discussion NVIDIA GTX 760 Review Comment Thread

[Sushi Warrior]

Testing at max OC speeds is ridiculous without points of reference (ie: stock clocks). Every sample OCs differently which means there will be a massive statistical variance....something that shouldn't be part of any review.

Linus is pretty clear that he doesn't expect to be a "standard" review, but rather provide something different than the dozens of stock clock reviews. No doubt that different people could get WAY different results. But what that test does show is Nvidia artificially limits the ability of their cards, while AMD gives them quite a bit more freedom to OC.

 

[AkG]

No what Linus does is needlessly muddy the waters. Every card will OC differently. So lets say he gets a lemon tier 1 GPU from either camp and a golden grade tier 2 from the other. Will his results (showing the tier 2 eating the tier 1's lunch) be accurate? Nope. He is doing a disservice to his readers. Want to include OC numbers...cool. BUT include apples to apples as well!

IMHO Linus should stick to unboxing videos. Its what he is good at (ie selling the 'item of the day' for his NCIX masters).

 

[SKYMTL]

Linus is pretty clear that he doesn't expect to be a "standard" review, but rather provide something different than the dozens of stock clock reviews. No doubt that different people could get WAY different results. But what that test does show is Nvidia artificially limits the ability of their cards, while AMD gives them quite a bit more freedom to OC.

Last I checked my average AMD card OC was about 13%. NVIDIA's cards average 12% sustained Boost. And that's from looking at the numbers from all of the roundups, launch day reviews, etc. that I have done.

I have no idea where people get this "AMD overclocks better" statement from.

 

[great_big_abyss]

Last I checked my average AMD card OC was about 13%. NVIDIA's cards average 12% sustained Boost. And that's from looking at the numbers from all of the roundups, launch day reviews, etc. that I have done.

I have no idea where people get this "AMD overclocks better" statement from.

Is that without increasing the voltages on the AMD cards? I suppose some people would argue that the AMD's overclock better because they can up the voltages. But, if you ask me, when you start upping the voltage on a graphics card, you're leaving the realm of 'free' overclocks and you're starting to enter dangerous territory.

 

[Sushi Warrior]

Last I checked my average AMD card OC was about 13%. NVIDIA's cards average 12% sustained Boost. And that's from looking at the numbers from all of the roundups, launch day reviews, etc. that I have done.

I have no idea where people get this "AMD overclocks better" statement from.

1. AMD can up Powertune to 120% instead of 106% on many Nvidia cards.
2. AMD has many voltage-unlocked cards (and AFAIK, it's just up to the AIB to lock/unlock it)

Considering that my 7950 hits 1200 core with an extra +50mv, and that there are pages and pages of people who have done the same, it's not that "AMD overclocks better" either, it's that Nvidia overclocks fine (for the average user, terrible for the enthusiast), 7950 OC's like a beast.

No what Linus does is needlessly muddy the waters. Every card will OC differently. So lets say he gets a lemon tier 1 GPU from either camp and a golden grade tier 2 from the other. Will his results (showing the tier 2 eating the tier 1's lunch) be accurate? Nope. He is doing a disservice to his readers. Want to include OC numbers...cool. BUT include apples to apples as well!

Personally I find the "real world" results like this pretty eye opening. I'm sure you could take 10 cards, average the results, and still have results very similar to what Linus gets.

 

[AkG]

Once again. Not necessarily. You can say oh take ten and the results will be the same...and actually DOING that. Hell is he is not even doing it on ten (and TBH anything under 300 is statistically irrelevant) he is using ONE. Anomalies will play merry hob with his results (trust me I speak for experience on this). Thus his numbers are about as realistic as taking darts and throwing them at a dartboard.

How many per page and how many pages. Lets say 20 and 20 for 400 people out of how many hundreds of K (or millions) SOLD? Could be that its just the best of the best being represented? Sky has done a lot of reviews. All of them to the same standard. Thus if he says its 1% difference its closer to 1% than ten or twenty (ie the amount some peeps are inferring). Are the '50s beasts...yes. But are ALL AMDs beasts? Nope. Thus the quick comment (I assume) by Sky.

 

[bliz]

Is that without increasing the voltages on the AMD cards? I suppose some people would argue that the AMD's overclock better because they can up the voltages. But, if you ask me, when you start upping the voltage on a graphics card, you're leaving the realm of 'free' overclocks and you're starting to enter dangerous territory.

