Spin Off: AT's Testing Methods & Uber Mode

[3DVagabond]

...continued

290X is 6% slower in quiet mode
780ti is 8% slower with reference cooling
780ti is 2% faster than 290X

290X is 8% slower in quiet mode
780ti is 11% slower with reference cooling
290X is 15% faster than 780ti

290X is 6% slower in quiet mode
780ti is 8% slower with reference cooling
0% difference in performance

290X is 7% slower in quiet mode
780ti is 8% slower with reference cooling
290X is 2% faster than 780ti

Overall in 14 games the 290X throttles performance by ~7% in quiet mode. Far from the 20% you are claiming.

In the same games the 780ti throttles performance by ~8% without adjusting for the effects of custom cooling. Again, not 1%-2% like you say.

When you average the 2 there is virtually no difference in performance over these 14 games. ~1% if you include Sleeping dogs. A virtual dead heat is you discard it as an outlier because of it's 15% difference favoring the 290X. Which obviously is not representative of the actual overall difference.

Where are you getting this 300MHz throttling for the 290X in quiet mode? You aren't referring to when W1zzard from TPU put his hand over the fan? I hope you have another example, because that would just be incredibly misleading to use that example and has already been debunked as representative of any real world conditions.

I would say the 1 or 2 bins you are claiming is pretty much disproved by the Hardware.fr results I've posted here. Both their 780 and 780ti are obviously throttling by far larger amounts than that.

Using your 780 ACX doesn't really apply to the discussion of "reference coolers", does it? I also use this to make my point that reference blowers on cards like these, even the Titan cooler, aren't desirable. Most people, including you, simply don't buy them unless they are forced to. Like with Titan and the current Hawaii situation when a card is new and custom cooled versions just aren't available.

Then you go off into your "cheap shroud" rant again. (We get it. You don't like the cooler.) Obviously you are using inaccurate figures to judge it by though. I think maybe you have read too much info and are misinterpreting and confusing it. The amount of passion you have against the Hawaii cooler seems genuine. I don't think you are purposely trying to be misleading, but I do think you are wrong.

Note: I'm not a proficient typist. This took forever to post. I hope you take the time to read it.

 

[Carfax83]

The heatsink dissipates watts, not degrees Celsius. Don't confuse heat and temperature. In terms of heat output the 780 Ti and 290X are very close, with review sites split between which consumes more power. The fact that the 290X is running 12 degrees hotter means that the ASIC is losing more power to leakage and helps the AMD reference cooler to achieve a higher performance and lower noise rating than it would be able to if AMD kept the maximum temperature at 83 degrees like the 780 Ti.

So yes, given that the 780 Ti heatsink is much quieter despite the fact that power dissipation is comparable between both products (and is being handicapped with the 83 degree temperature target to boot) it's fair to say that Nvidia has the far, far superior reference cooler.

That's a good explanation, thanks :thumbsup:

 

[Keysplayr]

The 290 cooler under the hood is exactly the same as the 5870, 6970, and 7970 coolers. Therefore the acoustic performance is directly related to using a 4 year design, and nothing but that - keep in mind that during actual gaming (non furmark) loads, that the Titan uses maybe 8-9 watts less than the 290X in uber mode. AMD could have easily made a better shroud to eliminate the need for a noise versus throttling quiet mode compromise, but they didn't. It's obvious as to why to anyone that has had experience with AMD Shrouds from the past 4 years. They never changed the under the hood design of the reference shroud for the past 4 years.

Further anecdotal evidence from those using accelero coolers and what not with the 290 are also showing shocking temperature drops of 40-50 degrees. The issue isn't the chip, per se - like I said, the Titan and 290 don't consume much differently in terms of power draw. Yet the Titan shroud is just superior in terms of acoustical performance, while the 4 year old cheap plastic design of the AMD shroud is beginning to show its age. This could have been completely prevented by AMD with a better shroud. The temperature differences that people are getting with 3rd party air coolers is precisely evidence of that, when a gelid or accelero cooler lowers temps by 40C, it makes it all the more obvious that AMD screwed the hell up in their design.

