Nvidia misses GTX470/480 launch targets

[cowie]

It says on the box of the card Overvoltage:



LINK

its just in..... afterburner beta4 for 480 volt control 860(1720) np with just jersey fresh air in sli even.
http://forums.guru3d.com/showthread....19#post3539619

this is news

[damha]

Use FRAPS in the in-game benchmark. The results given in the Test Summary screen are not accurate.

Uh oh. Bad news for built in benchmarks. They must all now be verified by FRAPs I suppose.

This means all of the Dirt2 figures we have are garbage.

[xdan]

Some GTX 470, Palit, Msi, Evga and GTX 480-Galaxy on stock on Newegg....
http://www.newegg.com/Product/Produc...20400%20series
GTX 480 dissapeared in 5 minutes from time i saw....

[Dumo]

I wonder if there is some sort of throttle control. GT, have you tried running the 470 at stock frequencies to see if performance improves?

Seems that happened with stock voltage and upped core/shader/mem.
Will be testing wuth AFBurner beta 4

[Eastcoasthandle]

Seems that happened with stock voltage and upped core/shader/mem.
Will be testing wuth AFBurner beta 4

Interesting...I look forward to your results.

[SKYMTL]

Uh oh. Bad news for built in benchmarks. They must all now be verified by FRAPs I suppose.

This means all of the Dirt2 figures we have are garbage.

It depends. I am sure most reviewers (I hope) validate built-in benchmark results with FRAPS before putting them into reviews. Here is what I have seen regarding built-in benchmarks:


Dawn of War II: Inaccurate

DiRT 2: Inaccurate

Far Cry 2: Accurate

HawX: Accurate

Just Cause 2: Accurate but none of the built-in demos reflect gameplay conditions

[trinibwoy]

By innacurate you mean that Fraps disagrees with the reported in-game result for the same demo sequence? Or inaccurate because it differs greatly from performance during actual gameplay?

[GoldenTiger]

Sorry for the late post...

Dirt 2 1920x1200 8x MSAA, maxed Ultra in-game settings stock GTX 470


Seems the overclocking is scaling properly.

Nice find on the Afterburner, hoping for a GTX 470 compatible version soon, since my card looks to have more room with volts .

[Eastcoasthandle]

Hey GT quick question. Was Dirt 2 already installed when you put that 470 in your system? If it was check the hardware_settings_info.xml file and make sure it's reading your PC correctly.

[GoldenTiger]

Hey GT quick question. Was Dirt 2 already installed when you put that 470 in your system? If it was check the hardware_settings_info.xml file and make sure it's reading your PC correctly.

Yes, it was... good catch, I will check into it.

EDIT:

<graphics_card name="NVIDIA GeForce GTX 470" vendorId="0x10DE" deviceId="0x06CD" vs_shader="3_0" ps_shader="3_0" memory="4059" rating="3" dx11="true">

Wonder if that bolded part is causing an issue... gonna adjust it and run again.

EDIT: Set file to read-only after editing, with "1280" as the memory value, same results.

[SKYMTL]

By innacurate you mean that Fraps disagrees with the reported in-game result for the same demo sequence? Or inaccurate because it differs greatly from performance during actual gameplay?

Innacurate means that the included, in-game benchmark results are out to lunch compared to what FRAPs reports. In the case of DiRT 2, it seems to be reading the loading screen and the "lead up" to the race as part of the benchmark sequence.

[Eastcoasthandle]

Innacurate means that the included, in-game benchmark results are out to lunch compared to what FRAPs reports. In the case of DiRT 2, it seems to be reading the loading screen and the "lead up" to the race as part of the benchmark sequence.

In that case it's not what fraps is reporting but when you decide to start fraps. So I would take what you say with a grain of salt. I just checked myself and there are no "...loading screen leading and the lead up to the race..." once the benchmark starts. There are however, "cutscreen(s)" just before the race begins. So I'm not sure what you going on about.

Therefore, regardless when Dirt 2 starts/ends their benchmark testing if it's the same for everyone those results reflect how CM wants their benchmark to run. Which should reflect any video card used. Making them valid from my point of view. This is because the benchmark is similar to what you see in game.
For example:
-game loads track
-cutscene(s)
-start race
with the exception of tweaking your car screen.

