GTX480 Vs. HD5870 Crysis benchmark

[Tempered81]

Apoppin, your 5870 is worse than I thought, that may be the worst voltage for 975MHz that I've seen.

Could be possible, I was actually thinkin the same thing earlier. Perhaps he just has the worst sample around. I did a poll of 5870's at XS, out of 73 5870's the average OC was like ~1020 @ 1.29v. 975 @ 1.35v is crappy, but def possible. Especially if he's aming for stability across a whole benchmark suite.

 

[dust]

Apoppin, your 5870 is worse than I thought, that may be the worst voltage for 975MHz that I've seen.

Cherry picked?

 

[MagickMan]

Cherry picked?

Cherry picked for optimal shittiness?

Could be possible, I was actually thinkin the same thing earlier. Perhaps he just has the worst sample around. I did a poll of 5870's at XS, out of 73 5870's the average OC was like ~1020 @ 1.29v. 975 @ 1.35v is crappy, but def possible. Especially if he's aming for stability across a whole benchmark suite.

Out of the 12ish 58x0 cards that I've worked with, none have been that bad. There was one Diamond 5870 that needed 1.32v to hit 1GHz, 100% stable with everything, but that was it. In fact, as far as core is concerned, I've seen better clocks at lower voltage from 5850s. Which kinda makes sense.

 

[3DVagabond]

Minimum and maximum frame rates are nearly worthless measurements without graphs at the very least.

Originally Posted by MagickMan
Apoppin, your 5870 is worse than I thought, that may be the worst voltage for 975MHz that I've seen.

Things that make you go Hmmm?

 

[MagickMan]

Things that make you go Hmmm?

It sure is odd. I've been searching other results for all 58x0 overclocks, and I've not found a single one that needed that much voltage for a midrange result. No wonder he's getting those temps and noise. It would be better if he had an average card to work with, something that would mirror what most people could actually expect.

 

[dust]

Well, my math could be a little off but it seems that there is a 17,85% increase for the core and a 19% mem clock for the GTX and only 14,7% core with a 8,33% memory clock increase for the Radeon.

We've already seen benches at stock showing the 480 lead, I find it only normal for the difference to increase if he keeps the above figures.

 

[MagickMan]

I was specifically talking about heat and noise, an average 5870 wouldn't be as loud or as hot as his at those speeds.

 

[apoppin]

My performance scaling results sure won't be any difference than any other HD 5870 at 975/1300. And i am certain there was complete 100% stability across every demanding benchmark i could give it including maxed out details at 2560x1600. i am sure i did not have to raise the fan speed as far as i did.

My GTX 480 scaled just as well as two others i know of; actually mine was also middling there also in comparison - and i did not touch the GTX voltage and i did not overclock its shaders.

i am done and it is finally time for bed (2:30 AM)

i feel bad that i was unable to enjoy a really quiet Radeon at its max OC; too bad about that Diamond HD 5870. The PCS+ is very quiet but has no voltage adjustment.

No matter what, IF i was going to really OC either video card, i would consider watercooling for quiet and peace of mind.

 

[3DVagabond]

You seem to be missing the point. Which is, the 5870 cards you have are not representative of what most people are getting. You must be just terribly unlucky to get two subpar 5870 or you need to change suppliers.

 

[Dribble]

You seem to be missing the point. Which is, the 5870 cards you have are not representative of what most people are getting. You must be just terribly unlucky to get two subpar 5870 or you need to change suppliers.

*most people* - is that an accurate study, or just "I read a few reviews, and a few peoples comments in forums and saw what I wanted to see"?

People in forums lie - just because they say their card is "stable" doesn't mean it really is? Equally the massive majority with lesser overclocks aren't going to be waving their epeen around in quite the same way. Reviewers often get cherry picked cards (there is probably only a few reviewer cards supplied direct by the manufacturer that get passed around most sites), and even then I doubt they would really run the cards at some of the settings they do and find they are stable in the longer term.

I expect appopin is actually pretty average.

 

[apoppin]

You seem to be missing the point. Which is, the 5870 cards you have are not representative of what most people are getting. You must be just terribly unlucky to get two subpar 5870 or you need to change suppliers.

Ridiculous.

The PC PCS+ came directly from AMD
- how do i change that supplier?

- i bought my Diamond reference HD 5870 from NewEgg less than a month ago

What "most people" ?

. . . do you mean, e-peen fanboy people who boot into windows at 1030/1335 with HD 5870 and call it 'stable' on a forum and then later complain because their warranty gets denied?


i got +125 MHz OC on the 5870 core; it is a decent overclock and you can see HD 5870 scaling with clock speed; 1300MHz on the RAM looks pretty good; would +25 MHz on DDR5 make a perceptible difference IF i could achieve it?

Do you really thing ANYTHING would be different with +25 MHz on the core - HD 5870 at 1000/1300 vs. GTX 480 at 825/1100?

Do you think the ranking would change much with 1025/1300? - if not, how far should i go to catch the overclocked GTX 480?
... 850 > 875 > 975 >1000 > 1025 .. do you really think 1025 MHz is "representative" of what "everyone" gets on air cooling (at 1.35V) ?