I should make it clear that noise isn't the big problem. The big problem is shedding 20% performance for any quiet fan profile. Again, that is pure freaking nonsense and consumers should not put up with that. ESPECIALLY when that is the default mode on the 290X.

Blackened, since most seem to want to stifle you here, perhaps another thread specific to the cooler would be in order? Just a suggestion.

 

[Grooveriding]

It would probably do nothing? Fairness?

I'm sure you've received your stipend of free hardware this generation. Go ahead and crank the fan to 100% and see if the card's clocks soars higher than they do with the fan at a lower setting, they won't.



Let's not resort to personal attacks, please.

-Rvenger

 

[Gikaseixas]

since most seem to want to stifle you here

Not even close, please do not start
He's derailing the thread period (with his emotions)

 

[Will Robinson]

Put me on ignore then. I have a right to express my opinion(S), even though i'm sure you don't want me to point out the obvious flaws with the 290. And I have to say, I've always been mostly positive with ATI GPUs and I was very happy with my former 7970s. I was also SUPER excited about the 290 series prior to launch and wanted it to be AMD's best GPU in years. Yet, the throttling situation with the 290X isn't acceptable - no prior ATI card has ever had a 20% performance deficiency for using a custom quiet fan profile. Again, if you don't like my opinion(S) --- ignore feature. Use it.

Fixed that for you.
If I had you on ignore that would've made the thread about 4 -5 pages shorter surely?

 

[blackened23]

What you do is spreading FUD. There is no 20% difference in performance between quiet and uber.

http://www.anandtech.com/show/7481/the-amd-radeon-r9-290-review/15

Stop claiming that this is off topic. The performance differences between a "warmed up" 290X, uber mode and quiet mode has EVERYTHING to do with the AMD reference shrould, and you know it. The only way that AMD can get this message is to let them know that we as consumers will not accept cards that perform great on a first run, but lower in performance as they warm up - this is what happens with the quiet mode profile. Additionally, throttling down to 727mhz in quiet mode? That's cool right? This has everything to do with the AMD reference shroud, stop pretending that it's unrelated. It has everything to do with the topic at hand.

 

[Saylick]

/snip
AMD's Uber mode in my eyes isn't the same as a factory overclocked card. At the same time, it's not the same as what we've done in the past - which is test a totally stock configuration (reference clocks and fan speed). I personally believe in the whole living document philosophy when it comes to things like constitutions or review policies, but here's where we can get into trouble. In the case of the 290X, AMD has two modes and you can make a good argument for why you should test both. Let's now take it one step further: what happens if NVIDIA shows up next round with 3 modes? Do we test all 3? Which modes do we then compare against AMD modes, particularly if they only line up along one vector (e.g. performance or acoustics, not both). What if AMD responds the next round with 4 modes, etc... It can quickly get out of hand.

What I'd like to do here is define a good policy for what to do if this turns into a fan speed arms race. Dealing with the 290X is simple: Ryan tested both quiet and uber modes, and I can totally appreciate the argument for including analysis based on both. What Ryan is concerned about is the future. This isn't a matter of him being lazy (me being the person he reports to, I can tell you that's definitely not the case - he's kept up an insane work schedule over these past several weeks in order to get everything done as best as possible. The launches aren't done yet for the year, add in short NDA windows, issues with cards/drivers and of course any travel and the pace you have to keep in order to put out these reviews is insane). The precedent we set here today will directly impact what manufacturers attempt to do with their reviews programs in the future. The safe bet is to stick with testing in default configurations. I am (and assuming Ryan is too) more than willing to expand/change/redefine that, but the question is how? Let's look beyond the present 290X situation and think about what happens next. If acoustics and performance become even more tightly coupled in future GPU designs, and multiple optimization points exist for each card (with 1 default setting obviously) how should we deal with that going forward? If things get crazy, we could be in a situation where there would even have to be a tradeoff in terms of review depth vs. card configuration combinations. E.g. would you be willing to give up a resolution setting across all games tested in order to get another operating mode included? What does this do to the complexity of graphs?
/snip

Take care,
Anand

Anand, first off, thanks for taking the time to ask your readers for their opinion. I've been a long time reader/lurker but didn't start being actively involved with the ATFs until recently but your gesture genuinely shows how much you value your readership and I appreciate that. AT reviews are some of the best I've seen and I wish to see that trend continue for many years to come.

