Which 970 model to choose?

[Dutu40]

Hello!

I'm looking to upgrade from my current gtx 560 ti to a new 970 model (and possibly keep the 560 ti as a PPU).

Price-wise the cards are pretty much the same, with around 20-30 euros difference in between.

I'd like to know your opinion on which card would be the best, I'm looking for the meanest one of the bunch, since they're all very good I know the difference is minimal but still I'm not looking to upgrade afterwards for a good few years and want to get the very best out of them.

I can choose between the MSI GTX 970 Gaming, Asus 970 STRIX, Zotac 970 Omega Core, Gainward 970 Phoenix or Phantom and Gigabyte GTX 970 Xtreme Gaming edition and G1 Gaming edition.

My PSU is a Corsair CX750 , I'm looking more towards the gigabyte Xtreme gaming one but would like to know all opinions and if my PSU can actually support the GPUs, including the 560 as a PPU (although that's not mandatory).

My Motherboard is an AsRock Z87 Extreme3.

Thanks in advance for your time reading this and for your answers if you choose to help me!

 

[Dark NighT]

I would pick the msi, just because it is so quiet, i love the silence.

 

[AlienALX]

Don't waste your time or electricity using the old card as a PPU.

The 970's Physx chip will decimate it.

 

[Kilbane]

Don't waste your time or electricity using the old card as a PPU.

The 970's Physx chip will decimate it.

Physx chip? Did I miss something?

Physx uses CUDA to compute physx. Using a seperate card as a PPU is great where games are Physx heavy (take the Batman Arkham series). The only downside is the increased power draw and heat. I have seen quite a lot of people using the 750 and 750ti as PPU's and they see a performance gain in such games. The 560ti is more than adequate as such, however will draw more power than your 970 (I have had both of these cards).

 

[Kushiro]

In regards to OP, the Gigabyte Xtreme gaming would get my vote.

 

[Kilbane]

Sorry got sidetracked, as for the OP, I have the MSI 970 Gaming 4G. Can't fault it. I don't think there is a huge difference between any of them. Pick one that suits your build.

 

[AlienALX]

Physx chip? Did I miss something?

Physx uses CUDA to compute physx. Using a seperate card as a PPU is great where games are Physx heavy (take the Batman Arkham series). The only downside is the increased power draw and heat. I have seen quite a lot of people using the 750 and 750ti as PPU's and they see a performance gain in such games. The 560ti is more than adequate as such, however will draw more power than your 970 (I have had both of these cards).

Nvidia have integrated the Physx PPU onto their die. So as you say the more CUDA cores and the higher the clock etc the better it runs.

The 560ti will not compute Physx anywhere near as fast as a 970 and would actually be a downgrade. I tried numerous other GPUs with my Titan Blacks including a 770 and all of them were much worse.

However, given you have both of those cards why not run Fluidmark and post the results? One from the 970 and one using the 560ti.

 

[Kushiro]

Personally I've never seen much point in having a weaker card for PhysX. For me if I am going to have another card dumping heat and consume power, I would much rather just grab a second GPU for dual.

Then again if one doesn't have the budget or has their own reason, I see no reason to judge them for it.

 

[Kilbane]

Nvidia have integrated the Physx PPU onto their die. So as you say the more CUDA cores and the higher the clock etc the better it runs.

The 560ti will not compute Physx anywhere near as fast as a 970 and would actually be a downgrade. I tried numerous other GPUs with my Titan Blacks including a 770 and all of them were much worse.

However, given you have both of those cards why not run Fluidmark and post the results? One from the 970 and one using the 560ti.

Dude, I think your missing the point. Offloading the Physx operations to a secondary stand alone card will reduce the load on the main GPU allowing higher frame rates due to the cores dealing solely with the rendering operations rather than having to do Physx also.
Of course Fluidmark runs better on a card with more cores, it's testing Physx and not rendering.

I'll just leave THIS here for you to have a look at.

 

[AlienALX]

Dude, I think your missing the point. Offloading the Physx operations to a secondary stand alone card will reduce the load on the main GPU allowing higher frame rates due to the cores dealing solely with the rendering operations rather than having to do Physx also.
Of course Fluidmark runs better on a card with more cores, it's testing Physx and not rendering.

I'll just leave THIS here for you to have a look at.

That's probably the only case I have ever seen where more performance was gained using an inferior GPU to cope with the Physx.

Not that any of it matters of course, given Physx is pretty much a dead technology.