 

[badb0y]

*most people* - is that an accurate study, or just "I read a few reviews, and a few peoples comments in forums and saw what I wanted to see"?

People in forums lie - just because they say their card is "stable" doesn't mean it really is? Equally the massive majority with lesser overclocks aren't going to be waving their epeen around in quite the same way. Reviewers often get cherry picked cards (there is probably only a few reviewer cards supplied direct by the manufacturer that get passed around most sites), and even then I doubt they would really run the cards at some of the settings they do and find they are stable in the longer term.

I expect appopin is actually pretty average.

So your refuting his comment by stating people are liars?

Genius.

 

[MagickMan]

No, it isn't "average", it's extraordinary (extraordinarily bad). That's quite a bit more voltage needed for 975MHz than I've seen from other 58x0s, and I myself have worked with more than a dozen from different companies. Not to mention, I'm looking at results from all over the web. Yeah, there's no doubt that some OC reports are fudged, but not all of them. If the average card needed 1.35v to be stable @975 there would be an uproar. From what I've been able to compile, from personal experiences and online sources, the average juice required for that speed is 1.25v.

That much extra voltage is going to add a lot more heat, which in turn will make the card louder due to the fan ramping up higher to compensate.

 

[apoppin]

No, it isn't "average", it's extraordinary (extraordinarily bad). That's quite a bit more voltage needed for 975MHz than I've seen from other 58x0s, and I myself have worked with more than a dozen from different companies. Not to mention, I'm looking at results from all over the web. Yeah, there's no doubt that some OC reports are fudged, but not all of them. If the average card needed 1.35v to be stable @975 there would be an uproar. From what I've been able to compile, from personal experiences and online sources, the average juice required for that speed is 1.25v.

That much extra voltage is going to add a lot more heat, which in turn will make the card louder due to the fan ramping up higher to compensate.

Let me put it another way:

Does it make ANY difference to the performance numbers to running a 'hot-running' HD 5870 vs. a "cool" 5870 at the *same clock* speeds - where both are 100% stable?


As i said before, "too bad" i got a card that couldn't get 1000MHz on the core with 1.35V - but it does not invalidate ANY benchmarking i did with it because it is 'loud'.

 

[Madcatatlas]

I just had to take the time to sift trough some 5870 OC reviews and found out that at least 4 out of the 5 i checked couldnt even turn up the volt.. (edit. lol they did try it with the MSI afterburner, but it wouldnt allow volt upping, my bad )

The one out of 5 reviews i did see go higher than 950 on the core, was overclockersclub , and im going to copy paste the following:



Overclocked settings:

• PowerColor HD 5870 PCS+ 950/1320

960 MHz without a voltage increase was very nice, the card pushed higher but wouldn't achieve complete stability. It went as high as 1000 MHz and 1400 MHz but was not stable enough for testing so it got cut back a bit to ensure stability. Oddly, the memory was not very willing to overclock at first and wouldn't go very high. After a few days of pushing, the card was more willing and hit 1320 MHz. Nearly 10% and 8% gains on both core and memory respectively just with ATI Overdrive - not to mention that the card is already overclocked 25 MHz on both the core and memory from factory. Voltage programs such as [COLOR=#289e28 !important][COLOR=#289e28 !important]MSI[/COLOR][/COLOR] Afterburner were unable to adjust the voltage and thus the card had to do with what it was set with. Temperatures were great, as the highest temperatures loaded with benchmarking and gaming were low 50's and high 40's Celsius anywhere on the card and at full speed the fan is definitely audible but not ear piercing. Next are the benchmark results!


960 without touching the volt, must have been really lucky with this card, seeing Apoppins needed volt upping to be stable.

edit: im also curious to see how 480s overclock in general. The coming weeks will be interesting.

 

[MrK6]

No, it isn't "average", it's extraordinary (extraordinarily bad). That's quite a bit more voltage needed for 975MHz than I've seen from other 58x0s, and I myself have worked with more than a dozen from different companies. Not to mention, I'm looking at results from all over the web. Yeah, there's no doubt that some OC reports are fudged, but not all of them. If the average card needed 1.35v to be stable @975 there would be an uproar. From what I've been able to compile, from personal experiences and online sources, the average juice required for that speed is 1.25v.

That much extra voltage is going to add a lot more heat, which in turn will make the card louder due to the fan ramping up higher to compensate.

Two explanations really - either apoppin has a horrible 5870 (which belittles his overclocking tests, shame he wasted the time) or he has no idea what he's doing when it comes to overclocking (which belittles everything he's doing). From XS: http://www.xtremesystems.org/forums/showthread.php?t=235693 . The average 5870 overclock is 1027/1286 @ 1.27V, but if you look at the stock cooling solutions specifically, you'll see 1000MHz on the core requires an average of 1.25-1.275V. Sounds to me like he overvolted the core past the point of stability, great fail.

Do you think the ranking would change much with 1025/1300? - if not, how far should i go to catch the overclocked GTX 480?
... 850 > 875 > 975 >1000 > 1025 .. do you really think 1025 MHz is "representative" of what "everyone" gets on air cooling (at 1.35V) ?