With that said, I understand your concern for not only producing a solution for the situation at hand but to also produce a viable long-term solution. To answer your question of what AT should do if nVidia/AMD start releasing cards with multiple presets, it is my understanding that there are in fact only TWO fundamental parameters - "vectors" as you call them - and these are framerate performance and acoustic performance. Every other property thus becomes a consequence of optimizing/balancing these two principal variables; the amount of FPS, heat, noise and power draw generated are going to be functions of clockspeed and fanspeed. Furthermore, because today's graphics cards are running some form of turbo mode, clockspeed and fanspeed are inevitably going to be functions of each other, which I believe simplifies the problem.

Consequently, should vendors release cards with multiple modes which are all covered under warranty, I propose that only a maximum of three (3) modes be tested. These three modes shall be: 1) the default settings, 2) the setting that offers maximum FPS performance/minimum acoustic performance, and 3) the setting that offers maximum acoustic performance/minimum FPS performance. Should vendors release cards with >3 presets, they will inevitably fall somewhere in between the min/max range which is described above, so it becomes unnecessary to test these modes. Realistically speaking, we shouldn't expect the min/max range to be excessively wide so we shouldn't see cards with 100% fanspeed "FPS = KING" modes and we shouldn't see cards with 30% fanspeed "SILENCE IS GOLDEN" modes either. You may take this one step further and begin plotting FPS vs. Acoustic data for each card. Using linear interpolation, a reader may then estimate how much FPS is to be expected at a given sound level. A card's FPS/dB efficiency can thus be assessed using these charts as well, where if FPS was on the vertical axis and dB was on the horizontal axis, cards who have plots closer to the top-left of the graph are more "efficient". This ultimately leaves reviewer subjectivity out of the frame and replaces it with stone-cold facts and objectivity.

Like you stated this begs the question of how review depth will be influenced by a new review policy. If what I proposed was to be implemented for each benchmark, Ryan would easily begin spending double the amount of time reviewing new cards. Clearly, this is not a long-term solution and is thus unfeasible. We need to produce a future-proof outline which covers most of the critical bases while including a small section which provides a summary of the performance vs. acoustics. I personally believe that if something has to give, it should be the ultra-high resolution benchmarks first. Now, I'm not saying that reviews should omit them entirely. I just believe that they could be condensed into a small section of their own simply because at the end of the day, the most popular resolutions are 1600p and below. In other words, I would rather have the reviews swap out the 3840x2160 benchmarks with only 1440p and 1600p and then proceed to only test 3840x2160 medium for the two games which nVidia and AMD typically do best in, respectively. This should free up some time such that you may test FPS vs. acoustics.

In short, it is my belief that the outline for all future video card reviews should follow something similar to what follows:
1) Title Page
2) Meet the XXX
3) New Features
4) Testing Setup and Methodologies
5 to N) Games 1 to X @ 3 different popular resolutions (NO Ultrahigh resolutions)
N+1) Ultrahigh resolution 3840x2160 (2 games: 1 nVidia optimized, 1 AMD optimized)
N+2) Synthetics
N+3) Compute
N+4) Max/Min Presets: Power Draw, Temperature, Noise (2 games at popular resolutions: 1 nVidia optimized, 1 AMD optimized)
N+5) Overclocking: Power Draw, Temperature, Noise
N+6) Conclusion

This should be the minimum. Should time permit, feel free to add additional benchmarks. I suggest adding in the rest of the benchmarks for all the games to the Max/Min section first before fleshing out the Ultrahigh resolution section.

 

[BoFox]

Overall in 14 games the 290X throttles performance by ~7% in quiet mode. Far from the 20% you are claiming.

In the same games the 780ti throttles performance by ~8% without adjusting for the effects of custom cooling. Again, not 1%-2% like you say.

When you average the 2 there is virtually no difference in performance over these 14 games. ~1% if you include Sleeping dogs. A virtual dead heat is you discard it as an outlier because of it's 15% difference favoring the 290X. Which obviously is not representative of the actual overall difference.