[Chickenfeed]

I tested that myself and as far as I can tell, the in game bench records starting from the moment it loads until the end of the race. I compared with FRAPS and in all honesty found the in game to FRAPS results to be pretty consistent. However the benchmark itself varries alot as your car seems to varry from last place ( highest strain due to having to render the other 7 cars and associated particle effects ) to a best case of 5th place ( this was after 25 runs ) What I've found is the average doesn't change much but the max / min can go up and down. For example after more extensive testing im getting 75-79 avg but the min can be from 55 on the low side ( trailing the race in 8th ) up to as high as 68 or so ( finishing 5th so a reduced rendering load ) I would start fraps the moment the map loaded and finish it once the player car goes across the finish line. Regardless of how you decide to bench this game, due to the AI element involved I'd disregard any results that aren't from a healthy average ( I'd consider a good 5 or so to be adequete, 3 however not so much in this case )

Anyways on the topic of the launch, a few cards trickled into Canada but they seemed to have been all preordered and were out of stock instantly. I find it odd how retailers here are asking only 10cnd more for the 470 but are asking 50 more for the 480... Most 5870s go for 440-460 here so at the actual msrp (which many seemed to have been able to preorder in the states ) that would have been a good deal. And our dollar is doing so good too! Don't you just love free trade....

[GoldenTiger]

I feel like a noob... I was forced to reinstall drivers (which should have been a first course of action when I saw Dirt 2 was performing oddly) when I changed things around to use the Asus voltage tweak through SmartDoctor on my card. At 1.012v and 800core/3650mem stable, I re-ran the Dirt 2 bench (same settings as before):



Much more like you'd expect... my previous driver install clearly had something borked with it!

GTX 470 vanilla with voltage tweaked:

(Edited to add info to pic)

[Chickenfeed]

Much much better Nearly exactly what im seeing on my 5870. Good stuff and good to see how much higher it got with a bit more voltage... That might still be enough of a game changer as It will most likely match or exceed the 5870 in places it was behind before hand. Any comment on noise / heat increases in doing so ( I find left 4 dead to be a great test for power consumption / noise levels )

[annihilat0r]

how hot is a voltage tweaked 470?

[GoldenTiger]

Thanks Chicken... I'm quite pleased with the performance/overclock now .

I am running 73% fan which is the last step before it becomes "loud" in my opinion (75% and higher become noticeable). It's able to be heard @ 73% but not loud to me, and vanishes completely with music or a game on very low volume. Temps are peaking at about 86c after an hour+ of Battlefield Bad Company 2 (maxed, DX11, 2560x1600, 4x MSAA, 16x AF), having held steady at that once reaching it (800/3650 1.012v). Comparatively, @ the lower clocks (720/3520 @ stock volts), temperatures peaked at 82c and held mostly at 81, with a 71% fan setting. It's a minor tradeoff, but it is definitely worth it IMO .

[GoldenTiger]

New result for 3dmark with the higher clocks .

[Eastcoasthandle]

Nice results GT. However, at over 70% fan speed that's just to much/to loud for me.

[damha]

Thanks GoldenTiger. Definitely a great alternative to the 5850, but they gotta bring the price to match it.

[tajoh111]

Thanks GoldenTiger. Definitely a great alternative to the 5850, but they gotta bring the price to match it.

This card with these type of OC is more inline with a 5870 OC.

20% OC without volts 32% with volts. These cards are exactly what xtremesystems is about, that a high performing card that like to overclock.

[annihilat0r]

Thanks GoldenTiger. Definitely a great alternative to the 5850, but they gotta bring the price to match it.

From users around the world it looks like Fermi's can be better overclockers than Cypresses lol, who would have thought.

I think $350/$500 are fair price points for GF100's.

[-Sweeper_]

Thanks Chicken... I'm quite pleased with the performance/overclock now .

I am running 73% fan which is the last step before it becomes "loud" in my opinion (75% and higher become noticeable). It's able to be heard @ 73% but not loud to me, and vanishes completely with music or a game on very low volume. Temps are peaking at about 86c after an hour+ of Battlefield Bad Company 2 (maxed, DX11, 2560x1600, 4x MSAA, 16x AF), having held steady at that once reaching it (800/3650 1.012v). Comparatively, @ the lower clocks (720/3520 @ stock volts), temperatures peaked at 82c and held mostly at 81, with a 71% fan setting. It's a minor tradeoff, but it is definitely worth it IMO .

whats the room temperature?

[Chickenfeed]

You mention the fan speed you are running but can you make any subjective comparison as to how loud the card was in reference to your experiance with a 5870? I find the 5870s cooler to become overly loud past 40% myself ( and I prefer 33% but once OCd the auto profile will often go to 37% ) so I'm just wondering how loud the 470 is with that in mind.

[kemo]

You mention the fan speed you are running but can you make any subjective comparison as to how loud the card was in reference to your experiance with a 5870? I find the 5870s cooler to become overly loud past 40% myself ( and I prefer 33% but once OCd the auto profile will often go to 37% ) so I'm just wondering how loud the 470 is with that in mind.

may be you mean that you start hearing the sound of the fan over the other fans in the system coz at 40% you wont hear it while turning your speakers on while gaming