Nvidia GeForce GTX 1050 (3GB) has a 96-bit memory bus

Published by

Click here to post a comment for Nvidia GeForce GTX 1050 (3GB) has a 96-bit memory bus on our message forum
https://forums.guru3d.com/data/avatars/m/49/49225.jpg
Looks like a lot of dud chips getting a new home.
data/avatar/default/avatar27.webp
Watch the Cuda Cores. It says 768, which means that they use a full 1050Ti with 3Gb Ram! Edit: Would this GPU be more powerful running older games? The boost clocks would intent this, but you lose memory bandwith or am i wrong?
https://forums.guru3d.com/data/avatars/m/258/258688.jpg
yoshi0597:

Watch the Cuda Cores. It says 768, which means that they use a full 1050Ti with 3Gb Ram! Edit: Would this GPU be more powerful running older games? The boost clocks would intent this, but you lose memory bandwith or am i wrong?
The 96-bit wide bus cripples the product. Even a 128-bit bus GPU is dog slow compared to a proper 256-bit bus GPU, etc., even if everything else is equal.
data/avatar/default/avatar32.webp
Thank you for the fact based informative reporting in this article. I first read this information on another site just minutes before and it was written as though some wildly nefarious action was taking place rather than a specification change as the fact it is. Consumer is aware, specification is public, decision can be made. Lower memory at higher bus width or higher memory at lower bus width on a product that remains a low end product.
data/avatar/default/avatar11.webp
Wonder how it will fare for upscaling to 4K with madvr.
https://forums.guru3d.com/data/avatars/m/34/34585.jpg
Another fk up by nVidia, really the GPP program was transparency and you do this sort of shit, like the 1030, one with GDDR5 and the other with GDDR4 WTF the latter has less than half the bandwidth it's technically a completely different SKU could have used the 1020 for the DDR4 model, then the GTX 1060 3GB (1152 shader) vs the 1060 with 6GB (1280 shaders) with the latter being 10-30% faster could have used the GTX1055. Now you have the 1050 2/4GB being faster in bandwidth scenarios than then 3GB model but in fill rate the 3GB should be faster. I bet bandwidth limitation will kick in before fillrate. Nvidias Line up is a total mess, GPP couldn't save this mess with this alleged transparency.
https://forums.guru3d.com/data/avatars/m/40/40086.jpg
Who do they aim products like this at? What purpose would this card serve to a consumer? Is it targeted at OEMs? Is it for retail? People who know what they are buying wouldn't look at this card, so is this just to add to the confusion for people who don't?
data/avatar/default/avatar05.webp
VAlbomb:

Wonder how it will fare for upscaling to 4K with madvr.
Precisely. Is this the perfect GPU for people looking for a 4k HEVC HDR ready Home Theater PC? If the extra RAM makes it a better choice performance-wise and the weaker gaming potential makes it better value-wise for this one specific usage case, so be it. Otherwise, who needs it?
data/avatar/default/avatar10.webp
Radeon 9700 was 256-bit. NVidia always sucked in this matter when compared to ATI. But those Pascal+ cards do use some sort of new memory compression algorithms. Much will depend on price and how well VRAM will overclock.
data/avatar/default/avatar23.webp
dragonlord:

Precisely. Is this the perfect GPU for people looking for a 4k HEVC HDR ready Home Theater PC? If the extra RAM makes it a better choice performance-wise and the weaker gaming potential makes it better value-wise for this one specific usage case, so be it. Otherwise, who needs it?
The regular 1050 does the job just fine for an HTPC with 4K HDR HEVC, no real benefit to adding more RAM, CUDA cores or clock speed. All of the video decoding is done by the NVENC engine built into the chip, the actual CUDA stuff is barely used. The 1050 can even do hardware HEVC encoding at a good clip (5x-10x playback speed depending on the content). I suppose if you wanted to not use NVENC for better quality it might have some benefits, but for most common HTPC builds the base 1050 is pretty much the perfect GPU.
https://forums.guru3d.com/data/avatars/m/246/246171.jpg
Size_Mick:

Who do they aim products like this at? What purpose would this card serve to a consumer? Is it targeted at OEMs? Is it for retail? People who know what they are buying wouldn't look at this card, so is this just to add to the confusion for people who don't?
Yeah I don't really get it either. When AMD does this, sure, it's annoying and stupid, but it's usually for OEMs. This isn't the first time Nvidia has done this. For example, I remember there were 3 different flavors of the GT 630, each of which were very different from each other (in fact, 2 were Fermi while one was Kepler based) and to my knowledge, none of them were specific to OEMs. It seems for anything below ##50Ti, Nvidia just recycles the performance tiers in whatever way conveniently gets rid of binned parts. I'm not sure if I prefer this, or AMD's method of "let's rebrand these same binned chips for the next 4 years".
https://forums.guru3d.com/data/avatars/m/243/243702.jpg
Size_Mick:

