GTX 970 upgrade to GTX 1070 (based on old 2500k for 1080p gaming)

STRIKER_RU

Honorable
Apr 17, 2015
54
0
10,640
Hi everyone! I've recently (7 months ago) got a GTX 970 Asus STRIX instead of my GTX 770 (GIGABYTE 2GB). So my question basically is should I sell my GTX 970 now and get GTX 1070 in June of 2016? My biggest concern is that I still run core i5 2500k slightly overclocked to 4.2 Ghz and I don't have plans to upgrade to skylake or other CPU and MOBO till 2017 (summer). Also I should mention here that I run all my games with 1080p.

Thank you in advance!
 
Solution


I run 970 and a 2500k at 4.2 Ghz also haha. I considered getting a new 1070 in June as well but decided that I'll wait. This setup if fine for 1080p. 2017 is probably a good time to upgrade and build a new computer entirely. You'll have more information on different versions of the...


Provided that its 1080p 60hz it's already overkill, but then again with the unreal 4 engine pushing newer games to their limits you should be good with 1080p 144hz. Btw Nvidia released the specs and the 1070 is going to be around the same performance or a bit faster than a titan X supposedly.
 


It appears you don't know much about the new graphics cards.....
It is a jump from 28nm to 16nm, allowing them to they put nearly 4X the amount of processing power in the same die area. Also given that the clock speeds released for the 1080 were in the high 1000s, with it being overclocked to 2114mhz at the live demo during yesterday's conference, this is not a surprise.
Combined with the same amount of memory on a new architecture this is pretty standard.
If you're confused, and asking yourself why the 1070 would perform so closely to the 1080 at such a big price gap, this is almost 100% because the 1080 will be an OC focused card, which was already displayed with the founders edition release with most likely binned chips.
 

STRIKER_RU

Honorable
Apr 17, 2015
54
0
10,640

I am concerned of a significant price drop on GTX 970 in June.
 


The previous three generations were all based on the 28nm process, meaning they were all eventually going to hit a brick wall in how far the performance would go.
16nm is a large jump, meaning that they can, as I said put more 'stuff' into more places.
No need to be so aggressive, its simply a fact that companies strive to create better products, and given that this has been in the works for two years now and the significant improvements that have been made in the hardware industry it is no surprise.
Typically thats what a new architecture does; it introduces new and better products.
Its like saying that you would buy a GTX 770 over a GTX 960, or a GTX 780ti over a 970. Its just silly right? Its the same performance, but just an older chip.
Going back to what I said, this is an example of them all being based on the same process, meaning similar yields.
The fact of the matter is that pascal is simply a larger jump in technology, and a 'new king' as nvidia put it during their conference, was in order.
They said that the 970 was supposedly better than the titan x, although they did not state in what applications, pointing towards VR hand picked benchmarks, meaning that it is probably a little slower, which is still a great upgrade.
However given the specs shown here: http://wccftech.com/nvidia-geforce-gtx-1070-launch/
It looks to be similar performance to the titan, especially with the speed per cuda and clock increase.
The price is due to them using standard GDDR5 memory which has been in production for years and is on the way out for mid range to high end cards, being replaced with GDDR5X and HBM and HBM2 memory.
This allows them to cut costs while keeping performance very high by using this already well established memory, given that it still is the same memory used on the titan x, albeit 2GB more of it.
Use the GTX 1080 as an example, the 980 and 980ti are now obsolete because of the direct replacement and odd 30% increase over the 980ti/titan.
If you want answers as to why ask Nvidia not me.
 


Performance is not inclusive of voltage.
Performance is performance, voltage is indicative of power consumption.
Its like saying this GT730 draws 30w from the wall, its performance must be great, it just doesn't make sense.
 


So all in all, keep your current rig you you're only running 1080p 60hz, but if you want a better experience when the new cards come out, get a high performance monitor and the GTX 1070.
 

anti-duck

Honorable
Something else that Suzuki doesn't seem to get is that 14/16nm FF is cheaper than 28nm as you get more chips from the same silicone wafer and once they perfect the process and get higher yields, it'll just get cheaper. It isn't a fair comparison to say that because GPU A with a 28nm process sells for $1000, it is de facto going to be ahead of GPU B manufactured using a 16nm FF process selling for $379 (we'll just completely ignore that certain aftermarket 980 Ti's beat a Titan X in everything) 14nm is more expensive right now because the yield will be around 70%. The 1070 launch price of $379 isn't that much higher than the $329 970 either.
 


No.... this is not true...
A smaller process means that more wafers are required, and the yields become smaller due to higher densities.
This means that it gets more expensive. Read up on Moore's law, which is quickly becoming obsolete.
The 1070 launch price is nearly 400 dollars, which is up from its predecessor, but it also appears that Nvidia have chosen to go for the price bracket with the most value with the current process in the 1070.
 

anti-duck

Honorable
The act of shrinking a die is to create a somewhat identical circuitry using a more advanced fabrication process, usually involving an advance of lithographic node. This reduces overall costs of a chip company, as the absence of major architectural changes to the processor lowers research and development costs, while at the same time allowing more processor dies to be manufactured on the same piece of silicon wafer, resulting in less cost per product sold

--https://en.wikipedia.org/wiki/Die_shrink

A smaller fab means you get more chips from a wafer, being a newer fab will mean the yield is lower until the process matures.



Not really sure why you said that either, when I gave the launch prices of both the 1070 and the 970 in my previous reply.
 

Correct, and as a result of this lower yield, it is more expensive no?
T.T Logic doesn't seem to apply apparently.
 

anti-duck

Honorable
I never said that more wafer aren't required lol, I said that you get more chips from a wafer but the yields are lower and it is overall cheaper as the yields go up, which they will. I even explicitly said " 14nm is more expensive right now because the yield will be around 70%.", which was admittedly a guess, it's a long time since I've seen what an average yield from a new process is.
 


Because I justified that in the post if you actually read it.
 


That is as the process matures over the next few years, this is relative to the new graphics cards being released.
I'm not even going to bother arguing anymore, ending this thread here. Seeya
 

anti-duck

Honorable


What the hell lol? I agree with you that Suzuki doesn't really know what he/she is talking about, you pick on some trivial little point and then storm off when I correct you? Is Tom's Hardware a fucking playground these days?
 

Stirfried

Distinguished
May 27, 2011
2
0
18,520


I run 970 and a 2500k at 4.2 Ghz also haha. I considered getting a new 1070 in June as well but decided that I'll wait. This setup if fine for 1080p. 2017 is probably a good time to upgrade and build a new computer entirely. You'll have more information on different versions of the 1070 and its benchmarks. Better to know more about what you're buying than worry on the potential loss on your current card. The only reason to get the new card is if you're gonna be using VR or you need it to power a 1440p/144hz monitor(s). No rush to get the new card its not going anywhere.
 
Solution