HD 6870 & HD 6850 vs. GTX 460 1GB: An Overclocking Study Comment Thread

[BrianHanna]

Originally Posted by ChinStrap

Another “first post”. SKYMTL, I follow your work from the posting you make on other forums on the internet. First, GREAT review. Thank you for putting in the time to really, give the community what it wanted. I legitimate overclocked review. Second, why did you choose to exclude (all) Call Of Duty titles? I know even MW2 is a year old now, but COD4 & 5 are really the only two games I play anymore (due to modding community).

my 4870 can max out any COD duty game at 1680x1050, pretty sure it's a joke to any of the newer cards

[heflys20]

Originally Posted by BrianHanna

my 4870 can max out any COD duty game at 1680x1050, pretty sure it's a joke to any of the newer cards

Pretty much..........

HIS Radeon HD 6870 Turbo 1 GB Review - Page 9/32 | techPowerUp

[r9800pro]

This so called "study" is flawed

1-With overclocking, it is always YMMV, a card would not OC to the same frequencies that another card would so overclocking (specially that high) should never be a purchase decision maker

"studies" should be done with constants and overclocking is not constant

2-Using a carefully picked GPU (By EVGA in FTW series) against ordinary GPUs from AMD !

3-There are still no software to show the full potential of HD6800 cards (correct me if I am wrong)

I for one had tested Asus HD6850 DirectCU and Oced it to 1030/1200 with stock 1.15v voltage but I mentioned clearly that the mileage may vary, so why didn't you say the same thing ?

[r9800pro]

I had to put pics in a secind post






4-And here is the last straw

From your review :



And here is my finding with the built in benchmark :

[Sagath]

Originally Posted by r9800pro

This so called "study" is flawed

1-With overclocking, it is always YMMV, a card would not OC to the same frequencies that another card would so overclocking (specially that high) should never be a purchase decision maker

"studies" should be done with constants and overclocking is not constant

2-Using a carefully picked GPU (By EVGA in FTW series) against ordinary GPUs from AMD !

3-There are still no software to show the full potential of HD6800 cards (correct me if I am wrong)

I for one had tested Asus HD6850 DirectCU and Oced it to 1030/1200 with stock 1.15v voltage but I mentioned clearly that the mileage may vary, so why didn't you say the same thing ?

Wow. Did you read the article at all? He addresses all your points and more in the original article, plus this follow up article.

And as to your 'built in demo' point, HWC choses not to use 'catered to company X' benchmarks, but their own custom run-throughs, to help minimize any biasing towards company X because its 'the way it was meant to be played' as seen in these two 'clarity for all' posts:
GPU Benchmarking Methods Investigated: Fact vs. Fiction
Your Guide to the Hardware Canucks GPU Benchmarking Process

Either that, or maybe we should just throw up Vantage scores and call it a day if you want to cherry pick numbers...

Maybe before coming to a site and trashing the hell out of it as a new member you should do some easy searching, reading, and then start name calling. Hell, you didnt even use a 1 gig, you used a 768 meg!

Seriously,

[sswilson]

Originally Posted by Sagath

Seriously,

That will depend entirely on how he responds to your post. If he responds reasonably to the points you've made, then the answer will probably be no. If OTOH he starts frothing at the mouth in classic fanboi fashion, it's a definite possiblilty. ;)

[_dangtx_]

800, what are your points again? and wait till the official response from sky.

[benchMark]

Originally Posted by _dangtx_

im sure sky would not only make it his passion to redo all the benches and what not if he didnt use the hotfix (doubt it meself) or make it his top priority.

I believe the overclocking review was done before hotfix 10.10c was released. I noticed the 6850 and 70 non-overclocked results, are identical to the Oct.20 review. I doubt if Michael would delay the review even further, if the hotfix was released after he had already completed the review. It would make more sense to followup with the difference in the two affected games, after his crossfire review.

[r9800pro]

Originally Posted by Sagath

Wow. Did you read the article at all? He addresses all your points and more in the original article, plus this follow up article.

And as to your 'built in demo' point, HWC choses not to use 'catered to company X' benchmarks, but their own custom run-throughs, to help minimize any biasing towards company X because its 'the way it was meant to be played' as seen in these two 'clarity for all' posts:
GPU Benchmarking Methods Investigated: Fact vs. Fiction
Your Guide to the Hardware Canucks GPU Benchmarking Process

Either that, or maybe we should just throw up Vantage scores and call it a day if you want to cherry pick numbers...

Maybe before coming to a site and trashing the hell out of it as a new member you should do some easy searching, reading, and then start name calling. Hell, you didnt even use a 1 gig, you used a 768 meg!

Seriously,

Please show me where he addresses my points in the "study"

I used 768MB GTX460 because that was what I had and I showed you Dirt 2 benchmark he did in the original HD6800 article in which GTX460 768MB was slightly better in average than HD6850 which contradicts with my findings and also with other reviews

These points still not answered yet :

With overclocking, it is always YMMV, a card would not OC to the same frequencies that another card would so overclocking (specially that high) should never be a purchase decision maker
"studies" should be done with constants and overclocking is not constant

Using a carefully picked GPU (By EVGA in FTW series) against ordinary GPUs from AMD !

[SKYMTL]

Originally Posted by benchMark

Is it safe to assume this review didn't include the Catalyst 10.10c Hotfix?
It includes:
Performance optimizations for AMD Radeon™ HD 6870 and AMD Radeon HD 6850:
Aliens versus Predator performance enhancements
Star Craft 2 performance enhancements.

It would be interesting to see how much of a performance boost AVP and SC2 get, if this review was done before hotfix 10.10c was out.

I tested those two games with the Hotfix and neither showed any improvements over the WHQL. From my understanding, the AvP improvements are for the stand alone BENCHMARK and were not applied to the game itself. As for Starcraft, as I said; they didn't show any improvement.

Plus the Hotfix introduced some instability when using Crossfire.

Originally Posted by r9800pro

1-With overclocking, it is always YMMV, a card would not OC to the same frequencies that another card would so overclocking (specially that high) should never be a purchase decision maker

"studies" should be done with constants and overclocking is not constant

I am not saying it should be a decision maker. As with all articles, it is posted for information purposes and will be considered valid to those who want to know how these cards perform when overclocked.

2-Using a carefully picked GPU (By EVGA in FTW series) against ordinary GPUs from AMD !

So? As you will see when looking through forum postings both here and elsewhere; achieving the GTX 460 FTW's clocks with an ordinary GTX 460 at stock voltage is more than possible. Plus, the FTW is priced between the cards and is available at retailers everywhere. How is that not a perfectly valid comparison?

3-There are still no software to show the full potential of HD6800 cards (correct me if I am wrong)

You are incorrect. Sapphire's Trixx can push both cards to 1.2Ghz. However, the increase in heat is too much for most of the heatsinks currently available on board partners' cards.

I for one had tested Asus HD6850 DirectCU and Oced it to 1030/1200 with stock 1.15v voltage but I mentioned clearly that the mileage may vary, so why didn't you say the same thing ?

So a few Mhz on the core make that much of a difference? And you said yourself: YMMV.

Plus, I can tell you that your memory may be clocked at this speed but it isn't actually WORKING at that speed. As I mentioned, the GDDR5 error correct routines kick in and it lowers performance to compensate for errors without a program like GPU Z detecting it.