avatars/m/115/115710.jpg
So this is definitely not aimed for 4K gaming then. They'd need to increase the ROP count a lot and that memory bandwidth does not help. Assuming that those numbers are correct.
data/avatar/default/avatar32.webp
I dunno I'm using a 750ti now that is 128bit and it performs like my 670 did, I don't see it being a bottleneck at all, this card is just amazing. I'm running a triple monitor setup at 5000something x 1080p and getting 60FPS while maxing out Assetto Corsa lol. Everyone on the forums talking about a 780ti for it.... well I spent £106 and am doing it just fine on this and it overclocks like mad with no external power connector. I'm never spending loads on a GPU again, seems to me the CPU is the major bottleneck right now as games for some reason love maxing 1-2 cores and ignoring the rest. The new Nvidia beta drivers have increased my FPS massively too, I'm getting better FPS in Dayz now than I ever did with my 670SLI setup. SLI doesn't work, just like Crossfire, it isn't worth it, I've tried both and like 2 games I've ever played have made use out of it.
https://forums.guru3d.com/data/avatars/m/182/182702.jpg
These are not final but... I don't know... I think I expected more for this 880 (on the paper). Besides CUDA cores, most of the numbers are similar/lower than 780. The memory bandwidth looks kinda narrow which is a shame because I thought about this 880 as an upgrade of my 680 suffering from its own limited memory bandwidth. But with the efficiency of the new chip, who knows... And the price also 🙂 Wait & See
https://forums.guru3d.com/data/avatars/m/182/182702.jpg
So this is definitely not aimed for 4K gaming then. They'd need to increase the ROP count a lot and that memory bandwidth does not help. Assuming that those numbers are correct.
Exactly! Maybe they will release a "monster" version (aka TI) for that matter, leaving the 880 as a high end product for smooth 1080p but no more.
https://forums.guru3d.com/data/avatars/m/245/245459.jpg
My thoughts exactly... 7400 MHz GDDR5 to push through that narrow bus width?
Maxwell works very efficiently with what we would consider a narrow memory bandwidth. I remember reading this in an article outlining the technology behind Maxwell and the 750ti when it launched. Effectively, it uses the memory bandwidth so efficiently that it's almost a factor of two, so the 238GB/s memory bandwidth figure would effectively be more like 400GB/s when making a comparison to Kepler architecture that we know so well. Part of this efficiency of memory bandwidth use is based on the Maxwell having a large L2 cache. EDIT: Just found a quote from a 750ti Maxwell article that indicates what I'm talking about: "The number of ROPs remains unchanged, while the size of the L2 cache has increased by 8X. The increase in L2 cache helps the card become less dependent on memory for bandwidth and makes memory less of an issue." Here's the link: http://www.brightsideofnews.com/news/2014/2/18/nvidias-gtx-750-ti-introduces-maxwell-and-a-whole-lotta-cache.aspx I'm sure there's lots of other articles out there saying the same thing. Don't fret about the lower memory bandwidth on Maxwell cards.
data/avatar/default/avatar29.webp
If that price is what Nvidia are going to charge for GPU's now especially when it's a GTX680 like product they can f*** right off, it's basically the same as a 780Ti!
https://forums.guru3d.com/data/avatars/m/254/254955.jpg
If that price is what Nvidia are going to charge for GPU's now especially when it's a GTX680 like product they can f*** right off, it's basically the same as a 780Ti!
I agree, if performance would be better for like 5-10% I would get 780Ti for replace my 670 with no doubt, but something tells me that there is MIGHT BE* a 880Ti 880SuperTi 880OMFGTi and etc. After all this can be just not truth.
https://forums.guru3d.com/data/avatars/m/230/230424.jpg
Maxwell works very efficiently with what we would consider a narrow memory bandwidth. I remember reading this in an article outlining the technology behind Maxwell and the 750ti when it launched. Effectively, it uses the memory bandwidth so efficiently that it's almost a factor of two, so the 238GB/s memory bandwidth figure would effectively be more like 400GB/s when making a comparison to Kepler architecture that we know so well. Part of this efficiency of memory bandwidth use is based on the Maxwell having a large L2 cache. EDIT: Just found a quote from a 750ti Maxwell article that indicates what I'm talking about: "The number of ROPs remains unchanged, while the size of the L2 cache has increased by 8X. The increase in L2 cache helps the card become less dependent on memory for bandwidth and makes memory less of an issue." Here's the link: http://www.brightsideofnews.com/news/2014/2/18/nvidias-gtx-750-ti-introduces-maxwell-and-a-whole-lotta-cache.aspx I'm sure there's lots of other articles out there saying the same thing. Don't fret about the lower memory bandwidth on Maxwell cards.
Ah, nice find. il give it a read. As for price, i think we know the drill now. Also, will we see a 8800gtx version for old time sake, if so the specs above arent gonna be enough to live up to the legacy.:D
https://forums.guru3d.com/data/avatars/m/175/175902.jpg
I suppose because 256-bit is sufficient and wont cause a bottleneck.
yes the number of bit is just to compare the size of the "door" but if the exchange are done faster, then 256 bit is sufficient and can be faster in overal perf than a slower 384 bit. about bottleneck... it's mainly an urban legend on modern PC despite what people write...