Let's Talk about Fairly Testing GPU Performance R.E. Warming Up GPUs

[Araxie]

every card behave different, every case behave different, I know for sure that every blower type cooler works way better and cooler inside my case than any other open-bench in the entire world.. so if they decide to use a crap case with crap airflow that for sure will mean a bad thing for the card.. it's just related to what case they used, at what ambient temperature and what kind of airflow had that cooler....

 

[Dayaks]

you shouldn't have to do any of this with a card to get it to where they claim it can go, especially a hard mod.

Dayaks, they used a 100% fan profile and still on a regular use they achieved on average 1785 MHz, which is nice, but quite a bit different than the 2100 they tried to pass off with the cherry picked Doom test of the original presentation.

I would think that if anyone wants a 1080 they should seriously consider waiting a couple weeks for non FE versions at the least.

100% fan and it throttled? Seems strange since every other site had 60-70C at 100% and OC'd?

You're right though, you shouldn't have to hard mod / BIOs mod for 2k+ Ghz given nVidia's presentation. On the flip side hitting the power limit is nothing new. Only one or two cards have one you can't hit without modding... And by the time they come out (kingpin?) you might as well wait for a 1080ti.

Seems like in most cases tweaking the fan profile slightly more aggressive gets you there (2k+). This is a blower, shouldn't be hard to not recirculate air...

I'll have to look at that German site when I get on a PC and can translate it.

 

[CSI_PC]

Dayaks, they used a 100% fan profile and still on a regular use they achieved on average 1785 MHz, which is nice, but quite a bit different than the 2100 they tried to pass off with the cherry picked Doom test of the original presentation.

I would think that if anyone wants a 1080 they should seriously consider waiting a couple weeks for non FE versions at the least.

That depends upon the reviewer regarding 100% fan.
The other sites I have mentioned do much better.
Tom's seems to have the worst example going with defaults but when they OC they managed to average above 2000MHz with 100% fan.
I still do not think the computerbase example is 100% fan but just temperature and power targets max'd when they average 1785MHz.
Another example with 100% fan is here HardOCP, and they also averaged above 2000MHz.
Plenty of other examples with 100% fan showing good OC after warmup.
That is not saying one should need 100% fan, just simple way to exact most performance.
However maybe some games have a strange behaviour, but that needs further investigating to see if those games behaved the same for all reviewers when trying to OC.

Cheers

 

[Revdarian]

For the two that quoted me, how many of those other sites that didn't find throttling tested in an actual enclosed case instead of a bench setup? that seems to have been the difference that forced the cards to throttle, and why it is smart to wait for different cooler versions even if you really really want a 1080.

 

[CSI_PC]

For the two that quoted me, how many of those other sites that didn't find throttling tested in an actual enclosed case instead of a bench setup? that seems to have been the difference that forced the cards to throttle, and why it is smart to wait for different cooler versions even if you really really want a 1080.

All the ones I am thinking were cased.
Throttling is a complex situation, but the simplest solutions is that NVIDIA were not aggressive enough with the fans (dB is lower than a 970/980ti), so while it is a pain for some they should look to custom profile.
The next aspect is to just ma
TBH, how many here do NOT custom profile their GPUs and various other fans including CPU....

Which reviews were you specifically talking about apart from computerbase?
Cheers

 

[CSI_PC]

For the two that quoted me, how many of those other sites that didn't find throttling tested in an actual enclosed case instead of a bench setup? that seems to have been the difference that forced the cards to throttle, and why it is smart to wait for different cooler versions even if you really really want a 1080.

This is an example of one of the better performances, although they also know what they are doing.
NVIDIA GeForce GTX 1080 Founders Edition Review & Benchmark
They made it even more of a stress test

This test was conducted over a 2-hour period (~8000 seconds). Rather than Kombustor, we used a real-world game to analyze performance and throttles in gaming scenarios. DiRT Rally was used at its maxed-out settings (4K, 8-tap MSAA, advanced blending, Ultra settings) to torture the card. To further amplify the thermal torture and create somewhat of a worst-case scenario, we also disabled all three front intake fans. This left the GPU entirely to its own devices – mostly the VRM blower fan and alloy heatsink / vapor chambers – to cool itself.

However I think part of the discussion on frequency boost comes down to various games, some seem to be very quirky such as the recent Anno games where they just use base clock unless temp&power targets.
Anyway I think a lot of the discussion is not a major issue as a lot of these can be tweaked and even resolved by the user, but I would be interested in how Anno is influencing Boost3/Pascal, just like Futuremark did as well for one review.
And this throttling issue is definitely one that will not exist for custom AIBs, due to the results we have seen when OC is done with custom settings for many reviews.
Cheers

 

[noko]

So in essence the stock 1080 cooler is good for default clocks while keeping it quiet. I rather Nvidia for the FE have a AIO water cooling system, 8pin and 6pin power connectors. Add the cost to change out the cooler on the FE makes the 1080 even more expensive. So AMD has great cooling but not much room to OC/Nvidia has great OC potential but a much poorer HSF combination. Basically the 1080FE cooler is as effective as a R9 Nano (Nano is quieter though at 100%).

