[H]ard Does New Heat / Noise Test (real world) for 480

[ronnn]

I thought the 480 tended to keep heating up for as long as 20 minutes with gaming. No need for an audio file that long, but it should be tested for a good period of time.

 

[apoppin]

I was hoping they could cook an egg on it

That would be next Fryday

You wouldn't have to record the entirety of the test - just the last 2 minutes (?) or so. You said you wanted this to be as 'real world' as possible. I'm not sure about you, but I don't fire up BFBC2 for only 2 minutes at a time.

Do you want to listen to a ten minute test if a 1 minute one will do?
- and .. actually .. yes, sometimes less than 2 minutes

---if i get one of them GSoDs

... or your Nvidia display driver has stopped responding and has recovered ....



seriously .. i want to illustrate each video card at idle and then as it starts to spin up and go to max load .. and then to the fan spin up.

... isn't that what we want to test?

 

[cbn]

who wants to listen to a long audio file?
- especially if a short one will do as well?

Let me see if i got this right:

i fire up a game demo and let the automatic fan profile take care of the fan RPMs as we listen to it as the GPU loads to 100%
--rinse and repeat with each video card with the testing parameters exactly the same ... i'll also include case temps.

Do we need a long demo to make the test valid?

How long does the demo need to be in order for 480 fan to hit the maximum speed?

I noticed in the [H] sound tests 480 GTX fan speed didn't hit maximum until 1:21 or 1:25 into the video (and that was using Furmark).

 

[apoppin]

I thought the 480 tended to keep heating up for as long as 20 minutes with gaming. No need for an audio file that long, but it should be tested for a good period of time.

A GPU heats up pretty quickly under max load; gaming for 20 minutes would give you the case temps.

i am thinking around 2 minutes .. that is why the Crysis GPU demo run would be good at 25x16 and load the GPU completely; it runs three times also continuously;

Crysis GPU demo starts light on the GPU in the clouds and goes to its own max pretty quickly especially if i add AA; the fan on any video card should kick up to its maximum RPMs within 1-2 minutes (of course i can pre test this)

You guys are helping me set the testing parameters. Thanks!

 

[cbn]

it heats up pretty quickly under max load

Yeah, but the fan still delays.

 

[cbn]

i am thinking 2 minutes .. that is why the Crysis GPU demo run would be good at 25x16 and load the GPU completely; it runs three times also continuously;
it starts light on the GPU in the clouds and goes to its own max pretty quickly especially if i add AA; the fan on any video card should kick up to maximum RPMs within 1-2 minutes (of course i can pre test this)

You guys are helping me set the testing parameters. Thanks!

I would check at 3 minutes, 4 minutes and 5 minutes. If the fan gets even louder at 3-5 minutes into the test....then you know the audio files need to be longer.

2 minutes is probably the bare minimum as I am pretty sure Furmark is more demanding than Crysis demo.

 

[BFG10K]

You wouldn't have to record the entirety of the test - just the last 2 minutes (?) or so. You said you wanted this to be as 'real world' as possible. I'm not sure about you, but I don't fire up BFBC2 for only 2 minutes at a time.

This is a very good point. All that’s important is the maximum noise level attained by each card. We don't really care when the cards are equal, we care when they’re different, and how different they are.

The videos don’t even have to be in sync, as long as we hear each card at its loudest volume. 30 seconds or so of each card at its loudest volume during the test will be fine for comparison purposes.

 

[blanketyblank]

That would be next Fryday



Do you want to listen to a ten minute test if a 1 minute one will do?
- and .. actually .. yes, sometimes less than 2 minutes

---if i get one of them GSoDs

... or your Nvidia display driver has stopped responding and has recovered ....



seriously .. i want to illustrate each video card at idle and then as it starts to spin up and go to max load .. and then to the fan spin up.

... isn't that what we want to test?

not quite. We don't really just want the fan speed/sound when it hits max load, but we want the fan speed/sound after a good period of time say over 30 minutes or over sustaining max load. It's potentially different since the fan might have to go higher rpm to sustain the temps. Based on [h]'s review metro 2033 seems to be able to push gpus to 98% load so that seems like a fair real world bench for these cards. Otherwise if the load is only 65% like in borderlands or something you'd be better off buying a 5850 or 470 anyways.