Where are you getting this 300MHz throttling for the 290X in quiet mode? You aren't referring to when W1zzard from TPU put his hand over the fan? I hope you have another example, because that would just be incredibly misleading to use that example and has already been debunked as representative of any real world conditions.

I would say the 1 or 2 bins you are claiming is pretty much disproved by the Hardware.fr results I've posted here. Both their 780 and 780ti are obviously throttling by far larger amounts than that.

Using your 780 ACX doesn't really apply to the discussion of "reference coolers", does it? I also use this to make my point that reference blowers on cards like these, even the Titan cooler, aren't desirable. Most people, including you, simply don't buy them unless they are forced to. Like with Titan and the current Hawaii situation when a card is new and custom cooled versions just aren't available.

Then you go off into your "cheap shroud" rant again. (We get it. You don't like the cooler.) Obviously you are using inaccurate figures to judge it by though. I think maybe you have read too much info and are misinterpreting and confusing it. The amount of passion you have against the Hawaii cooler seems genuine. I don't think you are purposely trying to be misleading, but I do think you are wrong.

Note: I'm not a proficient typist. This took forever to post. I hope you take the time to read it.

Good point!! I gotta double-check on this myself!

 

[Keysplayr]

Not even close, please do not start
He's derailing the thread period (with his emotions)

You should be thanking me for suggesting it. And you're welcome.

 

[Atreidin]

Not even close, please do not start
He's derailing the thread period (with his emotions)

His monomania is making this thread any many others annoying to read. I'm not sure what he wants, are we supposed to march on AMD's headquarters and write our congressmen to intervene?

Not to mention it isn't that bad. Things could be better but it isn't awful. I'd also respect someone's opinion of a product more if they (and this might sound drastic, but hear me out) ever even saw it in use in real life at one point.

 

[Erenhardt]

Stop claiming that this is off topic. The performance differences between a "warmed up" 290X, uber mode and quiet mode has EVERYTHING to do with the AMD reference shrould, and you know it.

Read the topic again. There is nowhere "reference amd shroud" in there. We are not discussing performance, not comparing throttling of cards. We are discussing testing methods which are influanced by introduction of "uber mode". STOP.

PS. Thank you for showing graphs that have absolutely nothing to do with topic. It is not about throttling, and 290 (non-X) doesn't even have uber mode.

 

[HurleyBird]

N+1) Ultrahigh resolution 3840x2160 (2 games only: 1 nVidia optimized, 1 AMD optimized)

Testing only two games, one of which is the AMD best case and the other of which is the Nvidia best case, would tell readers absolutely nothing. Perhaps worse than nothing. This is a poorly thought out idea.

 

[Saylick]

Testing only two games, one of which is the AMD best case and the other of which is the Nvidia best case, would tell readers absolutely nothing. Perhaps worse than nothing. This is a poorly thought out idea.

How many people actually run at resolutions that high to begin with?
Also, I made edits to my original post and stated that:

This should be the minimum. Should time permit, feel free to add additional benchmarks. I suggest adding in the rest of the benchmarks for all the games to the Max/Min section first before fleshing out the Ultrahigh resolution section.

 

[3DVagabond]

How about you raise the fan speed on the 780 Ti so that it gets maximum boost clock speed at all times with no down clocking?

That's basically what "uber mode" is on the 290x..

The two "boost" systems are not at all similar. They actually work exactly the opposite of each other. The thing I find strange is according to Ryan if AMD simply shipped the cards with the switch in the opposite position he would have tested in uber mode. Seems like a pretty arbitrary way to decide which setting to use.

 

[iiiankiii]

Blackened, since most seem to want to stifle you here, perhaps another thread specific to the cooler would be in order? Just a suggestion.

I second this. This isn't about the reference cooler. Maybe that can be discussed there? Might even help get the point across better b/c it's more focused.

With that said, this thread is about how Anandtech should handle multiple performance mode and its methodology.

Back on topic. I can definitely see a potential problem with cards shipping with multiple performance modes. There has to be a set default mode for testing purposes or else things might get out of hand quick. As others have suggested, I think testing the highest and the lowest performance modes might be the best solution.