You go as far as you can, not until you feel like it or your incompetency catches up. You're not doing anyone favors by screwing up a relatively simple procedure. Such an insult to the scientific method.

 

[apoppin]

Who gives a crap what *other* tech sites got, Madcatatlas?
- you are linking to an overclocked video card anyway; mine is reference

That reviewer *couldn't up* the voltage - there is NO adjustment


- i got 925 MHz on my PowerColor HD 5870 PCS+ - with no voltage adjustment
-- i guess i was unlucky *twice* - being *this time* -45 MHz lower than the example you sited on the same brand.

 

[SlowSpyder]

Ridiculous.

The PC PCS+ came directly from AMD
- how do i change that supplier?

- i bought my Diamond reference HD 5870 from NewEgg less than a month ago

What "most people" ?

. . . do you mean, e-peen fanboy people who boot into windows at 1030/1335 with HD 5870 and call it 'stable' on a forum and then later complain because their warranty gets denied?


i got +125 MHz OC on the 5870 core; it is a decent overclock and you can see HD 5870 scaling with clock speed; 1300MHz on the RAM looks pretty good; would +25 MHz on DDR5 make a perceptible difference IF i could achieve it?

Do you really thing ANYTHING would be different with +25 MHz on the core - HD 5870 at 1000/1300 vs. GTX 480 at 825/1100?

Do you think the ranking would change much with 1025/1300? - if not, how far should i go to catch the overclocked GTX 480?
... 850 > 875 > 975 >1000 > 1025 .. do you really think 1025 MHz is "representative" of what "everyone" gets on air cooling (at 1.35V) ?

For whatever it's worth, I have a 965MHz OC (yes, stable in all benches I've run) on my VaporX card. It has an improved cooler, but I cannot adjust voltages. So, 965MHz on stock voltage is pretty impressive considering you need a big bump to get 10MHz more than my card.

 

[apoppin]

You go as far as you can

that IS what i did

 

[MrK6]

that IS what i did

Then make sure you add a disclaimer to your review that you're an amateur overclocker and have little experience or results in getting the most out of your hardware.

 

[Madcatatlas]

Personally i think your overclock is all good. Magickmans argument of that much voltage resulting in more heat and therefore more noise from the cooler looks valid though, doesnt it?

But you have already said the noise levels are not "high" if i remember the other topic we talked about the article your doing, so rock on and give us something to read!

 

[dook43]

I needed 1.2V to get 850/1100 stable on my XFX 5850. 875/1100 artifacted, 850/1200 artifacted. Wasn't willing to go past 1.2v. I strongly doubt that bad OCs are that rare on 58x0s. My second 5850 (Sapphire) was perfectly happy at 950 at 1.2v but what's the point of that if both of your cards can't run at that speed?

Having said all that, my 5970BE is perfectly happy at 5870 speeds.

 

[apoppin]

Then make sure you add a disclaimer to your review that you're an amateur overclocker and have little experience or results in getting the most out of your hardware.

Why don't you add a disclaimer to all of you posts that you haven't got a clue what you are posting about?

Personally i think your overclock is all good. Magickmans argument of that much voltage resulting in more heat and therefore more noise from the cooler looks valid though, doesnt it?

But you have already said the noise levels are not "high" if i remember the other topic we talked about the article your doing, so rock on and give us something to read!

Of course, more voltage is going to add more heat and the fan has to be cranked up
- and my point is that it *doesn't matter* to the benchmarking or performance results.

relative noise is not what my review is about. Look it up; i am not allowed to post a link to my reviews on this forum (PM sent)

When i began testing, i pointed out that the HD 5870 is much quieter than the GTX 480 *until* you get the fan up around 90% - then they are both annoying as hell.

 

[MagickMan]

Let me put it another way:

Does it make ANY difference to the performance numbers to running a 'hot-running' HD 5870 vs. a "cool" 5870 at the *same clock* speeds - where both are 100% stable?


As i said before, "too bad" i got a card that couldn't get 1000MHz on the core with 1.35V - but it does not invalidate ANY benchmarking i did with it because it is 'loud'.

and I said that I'm not talking about the performance numbers, I'm talking about how much heat a card at 1.35v/975MHz produces. Your 5870 isn't normal in terms of heat and noise at those speeds, so any thermal and audible analysis of the thing is pointless.

 

[apoppin]

and I said that I'm not talking about the performance numbers, I'm talking about how much heat a card at 1.35v/975MHz produces. Your 5870 isn't normal in terms of heat and noise at those speeds, so any thermal and audible analysis of the thing is pointless.

You have to realize that i am NOT "analyzing thermal and audible" factors of HD 5870
- JUST *performance* of the overclocked HD 5870 at 975/1300 vs. GTX 480 at 825/1100

Maybe the title of my review will explain what it is about:

GTX 480 (800/1100 MHZ) vs. HD 5870 (975/1300 MHz), Overclocked Performance Analysis, Part 2

it is 22 pages long and only a couple pages briefly mention "noise"
- it is all about o/c'd *performance* of each card.