Who do they aim products like this at? What purpose would this card serve to a consumer? Is it targeted at OEMs? Is it for retail? People who know what they are buying wouldn't look at this card, so is this just to add to the confusion for people who don't?
Purpose is to get additional sales of bad apples. Meant to be sold to confused people.
https://forums.guru3d.com/data/avatars/m/270/270041.jpg
schmidtbag:

Yeah I don't really get it either. When AMD does this, sure, it's annoying and stupid, but it's usually for OEMs. This isn't the first time Nvidia has done this. For example, I remember there were 3 different flavors of the GT 630, each of which were very different from each other (in fact, 2 were Fermi while one was Kepler based) and to my knowledge, none of them were specific to OEMs. It seems for anything below ##50Ti, Nvidia just recycles the performance tiers in whatever way conveniently gets rid of binned parts. I'm not sure if I prefer this, or AMD's method of "let's rebrand these same binned chips for the next 4 years".
From what i understand these new cards are coming from Bad chips, in this case bad 1050ti chips, that don't quite meet the standards. Least this is what i have gathered... I just wish they had a better name scheme... i mean whats wrong with GTX 1055 or 1050x, least then less tech savvy customers can understand it a bit better. Though i guess the majority of us here this wont be an issue. The bandwidth on this card is a joke though, Then again this late into a cycle people unless they really need one this second shouldn't be buying any GPU's with rumours of new ones just around the corner
https://forums.guru3d.com/data/avatars/m/88/88351.jpg
waltc3:

The 96-bit wide bus cripples the product. Even a 128-bit bus GPU is dog slow compared to a proper 256-bit bus GPU, etc., even if everything else is equal.
Yeh, I don't get it, why does the bus keep going down? Does it really shave off a lot of the cost or something?? I remember back in the day 256 was scoffed at compared to 320 or whatever, then 128 became the norm. lol So now 96... I guess even lower soon. ha
https://forums.guru3d.com/data/avatars/m/34/34585.jpg
I disagree these are worse chips, since they have the same CUDA count as the 1050Ti and have higher clock speeds than it to boot, this is down to the memory where there is less memory chips and due to that you can have less layered PCB's which will cut down the costs quite a bit.
https://forums.guru3d.com/data/avatars/m/267/267641.jpg
Well. this card will perform well on paper for people who dont fully understand the specs.. I rember me "great" producs like Riva TNT 2 Vanta..
data/avatar/default/avatar13.webp
Another scam in such a short period of time, with 30gb less bandwidth it is gonna perform worse in a lot of games. Good job on using pc gamers on a budget as trash cans for broken chips.
https://forums.guru3d.com/data/avatars/m/88/88775.jpg
Great, so an underpowerd GPU now available in an even weaker form??
https://forums.guru3d.com/data/avatars/m/175/175902.jpg
allesclar:

Great, so an underpowerd GPU now available in an even weaker form??
Not so underpowered... the GTX1050Ti once OC is just below GTX1060 3g in most case, not bad for a 100Euro less GPU, and for the price you have 1 more gig. Also the new GTX1050 3g is intended for casual gaming and entry level, i think it's a good move as the original GTX1050 is weaker than the GTX950 that it's intended to replace. And stop crying about the memory interface: - again it's entry level 64, 96, 128... it's more or less the same with those GPU. - the bit is the "size" of the road the frequency is the "autorised speed", Why should NVidia would ever put a 4096 wide bus (as an exemple, there is no HBM GTX1050 of course) on such a GPU?
https://forums.guru3d.com/data/avatars/m/196/196284.jpg
schmidtbag:

Yeah I don't really get it either. When AMD does this, sure, it's annoying and stupid, but it's usually for OEMs. This isn't the first time Nvidia has done this. For example, I remember there were 3 different flavors of the GT 630, each of which were very different from each other (in fact, 2 were Fermi while one was Kepler based) and to my knowledge, none of them were specific to OEMs. It seems for anything below ##50Ti, Nvidia just recycles the performance tiers in whatever way conveniently gets rid of binned parts. I'm not sure if I prefer this, or AMD's method of "let's rebrand these same binned chips for the next 4 years".
They did the same thing with the GT640....