• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Unigine Heaven 4 benchmark

I left that locked because I'm not certain of safe values. The power limit slider will of course affect the voltage, but it's much more constrained. I'm not really tempted to push to the limits on it. Too new, too expensive, not enough reason. Gaming at 1920x1080, I'd be moving from far over vsync to slightly further over vysnc and wouldn't even notice. Any overclocking is only for a benchmark score and that's nowhere near enough motive to be unlocking voltages and trying my luck.

The limit is locked by the GPU so all you are doing is allowing it 100% of the max value. There is no extra mv options on pascal - if you don’t set it at that your 120% power has no effect really.

It’s safe bud, nvidia hard limited it which is one of the more annoying things on this chipset and the reason AIB cards don’t gain a lot on pascal.
 
The limit is locked by the GPU so all you are doing is allowing it 100% of the max value. There is no extra mv options on pascal - if you don’t set it at that your 120% power has no effect really.

It’s safe bud, nvidia hard limited it which is one of the more annoying things on this chipset and the reason AIB cards don’t gain a lot on pascal.

Well, I cautiously gave it a bit of a go with a bit higher voltage. Not a success. At the clock speeds I used for my benchmark results the card was being power throttled at times even with 133% max power draw, so increasing voltages just increased power draw and thus throttling. There's no point going over the 1.062V default and it doesn't seem possible to undervolt it, at least not with Afterburner. 0 on the GPU voltage is the default and there isn't a negative setting.

I had a poke around for some power consumption results and found what I expected, e.g.

http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1070-ti-8gb,5311-16.html
http://www.tomshardware.com/reviews/nvidia-geforce-gtx-1070-ti-8gb,5311-16.html
Their max stable overclocked MSI 1070Ti Gaming drew 233W. 133% power limit is 180x1.33 = 239W. Mine was clocked higher than theirs, so it would be drawing more power. Mine is much better cooled than theirs (I have very good case cooling, so the heat moved off the graphics card is shoved straight out the back of the case) and so I get much lower fan speeds and that would reduce power draw somewhat, but not enough to more than offset the increased power draw from the higher clocks.

It seems that I got lucky with the card and I have my case and cooling set up well, so I'm not being bottlenecked by either temperature or voltage. I'm being bottlenecked by total power draw, which is banging up against the 239W limiter and causing throttling. Afterburner monitoring confirms this - power throttling occurs frequently.

To get a higher result I'd need a higher maximum power draw, which as far as I know would require a different BIOS on the card and a different PCB. The ratio of power taken from the slot and the cables might have to be changed (depending on how it's handled at the moment) and an additional power cable would be required to stay in spec. Actually, it's already a little over spec at 239W (spec would be 225W). The power consumption monitoring in that review shows that nvidia/MSI have done it right - any excess is drawn from the cable, not the motherboard.

The only thing I could theoretically do would be to cool the card in a different way, one that's powered seperately. That would reduce the power draw of the card by whatever amount it takes to power the fans, giving a little more power for the card itself. But that would be silly for the small gain.

I'm going to count that as my final score and return to stock settings.
 
So...how many did you buy? :)

EDIT: That's a strangely low minimum FPS. I'm curious about it - do you have any thoughts as to why? I always get a momentary huge dip at the same point in the benchmark and I'm guessing everyone does, but yours is really low.
 
Last edited:
So...how many did you buy? :)

EDIT: That's a strangely low minimum FPS. I'm curious about it - do you have any thoughts as to why? I always get a momentary huge dip at the same point in the benchmark and I'm guessing everyone does, but yours is really low.

I got 2 before I found out there was no mGPU support.

As to the dips in the bench this is something that everyone gets. I could have probably reduced them a bit in the run I did above but it was my first ever go with the card.:)
 
For the GPU clocks do you want the OC speed or the boosted speed it hits during the bench?

Sorry new to this and just got a 1070 and applied an OC via GPU Tweak.

Is there an easy way to read the numbers?
 
I think I did this right.

14-12-17 OC Heaven Bench 01.png


1070 @2062/4666
Ryzen 1600 @ stock
1080p
388.59 Drivers
 
So...how many did you buy? :)

EDIT: That's a strangely low minimum FPS. I'm curious about it - do you have any thoughts as to why? I always get a momentary huge dip at the same point in the benchmark and I'm guessing everyone does, but yours is really low.

Minimums sorted, I overclocked the memory some more.

Titan V @1980/1000

7980XE @4.8

388.59 Drivers

1080p

sgDPxFH.jpg
 
My minimum looks very bad. I noticed during the bench it stuttered at certain points. Any idea could be causing it?

Looks to be the same as yours @Kaapstad - did you sort it wit some more memory OC?

I did get an improvement by overclocking the memory. Heaven 4 will almost always give low minimums and is to be expected and nothing to worry about.
 
Minimums sorted, I overclocked the memory some more.

Titan V @1980/1000

7980XE @4.8

388.59 Drivers

1080p

Decent score, does seem to add a reasonable bump over the xp but looks to be only 15% which is ok but not crazy. I don’t know why they have a 2 card limit on purchase if there is not real means of linking outside of specific applications.
 
Decent score, does seem to add a reasonable bump over the xp but looks to be only 15% which is ok but not crazy. I don’t know why they have a 2 card limit on purchase if there is not real means of linking outside of specific applications.

I can think of several reasons. Maybe it's one of these, maybe a combination, maybe something else, but here are my suggestions:

1) They're making very few of them and want to ensure that they don't run out. It looks bad if potential customers can't buy a card that has a lot of attention.
2) They're making very few of them and don't want to have people buying them to resell at a higher price because they can't be bought new without a long delay.
3) They're creating an artificially limited supply to make the extremely high price appear more justified.
4) The cards are a viable choice for some professional work and they don't want large customers buying a bunch of them instead of more expensive professional cards.

My guess would be mostly (4). The Titan V isn't cut down very much from the Tesla V100 flagship professional card but it's a fraction of the price. You don't get the same drivers and that does matter for pro work, but I think there are scenarios where the huge cost saving would be considered worth the reduced specs (but not much reduced) and the different drivers. It might also be possible to use the pro drivers on the Titan V.
 
I can think of several reasons. Maybe it's one of these, maybe a combination, maybe something else, but here are my suggestions:

1) They're making very few of them and want to ensure that they don't run out. It looks bad if potential customers can't buy a card that has a lot of attention.
2) They're making very few of them and don't want to have people buying them to resell at a higher price because they can't be bought new without a long delay.
3) They're creating an artificially limited supply to make the extremely high price appear more justified.
4) The cards are a viable choice for some professional work and they don't want large customers buying a bunch of them instead of more expensive professional cards.

My guess would be mostly (4). The Titan V isn't cut down very much from the Tesla V100 flagship professional card but it's a fraction of the price. You don't get the same drivers and that does matter for pro work, but I think there are scenarios where the huge cost saving would be considered worth the reduced specs (but not much reduced) and the different drivers. It might also be possible to use the pro drivers on the Titan V.

My guess is a combination of 1,2 and 4.

Bearing in mind NVidia have not cut anything from the actual GPU they are probably undercharging for the card as it is still a monster professional card.

I can also see these cards having a very high resale value if the supply is short like it is now, NVidia don't have anything to ship until 30 Dec.
 
Back
Top Bottom