GTX480 Vs. HD5870 Crysis benchmark

[MagickMan]

Stable at 1300 doesn't mean that there isn't a loss in minimum framerates. You can clock the memory on the 5800 series as high as the slider will go, it's error correction will compensate, but it will have a negative impact.

 

[HurleyBird]

Apoppin:

For the LAST time, I AM NOT talking about *VRAM* overheating and losing performance to error correction. I am theorizing that the *VRMs* could be overheating, which would cause the CORE (*not* the memory, that's COMPLETELY different) to throttle.

If you still can't grasp that, maybe this will help:

VOLTAGE REGULATOR MODULE != VIDEO RANDOM ACCESS MEMORY

Again, I'll lay it out for you. When the VRMs on a 5870 overheat it causes the card to throttle, which reduces performance.

Granted this is only a theory, and it could just as easily be proven wrong given the methodology I've described twice now. The simple fact of the matter is that when you apply a lot of voltage like you did to obtain your overclock, VRMs need to work harder and they get hotter. Any game that loads the 5870 to a significant degree is then more likely to exhibit throttling, just like furmark already does at stock clocks.

I don't know if you've been beating around the bush, or you simply don't know the difference between VRAM and a VRM, but either scenario isn't good if you want people to take you as a serious reviewer. Sorry for the all caps, but dude, you've exhausted the ways I have to get through to you.

Again, please don't come back and respond that you've tested the memory at 1150 and 1200MHz. That is not what I am talking about

 

[apoppin]

Stable at 1300 doesn't mean that there isn't a loss in minimum framerates. You can clock the memory on the 5800 series as high as the slider will go, it's error correction will compensate, but it will have a negative impact.

You guys are hilarious and should be a comedy team.

Throw out my minimums and look only at the averages. Feel free to add +1 FPS to every HD 5870 score well, just because .. because .. you "think so"

and what happens? ..

My chart with 14 games is still valid and the ranking remains the same except in Crysis where they are SO close ... and the conclusion is still the same

And to the somewhat broken AI that just cannot grasp that my vRAM is stable and not overheating my core, all i can say is that i got consistent (if tiny) performance gain with my HD 5870 clocked at 975/1300 MHz over when it is clocked at 975/1250 MHz (and below).

 

[Cookie Monster]

Since when did video cards worry about ECC?

 

[Voo]

There's a difference between error correction on the memory, which is not the same as throttling, and the VRMs overheating and throttling the shader units, like they already are known to do with fur mark. It seems like a likely scenario based on how little scaling Apoppin found in most of his games, and the fact that he had to give the card a lot of voltage to get his overclock.

So you say that his lesser overclocks (I assume he didn't up his voltage for the lower ones), were magically also throtteled back and that's why he's getting a steady, but small performance gain? If the card throttles the shaders I assume that would be a noticeable performance hit (and not just staying the same - but even if that was the case.. it'd be still faster than the lower oc, so not a problem, as long as you don't have a problem with a early dead card), so how exactly should that work out with his data?

@ Cookie Monster: Since GDDR5.

 

[SRoode]

Absolutely not. Utter nonsense.

Both of my HD 5870s have no problem with 1300MHz on the memory; it is just as stable at 1200 MHz as it is at 1300 MHz

i test each clock independently of the other. When my diamond HD 5870 is at 975/1300 it is just as stable as it was at 800/1200

There is no min frame rate being killed. i overclocked it to its absolute limit on air cooling but i did not exceed its limits.

From what you are saying, i don't think some have a real clue about finding a truly stable overclock

I have never said anything nasty to you, so why you choose to take this tone with me is really confusing. Anyway, you are incorrect. Although stable at higher clocks, at least on the HD5850, clocks that are too high does cause the min frame rate to drop off the table. You can check it yourself if you do not believe me, although others have backed my statement.

I'm really quite disappointed in your tone with me, and I think for now I have nothing left to say to you (since I don't know what I'm talking about anyway). Good luck with your endeavors.

 

[HurleyBird]

So you say that his lesser overclocks (I assume he didn't up his voltage for the lower ones), were magically also throtteled back and that's why he's getting a steady, but small performance gain?

I've only seen him mentioning testing 1250 and 1300 on memory. I don't think that would have a significant change to load on the VRMs like changing core clock speed would.

 

[v8envy]

The throttling theory has lots of merit. Let's think about the easiest way to safeguard against VRMs burning up (VRMs, not VRAM). A trivial feedback loop could have the GPU idling for 1 cycle out of 100 if the VRMs are at 90C. Then 2 cycles out of 100 at 100C, 5 out of 110 at 90 and all the way down to 99 idle cycles at 140C.

So the % of idle cycles goes up at the same time as the number of cycles also goes up. Resulting in a small but steady gain which would be less than the gain with no thermal throttling present.

There have been plenty of weird performance anomalies reported on ATI cards of this and previous generation attributed to insufficient or unstable power. GPU throttling due to VRM thermals was documented on the 4 series by this very site. It wouldn't surprise me in the least if this was the case here.

 

[Patrick Wolf]

I'd like to see GTX480 vs. 5830 CF since two of them can had for $500 or less atm.

 

[ViRGE]

Since when did video cards worry about ECC?

Ever since doing 4 transfers on a GHz+ bus. It's hard to make a bus like that behave.