over-clocked 860m ~ 90% 870m stock

"A overclocked 860m will be roughly 10 percent slower (using my overclock) than an 870m if the 870m does not throttle. Also a clocked 860m will still produce less heat than an 870m"

^ read the above statement in this forum, regarding the MSI GS60.
My question: True or just an exaggeration by the 860m version GS60 guy to feel better about his 860m ?

 

100% true.

 

Not sure. It depends on how overclocked it is, and it probably means only in terms of benchmarking scores, most likely at lower res. Since 870m has more resources in rops, TMUs, memory bandwidth etc etc... I really doubt that claim. Specially at higher resolutions.

A non throttled 870m is basically a 940+ core 680m right? That is similar to 780m on stock? Sadly I don't know much of the 870m.

 

It seems to be true. It's quite impressive. I hope the 980M has the same ability. That would add a good 25%-30% to the performance output.

 

GTX 870M is a 100W+ GPU
GTX 860M is a 45W GPU

I dont think an extreme overclock with 860M would ever get close to 100W. I think it will die a long time before it does

 

GeForce GTX 870M vs 860M

I think that's the maxwell version you are talking about.

 

GPUboss...ugh. They post a lot of inaccurate stuff :/

But of course a GTX 870M will win against GTX 860M. I`m not debating that. But efficiency wise: Performance/watt, the GTX 860M Maxwell runs circles around GTX 870M. Which is very helpful when you apply overclock.

 

The gs60 uses the kepler gtx860m.

 

yeah GPU boss...
But that was to show you the TDP = 75W.

 

I don't know what the hell is wrong with these companies. Why the hell they don't use Maxwell 860m? (except Lenovo & Alienware)
MSI didnt use it. Aorus didnt use it (in their defence I heard maxwell and sli dont go together) but still.....

 

This explains,

 

I've been meaning to publish these, just never took the time. I got my hands on a Razer Blade 14 2014 with GTX 870m and ran some benches... so what the heck, here goes...

Compared with GTX 860m Maxwell in a Sager NP7338 based on Clevo W230SS 13.3" notebook.

(1) GTX 860m Stock (W230SS - Maxwell)
Core: 1029MHz / 1097MHz Boost
vRAM: 5000MHz
vCore: 1.15V

(2) GTX 860m OC (Prema Mod on W230SS - Maxwell)
Core: 1309MHz / 1377MHz Boost (equates to +250MHz over stock)
vRAM: 5850MHz (equates to +425MHz over stock)
vCore: 1.20V

(3) GTX 870m (Razer Blade 2014)
Core: 941MHz
vRAM: 5000MHz
vCore: ?


ARTIFICIAL BENCHMARKS:

3DMark Fire Strike


3DMark 11 'P'


3DMark 11 'X'


Catzilla


Unigine Valley




GAME BENCHMARKS @ 1080p:

Bioshock Infinite


Crysis 3


Final Fantasy XIV




Grid 2


Metro Last Light


Resident Evil 6


Sleeping Dogs


Thief (2014)


Titanfall


Tomb Raider (2013)



TEMPERATURES:






POWER CONSUMPTION:






SIZE COMPARISON:






BLADE 14 2014 COOLING SYSTEM:

 

Wow..... awesome work bro.... (y)
But thats clevo right? I dont think the GS60 body can dissipate the heat well enough to keep temperatures that low and secondly its a keplar so.....
Also blade kept the temps to 80s.... The GS60 shoots the temps to 93C straight away (870m version)....
Desperately waiting for the unveil....

 

But the question was well-answered by this comparison....
OCed maxwell 860m does reach the 90% 870m mark often enough, in terms of fps

 

Those overclocks are massive. No wonder it reaches similar performance.

 

Well done HTWingNut. Good overview

That is a GTX 860M running inside a 13" and still staying in the 70s despite running a massive overclock, ladies and gentlemen.
Can`t wait until Maxwell GM204 is here

 

Keep in mind that I ended up repasting the Blade CPU and GPU. Before the repaste the 870m was in the mid 90's. The repaste helped that much.

Can't be soon enough! But we've been speculating about a year now haven't we? lol.

 

Those are some really impressive temps you got there with the Blade, considering it's 870M in a 0.7" thick chassis. Well done!