 

[Revdarian]

CSI_PC that actually does show what i said, they got under ideal conditions (all case fans running max, 1080 fans running max) that it was at 1800 MHz, which like i said is very nice, but still short of what was promised on the original presentation by a fairly large margin, and they also got the 80C+ temperature.

Just wait for the other cooling solutions, even one month shouldn't kill you (general you) and if it brings way better heatsink solution then so be it!

 

[CSI_PC]

CSI_PC that actually does show what i said, they got under ideal conditions (all case fans running max, 1080 fans running max) that it was at 1800 MHz, which like i said is very nice, but still short of what was promised on the original presentation by a fairly large margin, and they also got the 80C+ temperature.

Just wait for the other cooling solutions, even one month shouldn't kill you (general you) and if it brings way better heatsink solution then so be it!

Which site you talking about?
I assume GamersNexus in my last post.
If you read the quote they stopped all fans - see section: GTX 1080 Thermals versus Frequency (And Thermal Throttling)
Again they said on the specific page I am looking at and its data (not the others)

This test was conducted over a 2-hour period (~8000 seconds). Rather than Kombustor, we used a real-world game to analyze performance and throttles in gaming scenarios. DiRT Rally was used at its maxed-out settings (4K, 8-tap MSAA, advanced blending, Ultra settings) to torture the card. To further amplify the thermal torture and create somewhat of a worst-case scenario, we also disabled all three front intake fans. This left the GPU entirely to its own devices – mostly the VRM blower fan and alloy heatsink / vapor chambers – to cool itself.

Those core clocks are very acceptable, as shown as well in the tables, so not sure how you could argue that is showing terrible throttling experience.

Cheers

 

[Revdarian]

NVIDIA GeForce GTX 1080 Founders Edition Review & Benchmark

Well their own methodology says otherwise, just saying that they make it seem like it is actually pretty darn close to testing w/o a case. /shrugs, it is still the smart thing to wait for the newer offerings.



Thermal Test Methodology

All open bench fans are configured to their maximum speed and connected straight to the PSU. This ensures minimal variance when testing, as automatically controlled fan speeds will reduce reliability of benchmarking. The CPU fan is set to use a custom fan curve that was devised in-house after a series of testing. We use a custom-built open air bench that mounts the CPU radiator out of the way of the airflow channels influencing the GPU, so the CPU heat is dumped where it will have no measurable impact on GPU temperatures.

 

[Bahanime]

Those core clocks are very acceptable, as shown as well in the tables, so not sure how you could argue that is showing terrible throttling experience.
Cheers

The red line is the air cooler when OC. The left side is clock speeds.

They noted terrible stutters in their game when it starts to throttle.

The first thermal test in a case was a stock settings just to see how the clocks behave on stock in the worse case scenario.

The latter test:

"We tested using our GPU test bench". There's more in their Thermal Test Methodology. It's an open bench setup.

DIY GTX 1080 'Hybrid' Results – Higher Stable Clock, 102% Lower Thermals

 

[Bahanime]

^ Steve @ Gamers Nexus talks more about it and what the results mean.

He claims all the sites do not test clock endurance enough to see the kind of throttling this card gets to after a long gaming session.

It's both a thermal throttling when OC as well as power limit throttling. Basically by now, people should realize, the reference card blows and it's "premium components" are nothing more than marketing speech. A premium card is one that holds it's great performance over a long duration, running cool, quiet, and maybe it can justify a $$ premium.

 

[CSI_PC]

The red line is the air cooler when OC. The left side is clock speeds.

They noted terrible stutters in their game when it starts to throttle.

The first thermal test in a case was a stock settings just to see how the clocks behave on stock in the worse case scenario.

The latter test:

"We tested using our GPU test bench". There's more in their Thermal Test Methodology. It's an open bench setup.

DIY GTX 1080 'Hybrid' Results – Higher Stable Clock, 102% Lower Thermals

That is OVERCLOCKED with Offsets and not default operation
And they are not using fixed fan speeds but variable and part of the crash is the fan speed dropping greatly.
They say for the OCing:

The above is a VRM fan RPM vs. time chart. You'll notice that the fan RPM dips hard in the middle of the endurance test – that's because the data is from the same clock-rate stability test as the above chart (derived from a +210MHz offset sustained OC). The dip in fan RPM occurs around when the clock falls off a cliff and struggles to recover, shortly after the initial throttle and clock dive.

Now, running at ~100% RPM would resolve some stability issues and reduce overall thermals – but that's really not a good solution.

They could try a fixed 75% to see if it overcame the crash in profile behaviour, and work back from that
However that is not their focus-objective and I can accept that.

Cheers

 

[CSI_PC]

He claims all the sites do not test clock endurance enough to see the kind of throttling this card gets to after a long gaming session.

All the sites I use in these discussions do warm-up the card 1st.
It takes 4mins to hit around 80 degrees, but some of those sites I linked/quoted earlier go longer before benchmarking.
Cheers