• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

GTX 1070 specs and the reason that nvidia didn't unveil it!

idx

Joined
Apr 2, 2009
Messages
98 (0.02/day)
It appears that the GTX 1070 is even funnier than its name. Nvidia did not unveil the full spec of its GTX1070 because ... lol guess what:

GTX 1070 has 1920 SP !
Clocks : 1530MHz, boost 1670MHz .
Ingame it can boost upto 1800Mhz

I just wanted to share this because Its so freaking funny that both of Nvidia cards has 1080p in its name. and believe it or not this is why they didn't want to unveil everything about the GTX1070.

Just to be clear this is not 100% sure!

*source: ... umm... I promised to not tell.
 
Like the 970 before it with 0.5MB of crippled slow RAM out of 4, I bet this has 1GB crippled out of 8. Can we say memorygate all over again? lol

I'll bet this is the real reason.
 
Like the 970 before it with 0.5MB of crippled slow RAM out of 4, I bet this has 1GB crippled out of 8. Can we say memorygate all over again? lol

I'll bet this is the real reason.

Apparently the GTX 1070 is nothing but a damaged 1080 chip.
 
I don't mind 1GB slow memory out of 8GB total :p
 
I don't get it. So what? It has better than 980 performance for $100 less MSRP. Edit: Hmm, just checked 980 prices. $400-425 is typical now. So, better than 980 for a little less. Still don't see a problem.
 
Last edited:
nvidia has been doing the memory split thing since the 580 ...
the issue with the 970 is that no vendor used large enough memory chips to offset the issue
and its fairly easy to program around as well you simply need to allocate low priority objects to the .5GB section first and load them into the faster 3.5GB partition when needed
then you simply just swap back and forth
 
I don't get it. So what? It has better than 980 performance for $100 less MSRP. Edit: Hmm, just checked 980 prices. $400-425 is typical now. So, better than 980 for a little less. Still don't see a problem.

Unless you already own GTX 980...
 
and its fairly easy to program around as well you simply need to allocate low priority objects to the .5GB section first and load them into the faster 3.5GB partition when needed
then you simply just swap back and forth
Sounds like a recipe for hitching.
 
It appears that the GTX 1070 is even funnier than its name. Nvidia did not unveil the full spec of its GTX1070 because ... lol guess what:

GTX 1070 has 1920 SP !
Clocks : 1530MHz, boost 1670MHz .
Ingame it can boost upto 1800Mhz

I just wanted to share this because Its so freaking funny that both of Nvidia cards has 1080p in its name. and believe it or not this is why they didn't want to unveil everything about the GTX1070.

Just to be clear this is not 100% sure!

*source: ... umm... I promised to not tell.
I can't say I buy this... nor care I have to admit...LOL



On a side note, can someone explain why that means they will have a slower chunk of memory please....do we know the SMM breakdown to make that call?
 
On a side note, can someone explain why that means they will have a slower chunk of memory please....do we know the SMM breakdown to make that call?
If you're referring to my post above, then I'm only surmising it based on the previous 970 and the fact that NVIDIA is being cagey about the specs for the new version, that's all.
 
On a side note, can someone explain why that means they will have a slower chunk of memory please....do we know the SMM breakdown to make that call?
It's just someones guess... My i guess is fully functioning 8GB of VRAM. Nvidia can't fukup twice in a row.

PS. If it's at least 40% faster than my GTX970 - i buy it, no matter what.
 
Can't say I would make that leap with you qubit...at all. Heck of leap...

Just need to know how the SMMs are setup to figure it out though. Do we know what the 1080 SMM setup is? You can have a cut down core and still full access to the memory, just with less bandwidth (bits, say 256 to 192)... that is, if I have an understanding of it all, LOL!
 
nvidia has been doing the memory split thing since the 580 ...
the issue with the 970 is that no vendor used large enough memory chips to offset the issue
and its fairly easy to program around as well you simply need to allocate low priority objects to the .5GB section first and load them into the faster 3.5GB partition when needed
then you simply just swap back and forth
Code specifically to make a handful of cards not run at a snails pace is bad practice. If NVIDIA does the same thing with GTX 1070...
latest
 
let's pray hard that the 1070 will not suffer the same fate that the 970 has gone through... Despite the 970's 0.5GB VRAM being a reserve space, it still win 1080p benches across all kind of games being thrown at it, with some potential in 1440p.
 
And if it has 8GB, with 7 usable, that is only going to affect, what 4K users with gobs of AA? Not sure I see it as an issue... but the speculation is premature at best, particularly with the reason why it started was meritless to begin with.
 
Can't say I would make that leap with you qubit...at all. Heck of leap...

Just need to know how the SMMs are setup to figure it out though. Do we know what the 1080 SMM setup is? You can have a cut down core and still full access to the memory, just with less bandwidth (bits, say 256 to 192)... that is, if I have an understanding of it all, LOL!
Oh I dunno you could well be right. Just one of my wild ruminations that's all. :p

Alas, I don't know anything more about the specs than what's been reported in the news articles. Can't wait for the reviews on the 27th.
 
Ta... just 4 days to go. :)
 
Sounds like a recipe for hitching.
not if its done correctly
you swap from system memory all the time so why let it go to waste
 
Last edited:
Apparently the GTX 1070 is nothing but a damaged 1080 chip.

Just like the 970, and the 670, and the 570, and the 470... Did you expect something else?

nvidia has been doing the memory split thing since the 580 ...

Do you have a source for this, because before the 970, they used lower memory bus sizes, not a memory partition.
 
Do you have a source for this, because before the 970, they used lower memory bus sizes, not a memory partition.

Regrettably, at this moment, no I can't get you a source to OneMoar's claim. However, it was pretty common knowledge that the 660 was partitioned into 1.5GB memory portions.

This is why the 3GB version was preferred, despite nothing using 3GB back then, because the memory was symmetrical.

On the 2GB versions though (with a 1.5GB segment and a .5GB segment), no one really noticed, because games were not using all that VRAM anyway.
 
Last edited:
The upshot here is that while this is the first time NVIDIA has used this specific ROP/MC configuration in a product, this is not the first product they have designed with segmented or otherwise unbalanced memory configurations. Since the GTX 500 series, on some midrange SKUs NVIDIA has used unbalanced/asymmetrical memory configurations, most recently on the GTX 660 and GTX 660 Ti. In the case of both of those cards, NVIDIA utilized a 192-bit memory bus with 2GB of VRAM attached, which meant that some memory controllers had more VRAM attached to them than others. The end result as it turns out is very similar, and while NVIDIA has never explained in-depth how they handle memory allocation on those cards, it turns out that it’s very similar to GTX 970’s memory segmentation. Which is to say that NVIDIA actually has multiple generations of experience with segmented memory, and this is not the first time they have implemented it. Rather this is first time we’ve seen such a configuration on a high-performance card such as the GTX 970
cite
http://www.anandtech.com/show/8935/...cting-the-specs-exploring-memory-allocation/2
toward the bottom
 
The split they did in 660 and 660ti was almost symmetrical, 970 split was totally asymmetrical and first of a kind in that regard (gimped part is completely ignored by the driver until absolutely needed because it's still better than ram to vram over pci-e transfer)

edit: crossed the whole section because @rtwjunkie made me remember 600 series was first with asymmetric bus and memory, and the 550 ti was the almost symmetrical one.
 
Last edited:
Back
Top