 

[Kilbane]

:banghead: :banghead: :banghead:

 

[AlienALX]

Watch it your brains will fall out of your ears.

 

[barnsley]

Dude, I think your missing the point. Offloading the Physx operations to a secondary stand alone card will reduce the load on the main GPU allowing higher frame rates due to the cores dealing solely with the rendering operations rather than having to do Physx also.
Of course Fluidmark runs better on a card with more cores, it's testing Physx and not rendering.

I'll just leave THIS here for you to have a look at.

Thats from 2013. Its also using older GPUs (Those Titans are slower than 970s), a newer physX card and some pretty odd choices of games. His recent post reflects on what the current situation is.

Ok, so I did some tests.

My specs:
CPU: core i7 4770k
GPUs: Titan SLI (x2), GTX650ti for PhysX
Driver: 352.86 ("Witcher 3 ready")


I went to areas where there were a lot of PhysX enabled objects (flags, banners, hanging herbs, etc.) and tested the framerate both while standing still, and while using the Aard sign to fling everything about. I then averaged the results below.

Without dedicated PhysX card: 50.1
With dedicated PhysX card: 50.8

Yup. Not very impressive at all. I'll add that I don't think the 0.7fps is mere statistical noise. I really did notice a ~1 fps difference in most of the areas I tested in, especially when throwing Aard. But 1fps is 1fps, so who cares, right?

Taken from his article "Does a dedicated PhysX card help in Witcher 3?"


There has been no need to run a dedicated PhysX card for a long time . It does not impact your framerate nearly as much as you seem to think (and you can see that by disabling it through settings if you have an nVidia card). Currently, I'd argue all it does is give a few older games a bit of a boost but at the expense of more driver problems. IIRC nVidia were making noise about dropping support for a dedicated physX card not too long ago anyway.

 

[Kilbane]

Thats from 2013. Its also using older GPUs (Those Titans are slower than 970s), a newer physX card and some pretty odd choices of games. His recent post reflects on what the current situation is.


Taken from his article "Does a dedicated PhysX card help in Witcher 3?"


There has been no need to run a dedicated PhysX card for a long time . It does not impact your framerate nearly as much as you seem to think (and you can see that by disabling it through settings if you have an nVidia card). Currently, I'd argue all it does is give a few older games a bit of a boost but at the expense of more driver problems. IIRC nVidia were making noise about dropping support for a dedicated physX card not too long ago anyway.

An unfair example as to how modern games use Physx. The Witcher 3 has its Physx operations locked to CPU so will not take advantage of a stand alone card. However this does go to show how the Physx are being handled in recent releases. Hairworks for example will not use a Physx card as it must run on the main GPU. So yes having a stand alone card is something that isn't strictly needed. But as my case was about a stand alone card not negatively impacting performance and will in fact help in games that rely on Physx such as the arkham series, I feel my advice was good.

 

[AlienALX]

Whenever I have tried it (and I have, many times) it always performed worse than just running it from the faster card. And that wasn't just one test it was plenty of times (pretty much every time I have upgraded GPUs).

The last time I tried it I used two Titan Blacks with one GTX 770 sandwiched in the middle. Performance in Physx games tanked by about 15% and Fluidmark the 770 performed around 50% of one Titan Black.

Im summation it's not worth it. Not by any means neither fair nor foul.

I'm not surprised Nvidia are dropping support for it either. They have probably realised that once the truth went out that people were not going to fork out extra money/pay for extra electricity when it's such an epic waste of time.

The amount of times I have seen people do it on the EVGA forums (mostly yanks desperate to spend as much money as they can) it has always worsened performance.

So there's only one instance where some one actually said that it performs better, and that's the article you posted.

Maybe if I hadn't tried it myself and seen it for myself I would maybe believe that article. But I have, and so I don't.

But as my case was about a stand alone card not negatively impacting performance and will in fact help in games that rely on Physx such as the arkham series, I feel my advice was good.

The very first game I tried it with was Batman. Mafia 2 took the biggest hit though.

 

[Pr3d4t0r]

msi or gigabyte

 

[Kilbane]

I would go MSI personally

 

[AlienALX]

Gigabyte. If you want to see why check out the MSI forum and look at the nightmares I had with their RMA.

That or EVGA, who are setting up shop in the UK very soon.

 

[Doomslayer]

EVGA has come out with a GTX 970 Hybrid might be worth a look

 

[garih]

i went for the msi card last week. seems pretty decent, no sag at all, very quiet whilst at full load. nice msi white led logo on front. no backplate tho.