However what's the GPU load on these cards from the Crysis demo? If it is consistently over 90% I guess that's good enough.

 

[BFG10K]

How long does the demo need to be in order for 480 fan to hit the maximum speed?

He can keep running the test until the noise isn’t getting any louder (a constant volume for 5-10 minutes or so). When that happens, he can record 30 seconds or so of the card in that state.

 

[cbn]

He can keep running the test until the noise isn’t getting any louder (a constant volume for 5-10 minutes or so). When that happens, he can record 30 seconds or so of the card in that state.

Good idea.

 

[apoppin]

He can keep running the test until the noise isn’t getting any louder (a constant volume for 5-10 minutes or so). When that happens, he can record 30 seconds or so of the card in that state.

No worries with loading the GPU with Crysis at 2560x1600 and with AA; i would say 3 minutes total is plenty to get any of the video card's fan to max RPM
(of course, i will pre test it to be certain and watch temps to see if they climb over time)

We also want to hear the start of Crysis - when it is near idle during the last few seconds of the loading screen it is pretty quiet and as it starts to ramp up as your avatar falls through the clouds, and then as it hits max RPMs

ALSO, after looping it for nearly 3 times, i should also catch the ending sounds of each video card.

You really want to hear all six cards?

 

[blanketyblank]

i would say 3 minutes total is plenty to get the fan to max RPM
(of course, i will pre test it)

We also want to hear the start of Crysis - when it is at idle (during the loading screen it is pretty quiet), as it starts to ramp up as your avatar falls thru the clouds, and then as it hits max

ALSO, after looping it for nearly 3 times, i should also catch the ending sounds of each video card.

You really want to hear all six cards?

I want to hear 2900XT the loudest, 480, 5870, and maybe 4870.
Not sure if you need the others, but if you're going to do 4 why not?

 

[BFG10K]

No worries with loading the GPU with Crysis at 2560x1600 and with AA; i would say 3 minutes total is plenty to get any of the video card's fan to max RPM

3 minutes isn't nearly enough; it often takes 10-15 minutes for my GTX285 to achieve a constant loaded fan speed. If I fire up a game for just 5 minutes, it often doesn't even ramp up its fan from stock.

You should keep looping the benchmark until the fan noise isn't getting louder (i.e. you get a constant volume for about 5-10 seconds). Once that happens, record each card in that state for about 30 seconds or so.

You can do fan ramping tests in other videos, but what's most important is capturing each card’s maximum noise level. Nobody cares when the cards are equal; what we want to see is when the cards are different, and how different they are.

 

[apoppin]

i tell you what

*i* want to demonstrate fan ramping up - same test; same audio files

i will post the first part of Crysis as i said .. loop it for 15 minutes - and then record the ending 20 or thirty seconds.

How's that?

 

[digitaldurandal]

"Reputable warranty" will mean shit if they don't have it in stock at the time you RMA one. And judging from the limited availability of Fermi to retail vendors, I suspect a lot of people will be shit out of luck when RMA'ing this thing.

Let's see how long before I'll see a post in here with some title like:
"...(so and so company).. warranty sucks because they don't have GTX480 available"

Theyll give you a card from the next generation and you will win out. At least thats my experience with out of stock card replacements

 

[BFG10K]

i tell you what

*i* want to demonstrate fan ramping up - same test; same audio files

i will post the first part of Crysis as i said .. loop it for 15 minutes - and then record the ending 20 or thirty seconds.

How's that?

Oh sure, fan ramping is fine, just be sure you get the main bit correct.

Be absolutely certain those cards are as loud as they'll go in your test before you start recording. Also make sure you’re getting a 95% (or higher) GPU load from the game.

 

[apoppin]

Oh sure, fan ramping is fine, just be sure you get the main bit correct.

Be absolutely certain those cards are as loud as they'll go in your test before you start recording. Also make sure you’re getting a 95% (or higher) GPU load from the game.