 

I just added some pics at the bottom. You can see the pess poor job Razer does with the cooling touching the heatpipes directly to the CPU and GPU dies. Good in theory, but there's gaps between the heatpipes and no contact on the corners of the CPU. It could probably be a bit better if they soldered the sinks together and had a flat surface contacting the entire die.

Razer cooling does get loud though, holy cow. It's silent at idle but fire up a game and man you'll look like Psy from Gangnam Style video:

 

Wow... those pics... Wow...

 

That's unbelievable...

 

Those gaps... dafuq was Razer thinking.

But see, this means that if Razer designed a proper cooling system, it's very likely the components won't overheat at all. 870M + 4702HQ all in a 0.7" thick chassis and not have them overheat, now THAT would've been really impressive.

 

Direct contact heatpipes always have gaps but the benefits of it outweigh that penalty. See desktop versions of that sort of heatsink.

 

I think I know which ones you're talking about, but in those desktop heatsinks the heatpipes are soldered into a special base that effectively fills the gaps, plus the part that comes into contact with the CPU is machined flat.

Like this:

 

I have a Hyper 212 EVO in my desktop. No gaps; it looks nothing like Razer's abomination. See above pic.

 

Exactly. I've seen that and used that type of heatsink before. That's what they needed to do. Instead they just use the heatsink as is with no lapping whatsoever and keep the gaps clear. Plus a desktop has the IHS which helps distribute the heat, the laptop chips have the actual die exposed and hot spots are not a good thing.

Plus there's all those pits and imperfections in the heatpipe that just contribute to a worse cooling condition. Overall I was impressed with the Razer, but it did get hot, like 55-60C surface temps at the face of the laptop by the LCD, and the right hand key area was 45-50C at times as well. I think they would have been better off asking nVidia for a special clocked 860m (i.e. 865m) with a factory OC of 1200MHz with 6000MHz vRAM. It would have run 5-10C cooler and had comparable performance and also less power draw.

 

Added power consumption to the benchmarks, near bottom...

 

Still curious how it can be that a higher clocked gpu delivers even worse min fps than the exactly same gpu with lower clocks (Thief)...

 

Wow...that puts things into prespective (heat wise).
So the heat won't be that bad even when overclocked that heavily. That's cool I think.

Wingnut, did you also try OCing the 870m on any laptop and matched the results with the stock 880m?

 

Sometimes there's anomalies. That min FPS is the bottom 0.5% min FPS. I usually use FRAPS to record FPS during benchmarking and then take the bottom 0.5% because sometimes there's spikes for whatever reason. I can re-run the bench and see if I get the same result.

Nope, no other 870m I was able to get my hands on.

 

Thanks HTWingNut, but I already observed this strange behaviour by myself, too.
Especially Unigine Valley seems to be very very inconsistent regarding min FPS. I tested with a Maxwell 860m and i get quite different min fps values nearly every time with a variation of up to 20%... The avg fps although are quite consistent with only a expected very small variation (far under 5%).

For me that's very surprising as that is a synthetical benchmark, which is (or at least should be??) to 100% absolutely identical in every benchmark run, but it's quite strange to get such big variations on min fps values. Although i observed that the min fps seem to be achieved not in the acutal benchmark scenes, but in these little intervals between two scenes (when one scenes fades out, and the next one is being loaded the framerate drops quite heavily). But that doesn't really explain these big differences as these loading times respective the whole bechmark itself is absolutely identical in every run, isn't it?? So technically it's not really understandable to me, where these differences can come from...

 

Anomalies in FPS variations I guess

 

The best part... you picked 10 recent games with both average and minimum frame rates... with the temps.
The truth is... the OC'd 860m scored BETTER than within 10% average in those games in both average frame rates and minimums and stayed significantly cooler despite the overclock.

game benchmarks 1080p Ultra/Very High
(% performance of 870m - Blade 870m)

Average FPS (average % from all ten games)
860m stock 78.2%
860m OC 92.8%

Minimum FPS (average % from all ten games)
860m stock 79.9%
860m OC 91.6%

Great job HT!

 

Thanks for doing the math... I didn't want to, lol.

 

I'm still thinking how about OCing the 870m?

 

Sure, give it a go and let us know.

 

Nah I'm leaving that to you


I'm getting the p34g