As for exploring all of the modes available (if more are present), the initial review should cover all of the modes. All sequential reviews should only include the highest and the lowest performance modes.

 

[Gikaseixas]

You should be thanking me for suggesting it. And you're welcome.

Ok thank you Keys
wanna bet this aint gonna stop him?

 

[Abwx]

http://www.anandtech.com/show/7481/the-amd-radeon-r9-290-review/15

Stop claiming that this is off topic. The performance differences between a "warmed up" 290X, uber mode and quiet mode has EVERYTHING to do with the AMD reference shrould, and you know it. The only way that AMD can get this message is to let them know that we as consumers will not accept cards that perform great on a first run, but lower in performance as they warm up - this is what happens with the quiet mode profile. Additionally, throttling down to 727mhz in quiet mode? That's cool right? This has everything to do with the AMD reference shroud, stop pretending that it's unrelated. It has everything to do with the topic at hand.

Nice , but let s see with 15 games at 3 resolutions and do
an average for each res...

But , what happens at hardware.fr.??..Yet their test room
temp is between 26°C and 26.9°C , not exactly kind for
the card but in line with the usual number we take in electronic
design simulation , that is, 300°K , quite a professional approach
from said site.

Nowhere we see 20% between quiet and uber.

http://www.hardware.fr/articles/910-25/recapitulatif-performances.html

 

[raghu78]

Oh, if you say so. What really bothers me is that using a quiet fan profile results in a 20% loss of performance. Yet you keep telling us it's cool. 20% performance lost, hey no problem.

Just FYI, I loved my former ATI cards (5870 - 7970) but none of those cards required an insanely high fan profile to perform at their peak; my 7970s overclocked like crazy and could do so with a sub 45% fan speed. Sadly this isn't the case with the 290 - as i've mentioned, noise can be a problem but was easily solved with the 7970 (USE A LOWER FAN SPEED). This isn't a solution with the 290X because you lose 20% performance in the process.

But, according to you, this is cool right? We should accept this as consumers? I guess so, according to you, because from your perspective AMD can do no wrong. Which is hilarious because I guarantee that for every 1 person that buys a 290 based on value, another 2 people will avoid it because of the huge web controversy over throttling and retail sample variance. But according to you, that's all cool. 20% performance lost, no problem, right? And you call this a good product? Please.

the R9 290 is is an outstanding product with one weakness. fan noise. R9 290X is good but its value proposition compared to R9 290X is not so strong. in fact my opinion is clear. the uber fan speed should have been the only mode on R9 290X / R9 290. in fact AMD should have kept the fan speed at a fixed 2650 rpm for R9 290 and a fixed 2800 rpm for R9 290X and avoided all the performance inconsistency problems :whiste:

http://www.hardocp.com/article/2013/11/07/nvidia_geforce_gtx_780_ti_video_card_review/8

"AMD gave us two fan modes on the R9 290X. On one hand we have "Quiet Mode" with its 40% fan RPM cap. On the other hand we can have a louder, but better gaming experience, with "Uber Mode" using a 55% RPM fan cap. These modes have fueled a lot of debate and a lot of FUD. Using 40% Quiet Mode, the fan is fairly quiet. With Uber Mode’s 55% the sound profile becomes more debatable. We would call Uber Mode moderately loud, yes you can hear it, but it is not annoying loud. Manually setting a 65%+ fan cap, the reference cooling solution is very loud and will likely a agitate many users. We are running the R9 290X in Uber Mode for the review today. In Uber Mode this reference-cooled R9 290X is no doubt louder than the other cards, but to refer to it as a "jet engine" at these speeds is not close to a fair comparison. We have no issues with gaming with the R9 290 in Uber Mode sitting next to us on an open test bench."

http://techreport.com/review/25602/amd-radeon-r9-290-graphics-card-reviewed/9

"Much has been made of the R9 290X's relatively high power draw, operating temperatures, and noise levels. Obviously, the R9 290 shares these same characteristics, with a somewhat louder default fan profile. In my view, the only one of these properties that's really worth fussing over is the noise, since it's the thing you'll notice in day-to-day use.