Dude you pretty much can't overclock a card w/o raising voltage,i could not even overclock mine +20mv without raising voltage...Because the voltage is set for the actual core clock of the card,so...

and boosting the voltage a little bit,less than 100mv is not really dangerous imo,the card will still have chances to last 8 years,so...Just have to be careful.

 

[Sushi Warrior]

Once again. Not necessarily. You can say oh take ten and the results will be the same...and actually DOING that. Hell is he is not even doing it on ten (and TBH anything under 300 is statistically irrelevant) he is using ONE. Anomalies will play merry hob with his results (trust me I speak for experience on this). Thus his numbers are about as realistic as taking darts and throwing them at a dartboard.

How many per page and how many pages. Lets say 20 and 20 for 400 people out of how many hundreds of K (or millions) SOLD? Could be that its just the best of the best being represented? Sky has done a lot of reviews. All of them to the same standard. Thus if he says its 1% difference its closer to 1% than ten or twenty (ie the amount some peeps are inferring). Are the '50s beasts...yes. But are ALL AMDs beasts? Nope. Thus the quick comment (I assume) by Sky.

I know he isn't using 10, no reviewers I know of use 10, but I am saying based on observation it can be seen than the 7950 overclocks very well, and the 760 doesn't appear to be as extraordinary. Anomalies make a difference, but when you consider the fact that the "silicon lottery" is as likely to give you a terrible card as it is a bad card, and that both those cases are much less likely than an "average" card, you can see that by picking 1 sample of each, it's likely he got an average model of each, and extremely improbable he got a good 7950 and a bad 760 (and 660).

Yes, the thread is 23 pages * 20 posts, cutting out random posts (lets say 50%) then that is 230 users. Now, what you may notice is the thread isn't just "Who got awesome results", but rather "what were your results, good or bad". Given that, I read through the first 3 pages and I have yet to find anyone reporting an OC of less than 1100 core (which is what Linus used). So, his results are on the low side for a 7950 (for reference, most posters were hitting 1200 core with ~1.2-1.25v). All that I can gather from this is everyone who tried to overclock their card got at LEAST 1100 core out of it, out of approx. 20-30 users. I would say that's a large enough sample to notice a trend!

 

[AkG]

Oh yes I 100% agree the '50 specifically OCs well and the 760 is a ?? right now. But 230 cards out of how many cards is my point. And like I said I think they are good OC'ers but I wouldnt judge the entire lineup based on that one model

Plus Linus is not even trying to be unbiased. He is doing is level best to provide skewed results (to make whatever card he is reviewing RIGHT THEN look good is open to debate. I have no idea. I dont trust him and havent since he tried to convince ME that a OCZ Solid was just as good as Vertex...and the difference was only 'marketing'...because a sale on the Solids was starting in two days and he didnt want bad advertising to hurt sales). Give apples to apples or GTFO. Its not rock science. You want to be different...awesome. Just dont be a jackass in how you do 'different'.

 

[Sushi Warrior]

Oh yes I 100% agree the '50 specifically OCs well and the 760 is a ?? right now. But 230 cards out of how many cards is my point. And like I said I think they are good OC'ers but I wouldnt judge the entire lineup based on that one model

Plus Linus is not even trying to be unbiased. He is doing is level best to provide skewed results (to make whatever card he is reviewing RIGHT THEN look good is open to debate. I have no idea. I dont trust him and havent since he tried to convince ME that a OCZ Solid was just as good as Vertex...and the difference was only 'marketing'...because a sale on the Solids was starting in two days and he didnt want bad advertising to hurt sales). Give apples to apples or GTFO. Its not rock science. You want to be different...awesome. Just dont be a jackass in how you do 'different'.

All I can vouch for is the 7950 does OC better than most from my experience and what I have seen, and I can hypothesize that AMD gives the user more freedom to OC than Nvidia. More freedom =/= better overclocks though, because it doesn't matter if the slider is locked if you can't hit that max limit! :doh:

Linus definitely seems like shill city (as if the Intel ad at the start of the video didn't give it away :shok. TBH, he didn't really even mention the 7950 beating it through the video, but just focused on Nvidia vs. Nvidia performance...

While I agree apples to apples is most truthful, a review like his makes for an interesting read/listen/watch when there are hundreds of apples to apples reviews!