Sager NP9772 - To OC or not to OC?

So here is the deal, I'm buying the Sager NP9772 as a mobile solution to run my new DK2 Oculus Rift on. The Oculus Rift Virtual Reality headset (DK2) requires being able to run games at a consistent 75 FPS 1080p (with as high settings as possible) to avoid any judder.

I am buying the laptop from XoticPC, which offers a Redline Boost service where they overclock the GPU to maximum stable performance. However, I've heard some reports that OC'ing the GPU will result in blazing hot temps, especially where the 4790k CPU is concerned. Is this true? What kind of performance gain in terms of FPS will I be looking at on an OCed 980m? Should I have them OC the GPU or no?

Thanks

 

Why not go for an sli based notebook? It's more gpu power you need over cpu power.

 

Because VR

 

Nvidia are making a big push to get sli working on vr especially with two, one gpu for each eye display which should give near perfect scaling.

 

http://forum.notebookreview.com/thr...eculation-thread.775782/page-20#post-10004650

VR SLI trades scaling for lower latency

 

That would put it at worse scaling than regular sli which would make no sense. If that's true though and let's go for 50% scaling it will still beat out the single card setup.

 

Scaling is worse than regular AFR SLI because view-independent rendering causes duplicate work to be done on each GPU when in VR SLI

 

Work is duplicated in normal sli as well, but you have doubled your resources so that explains why you don't get more than double processing (single setup) but not why less.

But again this does not address that you still want as much resources as possible and having a second gpu is still better than one.

 

Overclocking your GPU does not make it blazing hot. In fact its surprisingly cool with the proper precautions. These are my results with +200 MHz:

Spoiler

Min fps +10-20%, avg fps +10%

Spoiler

CPU 70-80C, GPU 60-70C


This is with max fans (Fn+1), elevated on cooler. Max fans does at least 10C, undervolting the CPU another 10C. As you can see my 980M never passes 70C even overclocked.

 

Thank you for being the only person to answer my question. -_- Computer geeks seem to have a problem with staying on topic, this happened to me over at the Oculus forums as well.

Probably not going to overclock it from what you have shown me. I will not have a laptop cooler and I don't want to be running max fans all the time, and a 10% fps increase is not really worth it IMO. Thanks

 

The coolers main benefit is to keep it elevated. Mine is actually broken. OC or not I strongly suggest you prop the back up a bit to increase air flow. Use bottlecaps or whatever to raise it a few cm (this goes for all laptop models).

You dont have to use max fans for gaming. Im at 80-90C on the CPU with auto fans instead of 70-80C. Its noice vs temps.

I havent specifically checked GPU temps when overclocked and auto fans, can do that if youre interested.

Edit: Ok, I ran a test. Highest temps I get with OC 980M and max fans is in Witcher 2 with CPU 80C and GPU 69C, so I ran the same loop with auto fans. CPU 90C and GPU 79C, so +10C on each.

OC 980M, auto fans, still below 80C. Keep in mind my CPU is undervolted as far as I can, in my case -70 mV.

Unless it´s overly expensive I don´t really see a reason not to let them OC it for you. I suppose you can switch to stock clocks yourself if you want later. And if they do it, you should still have full warranty?

 

You have not doubled your resources because both GPUs are doing a lot of the same work. Each GPU/eye is rendering an offset image of the scene in stereo. But view-independent work (stuff that doesn't change from one eye to the other) like shadow mapping, GPU skinning, GPU particle simulation, etc. is duplicated on each card. In an app with absolutely no view-independent rendering, you get 2x, but since that doesn't happen in practice (there is always overlap), perfect scaling is not realistic. The proportion of view- dependent work determines scaling, the higher the better. Right now 40-50% (140-150% of single GPU) is already considered great.

 

Lets say 20% of your GPU is used for that stuff and it was NOT required for SLI. That means 100% of the other GPU could be used to render frames.

You would go to 180% performance instead of 80% which would be 125% increase in performance.

Now if both GPUs ARE required to do the same task they each use 80% of their shader resources then you go from 80% to 160% or a 100% increase in performance.

The 2nd card having to duplicate work does not explain why scaling is less than 100%.

 

Come on, Meaker. I've spent the last 2 pages repeating myself. Read Nvidia's documentation if I'm speaking gibberish to you.

 

I wouldn't waste money on all this.. I would just get it like this.. learn how to tweak your CPU+GPU... You'll need your CPU+GPU with Gelid GC Extreme etc... If you do that and flash premamod, your CPU+GPU can be overclocked quite a bit..

 

I did, but the scaling benefit is done at low latency and should not suffer the usual downsides of SLI, so while good scaling will be around 50%, that's all gravy as far as FPS goes.

Yes you can tweak a single card setup, it will not be as good as a two card setup tweaked.

 

I don't disagree with that, I'm just saying don't except 2x scaling in VR SLI like in perfectly load-balanced AFR SLI. Unless each eye is completely independent with its own command buffer, but then that's twice the CPU overhead and redundant as the two views are almost identical in stereo rendering. That's why VR SLI uses a single shared command stream to both GPUs.