Spin Off: AT's Testing Methods & Uber Mode

[3DVagabond]

When Tom's told AMD they had the retail card that was excessively throttling, they were told it was defective and to RMA it. So, any claims that you can't RMA a card because it can't maintain clocks are simply lies and smoke screens.

 

[BallaTheFeared]

If the 780 Ti came with a 100% fan speed its performance would be exactly the same because the chip can't simply keep increasing its frequency. If it also increase voltage and power consumption, NVIDIA would have to make sure the circuitry was up to the task of extra power.

So NVIDIA would have to have better components (more expensive) would have to have chips that could reach higher speeds (again higher costs on binning with lower chips reaching those speeds/voltages), etc.

How much does NVIDIA allow the power to be increased in the 780 TI? Is it something like 5-7%?

It is a contrived scenario.




Up to isn't a free pass for manufacturers, at least not in the EU.
Otherwise AMD would have said "UP to 5GHZ" instead of the "Up to 1GHz" that its uber mode pretty much hit all the time.

Only because Nvidia optimized for noise and performance, not raw performance.

Say let me add boost 3.0 on the 780 Ti and 100% fan.

Reviews show 780 Ti reference easily hitting 1250 to 1300 on reference air with +250/300 offsets, which is how you have to do it with boost 2.0. This performance is achieved with out touching the fan or voltage of the card.

Afaik the hardware of Titan and the 780 was capable of 340w consumption within specifications, they wouldn't have to change anything only increase the stock boost clock (offset for us) and the fan speed to allow it to maintain without throttle.

I don't really care what the EU does with "Up To".

 

[Abwx]

Only because Nvidia optimized for noise

They optimized the noise of their cards but unoptimized
the full PC noise by the same way....

From the other thread it seems that Nvidia s coolers are left
uncovered and as a consequence most of its heated air flow
will be blown inside the case...if there is one..

Measuring cards noises in isolation require careful study
of the measurement relevancy , in that case the Nvidia card
will heat the inside of the case much more than the Radeon
wich blow its air outside of the case , hence the card will be for
sure more silent but only to transfer the burden to the case
extracting fan that will spin harder to get out the heated air,
in short both the air extraction and subsequently a part of the
noise will be transfered to the case fans wich will negate
the noise advantage of the Nvidia cards when measured
in isolation.

 

[Z15CAM]

1,150MHz with a cooler that is arguably as good as any aftermarket cooler shipped?

That was all they could get with an overclock?

If I can match the performance of Tom's Accelero Xtreme III 290/X conversion, it's good enough for me, just as long as it's Relatively Quiet when OC'd.

Actually 1150Mhz is not too shabby of an OC for an AMD Reference Card considering it's not on a Water Block.

I may be a little off here but Reference Cooling Alternatives may be somewhat restricted trying to accommodate both Silent and Uber Thermal Mode Fan Profiles.

Enthusiasts no doubt will edit the Thermal Fan Profiles of the Cards Firmware to accommodate specifics. Of course this will likely void the cards warranty as the Flashing will not be AMD approved. Perhaps there will be a Driver Software solution for Editing FAN Spd Vs GPU Temperatures relative to the Environment the Card is in.

 

[wand3r3r]

There's currently a big difference between the boost implementation and up to implementation. Both have possibilities to skew the system.

If boost is actually boosting as high as it pleases and golden samples are reviewed, you get the point.

If the up to implementation doesn't actually reach maximum speeds often or in a uniform manner, same issue depending on review samples.

The key difference here, those in camp green feel let down that they cannot currently be guaranteed anything other than stock base clocks (which seems to be only in a terrible airflow situation) so they feel the need to "increase" the odds by some manual trick (fan speed/power boost). Next generation when team green outdoes red, the situation will be reversed and the merry go round will continue.

Turning up the fan, clocks, or power % is manual and isn't supported. Next generation I suspect it will be but until then it is not official and therefore is the same as manually overclocking the voltage neutered keplers. I'm interested in OC vs. OC but it's an entirely different issue.

If I had to choose one I'd only want to see uber mode. The 780 ti vs. "quiet" 290x review is so skewed it's ridiculous (totally ignoring that the "loud" factory supported 290x is dead even with the 780 ti is absurd).