OK then, is everyone agreed?
- i record about the first 15 seconds of the Crysis GPU demo, and then the last 15 seconds or so of the demo after it has been running for 15-20 minutes - each card in identical testing situations. PC on my desk right next to my LCD at ear height and about 3" away from the mic where my head would be ,,,

Crysis GPU demo at 2560x1600 , very high, will max out any GPU; it might even shut down the lesser cards to a complete slideshow where it barely completes one run.

Finally, do we really need *S-i-X* video cards to compare
= that is 2 hours of endless incessant whining i have to endure while i should be sleeping
.. at least my room will be warm in the wee hours of the morning

 

[BFG10K]

I think the two cards people mainly want to see is the 5870 and the GTX480. If you want to add more, by all means do so.

 

[apoppin]

I think the two cards people mainly want to see is the 5870 and the GTX480. If you want to add more, by all means do so.

OK

i think we are all agreed on the methodology of the tests. it seems quite sound

i need to get busy and set this up .. i will see you guys later
- with LUCK .. i will get it tonight; for sure i will get it by Sunday night

Catch you later; thanks for all the help!
- i got to get back to work

 

[BenSkywalker]

I think the two cards people mainly want to see is the 5870 and the GTX480. If you want to add more, by all means do so.

In terms of the enthusiasts perhaps, as someone using the review as a tool to make an informed choice it would be very nice to have a couple of other random cards that could be used as a point of reference. It isn't likely that too many people that own one of those boards is honestly cross shopping the other(not impossible, but it certainly won't be the norm).

 

[blanketyblank]

One other consideration if this is 15 minutes is the order of the test if there isn't a good break in between. Any card being tested later on could potentially be disadvantaged by the heat from the previous test. Especially if as you say your room is actually heating up (not sure if this is an exaggeration or not).

 

[BFG10K]

One other consideration if this is 15 minutes is the order of the test if there isn't a good break in between. Any card being tested later on could potentially be disadvantaged by the heat from the previous test. Especially if as you say your room is actually heating up (not sure if this is an exaggeration or not).

He’ll open his case when he swaps the cards anyway, so any trapped air will escape. While he does this, he can open the doors and windows to his room too.

 

[apoppin]

it's 3:20 AM and dead silent - except for the PC looping Crysis with HD 5870
- neither the PowerColor PCS+ nor the Diamond reference version really "spin up" so that you really notice. After 20 minutes of looping Crysis (25x16, 8xMSAA, very high) - at 10 FPS - they are still really really quiet and the case temps are not so high.

i am giving it a little break between cards so the temperature delta between inside and outside the case is not so great.

i got the 2 Radeons finished, -- i was going to test 4870 but it is an ASUS Dark Knight non reference version. i will skip it and head for HD 4870-X2 ... by then it will be after 4 AM when i am done and i *might* get to the GTX 480 this AM ... i don't know if i can get to GTX 280 ... and i might have to finish really early Mon morning

Refrigerator is off, gas pilots off, all fans off and the dogs are outside. i am using my hi-def HR 1080i video cam to capture the sound as well as you can see what is happening on screen. With luck, i will have it up on Monday.

 

[Larries]

You probably already did it, but in case you forgot, remember to mark the room temperature too (in case the temerature changed a lot due to weather).

 

[ViRGE]

it's 3:20 AM and dead silent - except for the PC looping Crysis with HD 5870
- neither the PowerColor PCS+ nor the Diamond reference version really "spin up" so that you really notice. After 20 minutes of looping Crysis (25x16, 8xMSAA, very high) - at 10 FPS - they are still really really quiet and the case temps are not so high.

i am giving it a little break between cards so the temperature delta between inside and outside the case is not so great.

i got the 2 Radeons finished, -- i was going to test 4870 but it is an ASUS Dark Knight non reference version. i will skip it and head for HD 4870-X2 ... by then it will be after 4 AM when i am done and i *might* get to the GTX 480 this AM ... i don't know if i can get to GTX 280 ... and i might have to finish really early Mon morning

Refrigerator is off, gas pilots off, all fans off and the dogs are outside. i am using my hi-def HR 1080i video cam to capture the sound as well as you can see what is happening on screen. With luck, i will have it up on Monday.

Isn't Crysis VRAM limited at 2560x1600 at 4x MSAA? I'm thinking you'd be better off at 1920x1200 at 4x MSAA so that you can dodge that scenario.