We're apparently going to have to face this price/performance-versus-acoustics tradeoff for a while, so I spent some quality time with the R9 290 trying to get a handle on what I think of the noise, beyond the readings on the decibel meter. I've gotta say, there are some mitigating factors. For one, I like AMD's choice to stick with a blower that exhausts hot air out of the case rather than going for a triple-fan cooler that doesn't. I've seen those fan-based aftermarket coolers perform poorly in multi-GPU configs, and they often occupy quite a bit more space—maybe even a third expansion slot—in order to work their magic. I'm also not convinced AMD's cooler is a poor performer and therefore noisy, as some folks seem to think. Remember, it has more heat to remove than any of the coolers on the other cards we tested. Finally, I don't think this blower's ~49 dBA reading is the worst of its type. The quality of the sound isn't grating. Subjectively speaking, there are much more annoying coolers in this territory on the decibel meter. The impressively smooth, gradual ramp of fan speeds up and down in the new PowerTune algorithm helps make the noise less noticeable, too. This ain't an FX-5800 Ultra, folks."

different people have different tolerances for acceptable fan noise and that adds even more subjective ness to the topic.

so I have to say your ranting is not based on any objective analysis and is pure drivel. :whiste:

 

[wand3r3r]

So I've got some thoughts here, and perhaps we'll turn these into a post on the main site but I wanted to get in this thread as you guys are honestly the source/inspiration for any such post.

Let's start with why we want to define a clear framework for how general performance/power/sound testing goes. Not only does it allow for fair comparisons between products, it also helps us deal with the inevitable situation where a manufacturer submits a ringer for review (e.g. factory overclocked card). I don't think there's much argument against this point - we all want a level playing field.

Similarly, it should be obvious why we'd want to include in such a framework the idea of testing a card at default settings. Having a strict policy there prevents a situation where AMD/Intel/NVIDIA show up and say hey we're selling the card in configuration x because of yields/experience/someothervalidreason, but it's really quite awesome and can run in configuration x+50% and that's how you should test it and btw we rule the world if you test it like that. This makes a lot of sense particularly when talking about encouraging factory overclocked comparisons.

The close relationship between fan speed and performance sort of throws a wrench in all of this. When Intel first started introducing aggressive turbo modes back in Lynnfield I was worried that it would completely corrupt our ability to reliably test CPUs. It turns out that wasn't the case. With graphics however the situation is a bit different, and with the 290/290X we're beginning to get a feel for exactly why that is.

I originally assumed the reason this was a problem now (and is going to be in the future) is because we're stuck on 28nm trying to get more performance without a good process tech solution until 14/16nm FinFET in 2015. Now I'm feeling like this is just going to be a part of the reality going forward, so we need a real solution.

AMD's Uber mode in my eyes isn't the same as a factory overclocked card. At the same time, it's not the same as what we've done in the past - which is test a totally stock configuration (reference clocks and fan speed). I personally believe in the whole living document philosophy when it comes to things like constitutions or review policies, but here's where we can get into trouble. In the case of the 290X, AMD has two modes and you can make a good argument for why you should test both. Let's now take it one step further: what happens if NVIDIA shows up next round with 3 modes? Do we test all 3? Which modes do we then compare against AMD modes, particularly if they only line up along one vector (e.g. performance or acoustics, not both). What if AMD responds the next round with 4 modes, etc... It can quickly get out of hand.

What I'd like to do here is define a good policy for what to do if this turns into a fan speed arms race. Dealing with the 290X is simple: Ryan tested both quiet and uber modes, and I can totally appreciate the argument for including analysis based on both. What Ryan is concerned about is the future. This isn't a matter of him being lazy (me being the person he reports to, I can tell you that's definitely not the case - he's kept up an insane work schedule over these past several weeks in order to get everything done as best as possible. The launches aren't done yet for the year, add in short NDA windows, issues with cards/drivers and of course any travel and the pace you have to keep in order to put out these reviews is insane). The precedent we set here today will directly impact what manufacturers attempt to do with their reviews programs in the future. The safe bet is to stick with testing in default configurations. I am (and assuming Ryan is too) more than willing to expand/change/redefine that, but the question is how? Let's look beyond the present 290X situation and think about what happens next. If acoustics and performance become even more tightly coupled in future GPU designs, and multiple optimization points exist for each card (with 1 default setting obviously) how should we deal with that going forward? If things get crazy, we could be in a situation where there would even have to be a tradeoff in terms of review depth vs. card configuration combinations. E.g. would you be willing to give up a resolution setting across all games tested in order to get another operating mode included? What does this do to the complexity of graphs?