I can see the dilemma between different modes, and this is potentially a whole can of worms in the future, so I'd say max performance mode and scathing comments on the drawbacks if relevant, however a fair comparison which considers all objective points.

Either way, from the debate, it's clear the brand fans will never agree. If one mode alone will be tested, it should be put to the community vote imo.

 

[GaiaHunter]

Only because Nvidia optimized for noise and performance, not raw performance.

Say let me add boost 3.0 on the 780 Ti and 100% fan.

Reviews show 780 Ti reference easily hitting 1250 to 1300 on reference air with +250/300 offsets, which is how you have to do it with boost 2.0. This performance is achieved with out touching the fan or voltage of the card.

Afaik the hardware of Titan and the 780 was capable of 340w consumption within specifications, they wouldn't have to change anything only increase the stock boost clock (offset for us) and the fan speed to allow it to maintain without throttle.

I don't really care what the EU does with "Up To".

Again more ifs.

Lets talk about those 1250-1300 reference.

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/30.html

Maximum overclock of our sample is 1120 MHz GPU (base) clock (28% overclocking) and 1975 MHz memory (13% overclock).

http://www.guru3d.com/articles_pages/geforce_gtx_780_ti_review,28.html
1075 MHz with boost of 1205-1276 with power target of 106%

• Power Target 106%
  Priority at Temperature target
  Temp Target 95 Degrees C
  CPU clock +100 MHz
  Mem clock +450 MHz

With this overclock we have extra performance at our hands, as the boost clock will now render at 1200~1276 MHz depending on the power and temperature signature. The GPU will continuously be dynamically altered on voltage and clock frequency to match the power and temperature targets versus the increased core clock. Have a peek at the results when overclocked.

http://www.hardwarecanucks.com/foru...-nvidia-geforce-gtx-780-ti-3gb-review-12.html

Even with NVIDIA's limits constraining frequencies, our sample hit a constant core speed 1143MHz with peaks around the 1176MHz mark and quite a bit higher than reference Boost clocks. Meanwhile, memory frequencies topped out at 7744MHz which is about average for 7Gbps modules.

The core in particular was a surprise but it needs to be mentioned that even with its extra overhead, NVIDIA’s Power Limit did step in the way and without it we have no doubt this card could have easily achieved even higher speeds. As it stands, with overclocks in place, the GTX 780 Ti is nearly a match for the GTX 690…and that is impressive to say the least!

http://www.hardware.fr/articles/912-24/gpu-boost-pratique.html

During our tests, carried out a reminder after the cards are mounted and temperature have stabilized, the frequency of the GTX Titan varies between 837 and 915 MHz, with an average of 863 MHz, far enough from the maximum frequency of 1006 MHz our sample. 's GeForce GTX 780 operates between 863 and 915 MHz, with an average of 876 MHz, against a maximum frequency of 1006 MHz. As for the GTX 780 Ti, its frequency varies between 876 and 967 MHz for an average of 915 MHz, whereas our sample can rise up to 1019 MHz. Once in 'Uber', for example with temperature limits and consumption to maximum or with alternative cooling system, the GeForce GTX 780 and GTX 780 Ti are close to their maximum or remains almost constantly. We then noted constant 993 MHz for the GTX 780 and 1006/1019 MHz for the GTX 780 Ti, which provides a significant performance gain. A note in the case of the latter in most tests consumption was very close the limit we had yet pushed 106%, the maximum allowed by Nvidia. This is why the frequency goes regularly to 1006 MHz. An insignificant decrease but indicates that there is very little margin available level of consumption and a more demanding sample may be more limited.

http://www.anandtech.com/show/7492/the-geforce-gtx-780-ti-review/16





http://www.hardwareheaven.com/revie...card-review-power-temps-and-overclocking.html





---------------------------------------------------------------------------------------------------
You might not care what EU does with up to but NVIDIA and AMD do since the EU is a big market. I guess they can create different cards for the EU and US market but that seems something very costly compared to the fact uber mode already reach 1000 MHz pretty much all the time (with maybe an exception or 2 where it will be 10-20Mhz short of that).

 

[BallaTheFeared]

Again more ifs

If you have a problem with ifs, take it up with Anand.

 

[GaiaHunter]

If you have a problem with ifs, take it up with Anand.

He himself said that it may look like a contrived scenario.

For me it definitely looks like a contrived scenario.

A preset or a switch under warranty will have to work otherwise it will cost them too much in RMAs and what not. That means that either NVIDIA or AMD cannot go too crazy on their presets that are under warranty - that is why uber mode is like ~330 more RPMs compared to quiet.

 

[BallaTheFeared]

Overclocking is warrantied, whats the difference?

330 more rpms between AMDs 40% and 55%, you can't honestly believe that.

 

[(sic)Klown12]

that is why uber mode is like ~330 more RPMs compared to quiet.

The difference between the two settings is actually ~1000 RPMs

 

[Jaydip]

Were reviewers have tried to equalize things, they lowered the R9s fan speed to 34% to match the noise levels of the geforces. They didn't speedup the geforces fans because that doesn't do to much for their performance as they will still throttle due to power limits yes?
Now you want them to up the geforce power settings because you know increased fanspeed won't do much for their performance compared to what it does for the R9s
Here, have a listen to to noise levels of a 780 at a setting that prevents thermal throttling compared the 47% fanspeed of the 290 and there goes the noise argument

I believe if users are smart enough to select a bios switch they are smart to maximize the power and temperature targets as well.

 

[AlNasty]

I kind of wonder if AMD called "Uber" mode "Standard" mode, along with a "Quiet" mode, would there be all of this controversy. Or, if they left them as "Uber" and "Quiet" but the dip switch had a
center location the did nothing, and shipped it in the center position so the user had to select.

Does anyone know if turning up the fan speed on a 780 or 780Ti to say, 50-60%, make it
run any faster with the stock voltage/settings?

 

[Keysplayr]

And when they do they are going to see results for quiet mode in the comparisons.

Wut? Is that automatic?

 

[GaiaHunter]

The difference between the two settings is actually ~1000 RPMs

Wrong logic on my part.

On the other hand would you like to post what temperatures you get with how much fan % in your 290 here?

 

[3DVagabond]

Wut? Is that automatic?

If they use the quiet mode as the default load those are the numbers they'll produce for the comparisons. That will handicap the 290X by about 7%.

 

[Silverforce11]

I kind of wonder if AMD called "Uber" mode "Standard" mode, along with a "Quiet" mode, would there be all of this controversy. Or, if they left them as "Uber" and "Quiet" but the dip switch had a
center location the did nothing, and shipped it in the center position so the user had to select.

Does anyone know if turning up the fan speed on a 780 or 780Ti to say, 50-60%, make it
run any faster with the stock voltage/settings?

Toms had a video on it for comparing noise levels, the reference 780 needed 80% fan speed to maintain its highest boost clocks. At lower fan, it would throttle. So yes, higher fan speed gives you more performance, but at stock settings, there is a limit determined by NV, so 80% to 100% makes no difference in performance.

@Balla, the difference between a flip on the actual card and loading up a 3rd party tool to enable vcore changes, power limits, clock speed and fan speed... it's not so obvious??

And becareful, not all AIBs will give you a new card once you brick yours OVERCLOCKING it, especially with vcore tweaks. Certainly here in Australia, if you RMA a GPU/CPU due to over-clocking.. well, lets say it won't end so well for you.

Really, if it was the reversed situation, the reference 780 or Titan came with a switch, 40% fan speed and 80% fan speed cap and the result of using the 80% fan speed is extra boost clocks, wouldn't you want good review sites to test both?? What exactly is the harm for the audience to have extra information available to them? Just extra work load for the reviewer, which in fact, the audience may appreciate greatly.

 

[(sic)Klown12]

Wrong logic on my part.

On the other hand would you like to post what temperatures you get with how much fan % in your 290 here?

With the stock fan profile with the overclock in my sig, I top out at 94C with 40-42%(~1900-2000 RPMs) fan. With a custom curve through Afterburner, I top out at 83c with a fan speed of 53%(~2800 RPMs). The ambient temps at the moment is ~20c and I have a case with good airflow and negative air pressure. These readings are all from Crysis 3 with a mixture of high and very high settings with SMAA.

 

[GaiaHunter]

With the stock fan profile with the overclock in my sig, I top out at 94C with 40-42%(~1900-2000 RPMs) fan. With a custom curve through Afterburner, I top out at 83c with a fan speed of 53%(~2800 RPMs). The ambient temps at the moment is ~20c and I have a case with good airflow and negative air pressure. These readings are all from Crysis 3 with a mixture of high and very high settings with SMAA.