I don't know that I've got the answer/a solution here, but this is the discussion I'd love to have.

As I mentioned earlier, this is a discussion that we might take to the main site at some point. We felt like we owed it to you guys to start it here given the time/effort you guys have put into it already. We're here to listen and will obviously take your input into account (e.g. the 290 fan noise update was a direct result of your feedback). All that I'd ask is please be respectful of Ryan in your discussions of his work. He really puts a ton of time and effort into this stuff and takes all of your feedback very seriously. Obviously you're free to post/say whatever (as long as it doesn't violate our ToS), but I've always been a fan of the golden rule

Thank you all for reading the site and for caring enough to engage in hundreds of comments on the forums and on the site itself. I'm off to bed for now but I'll check back tomorrow.

Take care,
Anand

Overall:

I think you are correct, the next round will almost certainly have NV including similar modes (unless they find they can benefit more by smearing), since clearly sacrificing something (noise in 290/x case) for more FPS is desirable for charts as long as FPS remain the foremost criteria. This may turn into another way to gain something over the competition which is why AMD went that route, and NV tries to smear it (continous cat and mouse). I agree testing 3 different modes in each card passing in the door will be tedious and it would get annoying to decipher how the cards relate and what the loss is due to the different modes.

The foremost criteria in GPUs are certainly performance, and price followed by a subjective mix with noise, and power usage.

Subjectivity:

I have an issue when subjective issues override everything else in a review unless they are extreme. When issues like that occur, it's best to keep the objective and subjective statements separate. However it's difficult on where to draw the line, several respected reviewers stated the fan noise is acceptable, while Ryan despised it. The smoothness viral campaign was noted for years by Brent and Kyle, and they repeated the subjective analysis however it didn't override the entire review since it was a subjective opinion without hard data (until a year ago). So subjective opinions are important, they just have to toe a line since everyone will never agree on them anyway.

The biggest issue is that the 290/x noise is subjective and some accept it, so while completely bashing the 290 on an entirely subjective opinion based criteria - noise, and essentially praising the 780 ti while in reality it's proven to be a tie to the 290x loud mode, comes across as subjectivity overriding objectivity. The opinions are based on ommiting data, that's about the worst thing a reviewer can do, it looks like (paid) bias at the worst, subjective at the least.

So what is the universal approach to the topic?

Perhaps a wait and see what the next generation approach to the boost ideas are? I am mixed on the overall idea of dual bios. If I had the card it would unquestionably be uber bios and cranked up as high as I could stand with an overclock. This is certainly not the same approach others would take so it certainly would be best to show all the data, and make opinions based on the data as a whole.

As for reducing test data to allow for these new variables to be tested, I would suggest avoiding pointless resolutions. High end = 2-3 high resolutions which could be voted on or simply picked. Mid range = 2-3 resolutions in the middle of the spectrum. If the cards should only be tested in one mode, maybe it should be asked from the community.

At this point I would want data on everything. I can see the wormhole that could be coming if there are 5 modes on every card.

 

[3DVagabond]

Was the review comparing the noise levels of 290x uber or quiet vs 780ti?

The noise levels on Crysis3 for the 780ti/290x quiet (51.7, 53.3) is a 1.6db difference, and I'm wondering myself if that a significant enough advantage to the 780ti. The furmark score is 1db advantage for the 780ti.

I don't like the way things are going with uber mode and the 94C max temp. And the 780ti max temps is also creeping north of 80C. I'm wondering if the lifespan of products are being shortened in this round of minor tweaks and rebadging.

TSMC's 28nm process is rated for 125°C (Which is military spec). 80° to 95° isn't a concern.