Cool, thanks.

What would be the minimum fan speed that would prevent throttle at your OC?

I also assume that the noise isn't that problematic considering you prefer lower temperature over lower noise?

 

[Erenhardt]

Also, as far as opinion vs objective 'reviews'. Well, reviewers have opinions. They are supposed to, and their opinions are important. Otherwise AMD or Nvidia would have no use for them and would just put the scores up on their website and let that be it. So, people getting on Ryan for supposedly not being objective in his review per their opinion I think are being unfair. One of the reasons you are here reading his review (among a few others) is because you value his (their) 'opinion' and feel as though his (their) opinion is reputable. As thus, that also means he is not obliged to always agree with you, or you him. Otherwise he would just be a puppet, so to speak...which is no good.

Far from it! If we want subjective data on a card, hell, go to manufacturers slides, it doesn't get more subjective (biased) then that. Reviewers are here to show us real world data. Without any backstabbing, and crippling competition, and that is what is going to happen here. One product will be shown in not the best possible way, to say the least.

I think default, or the quite mode should be compared to whatever is similar.

Then why having noise/power consumption charts? lets equalize everything! What an exciting products, the only difference is color of the sticker!
There are so many ways both companies try to distinguish themselves from each other, that equalizing is (sorry for the word) dumb.

How is increasing fan profile an overclock. Hell, look the other way, 290X works downclocked while in "quiet mode", a mode to play old console ports without having to worry about v-sync.

For the record, my card came with the Uber mode bios set. It was a powercolor, and it seemed to have higher out of the box clock speeds too. I haven't tried OCing it yet, just adjusted fan max and thermal target.

You Sir, just spilled fuel to the fire!

 

[Erenhardt]

dp

 

[GaiaHunter]

You Sir, just spilled fuel to the fire!

First it was about out of the box settings but then it already changed to something about obnoxious higher performance settings compared to the competition.

If I was a manufacturer the message I would be getting was "ship only the highest performance setting instead of giving choices".

Maybe reviewers then start to manually change the fans, voltage, clocks, etc, to equalize cards.

Then the message would be "lock everything and don't allow users and reviewers to mess with any setting".

That is exactly what we enthusiasts want...

 

[(sic)Klown12]

Cool, thanks.

What would be the minimum fan speed that would prevent throttle at your OC?

I also assume that the noise isn't that problematic considering you prefer lower temperature over lower noise?

I haven't tested it to see exactly where it throttles due to lack of cooling, but I'd expect capping it at 40% would. As for the noise, I game with headphones so the only noise I worry about is idle and my card is well under the noise coming from my case and CPU fans. Without the headphones, it would only become slightly annoying once I past 55%. At that speed it's no worse than the reference blower design AMD used on Cypress, but past 70% it does live up to the negative reputation.

 

[BallaTheFeared]

You Sir, just spilled fuel to the fire!

Apoppin got the same card, comes in uber mode stock and with a clock speed of 1030MHz, oh and he said his doesn't throttle and the noise is any worse than Fermi but not quite as bad as the 6990 in unofficial mode.

It's considered a factory overclocked model

 

[Silverforce11]

Apoppin got the same card, comes in uber mode stock and with a clock speed of 1030MHz, oh and he said his doesn't throttle and the noise is any worse than Fermi but not quite as bad as the 6990 in unofficial mode.

It's considered a factory overclocked model

Yes, It's a powercolor model, they gave it 30mhz extra and its now an OC model. You see they have the same one for the R290: http://www.pccasegear.com/index.php?main_page=product_info&cPath=193_1575&products_id=25631

Boost to 975mhz and its a factory OC model.

AIBs have been doing that for years, a tiny clock speed increase from reference specs and its marketed as an OC card.. nothing new. As for ABT's 55% fan speed uber card running at 1030mhz and not throttling, that's some pretty magical heatsink and fan, because other review sites have shown it still throttles on Uber mode..

 

[Keysplayr]

If they use the quiet mode as the default load those are the numbers they'll produce for the comparisons. That will handicap the 290X by about 7%.

Oh, well yeah I would rather see both modes in any given review for comparison.
I would also like to see all this being able to be done via software instead of having to physically open my case and flipping the AUSUM switch. Why cant they do this with software? I understand there are two BIOS but still.