 

[3DVagabond]

When you start at such a low clock speed as 662, it's easy to be impressed by the 290 if you're not an astute technically competent observer. :sneaky:

However, while the 290 will boost a higher percentage gain, it doesn't offer the kind of flexibility you see with GPU boost 2.0 in terms of not only clock speed, but fan speed as well.

The GTX 780 Ti will boost ABOVE it's normal boost clock speed.. In the review, it had a max boost of 1020 (regular boost is 928). With a better cooler, I wouldn't be surprised if went all the way to 1100.

And all of this is done with less noise than the AMD card, which is basically stuck at 47%. Any lower, and it begins to throttle..

Now you can argue that a AIB cooler will solve this issue, and it may. But, after reading more reviews, I don't think the reference cooler is as bad as people are saying. It just has to remove a lot more heat, as the 290 series is designed to run up to 95c..

No it doesn't have to remove a lot more heat. Heat is not determined by the temp that the chip is operating at. It's determined by how much power is being used and has to be dissipated. The average power usage between the two is not very different.

 

[3DVagabond]

I think you are forgetting that the 290 was designed to run at a high thermal threshold. In the temperature tests, it was consistently pegged at 94c.

You can't blame everything on the reference cooler. The 290 runs hotter period, because it was designed to.. Kepler has a much lower thermal threshold..

They are both made on the same 28nm TSMC process. One GPU is not designed to run hotter than the other. There are other components on the PCB though that very well might be designed to run hotter safely. We don't know that though. All of this stuff people say about being designed to run hotter is generally simply ill informed rationalizing.

 

[Keysplayr]

Ok thank you Keys
wanna bet this aint gonna stop him?

Stop him from what? Talking about something that you don't want discussed? AFAICT, I've seen him explain why what he is talking about is on topic. So, stop him from what? You telling him to knock off the "off topic" is more "off topic" than him discussing it. Put the "off topic" card away.

 

[wand3r3r]

...continued

Overall in 14 games the 290X throttles performance by ~7% in quiet mode. Far from the 20% you are claiming.

In the same games the 780ti throttles performance by ~8% without adjusting for the effects of custom cooling. Again, not 1%-2% like you say.

When you average the 2 there is virtually no difference in performance over these 14 games. ~1% if you include Sleeping dogs. A virtual dead heat is you discard it as an outlier because of it's 15% difference favoring the 290X. Which obviously is not representative of the actual overall difference.

Where are you getting this 300MHz throttling for the 290X in quiet mode? You aren't referring to when W1zzard from TPU put his hand over the fan? I hope you have another example, because that would just be incredibly misleading to use that example and has already been debunked as representative of any real world conditions.

I would say the 1 or 2 bins you are claiming is pretty much disproved by the Hardware.fr results I've posted here. Both their 780 and 780ti are obviously throttling by far larger amounts than that.


Then you go off into your "cheap shroud" rant again. (We get it. You don't like the cooler.) Obviously you are using inaccurate figures to judge it by though. I think maybe you have read too much info and are misinterpreting and confusing it. The amount of passion you have against the Hawaii cooler seems genuine. I don't think you are purposely trying to be misleading, but I do think you are wrong.

Interesting. So it's throttling by a massive 7%, while the 780 ti throttles by 8%. Where is the outrage? It's not what posters here would like you to believe.

Also, it's been demonstrated multiple times that the titan (and I believe 780?) throttles by quite a bit, somewhere in the same ballpark. That German site has demonstrated it as well as some others.

None of that justifies it (throttling), however the smearing on these forums appear a little one sided.

This is going off topic however, I would suggest a new thread.

Relating to this thread:

The key to this whole idea is relaying this information to the user. AT hasn't been investigating many of the 'issues' like this (or smoothness, whatever the latest viral campaigns are), however I would find it beneficial if they go beyond straightforward reviews and investigate stuff like this (in entirely different articles that is). The capability is definitely here, the writers are sharp, it would be great to see it used beyond conventional straightforward reviews. (Every now and then there are little gems, so actually I would just like to see more)

To try clarify, I would love in depth articles about throttling (and other topics). Both sides are guilty and it would be interesting to see the data behind it.

AT has generally been an excellent resource, I think a little examination on the future direction is in need particularly after the 290/780 ti debacle.