Sander Sassen after the fact

[Slides]

Not close at all.

Not really true

 

[Smurfie]

Two rats in this game, one that we skewered and grilled, and the other still running free in the fields.

Seriously, Sanders made himself an ass with his crazy drama, but the claims of ATI to refute Sanders weren't entirely true either. If anything, for a good while, ATI's PR simply reminded me of Nvidia's PR during the NV30s days.

Think about it, if I were to call on ATI's devrel now, will I be sure I won't get similar shenanigans in the reply?

 

[Tim]

Not really true

In Halflife 2 the real x1800xt is 35% faster reletively to the GTX than in Sanders "review", in Doom 3 in the Sander score was only 2% off but he had some strange nonexistent performance boost over the x850xt. In Farcry the real scores was 23% higher (still relatively to GTX). In SC:CT the real x1800xt anihilated the Sander x1800xt by being 45% faster.

(all based on the extremetech scores at 1600x1200+8AF&4xAA)

In short the real x1800xt is on average 25% faster (at 1600x1200+8AF&4xAA) in the extremetech review than in Sanders, something that hardly can be explained by a 4% speed bump and a couple of weeks newer drivers. If we ignore Doom the real x1800xt is 34% faster.

 

[Razor1]

In Halflife 2 the real x1800xt is 35% faster reletively to the GTX than in Sanders "review", in Doom 3 in the Sander score was only 2% off but he had some strange nonexistent performance boost over the x850xt. In Farcry the real scores was 23% higher (still relatively to GTX). In SC:CT the real x1800xt anihilated the Sander x1800xt by being 45% faster.

(all based on the extremetech scores at 1600x1200+8AF&4xAA)

In short the real x1800xt is on average 25% faster (at 1600x1200+8AF&4xAA) in the extremetech review than in Sanders, something that hardly can be explained by a 4% speed bump and a couple of weeks newer drivers. If we ignore Doom the real x1800xt is 34% faster.

well not all reviewers got the same benchs. Anandtech didn't get those scores, Techreport didn't either. Both of those sites are very similiar to Sander's % differences. There is still too much flucuation on benchmark % from different reviewers possible due to which part of the games were benchmarked. But it doesn't seem Sander wasn't lieing. If anything he showed areas that were favorable to nV cards.

 

[Tim]

well not all reviewers got the same benchs. Anandtech didn't get those scores, Techreport didn't either. Both of those sites are very similiar to Sander's % differences. There is still too much flucuation on benchmark % from different reviewers possible due to which part of the games were benchmarked. But it doesn't seem Sander wasn't lieing. If anything he showed areas that were favorable to nV cards.

Compared to techreport (who did not bench at with 8AF but 16AF, so the comparison is not completely valid), Sander got it even more wrong in farcry (35%), OK in SC:CT the real x1800xt was only 40% better (and not 45% like at ET). Once again the Doom 3 scores was only a couple percent off and there where no Halflife 2 scores. Compared to anandtech sanders scores in Farcry and SC:CT is also completely off.

The fact that the overall conclusions from Techreport and Anandtech is similar to Sanders does not change the fact that theire benchmark results is completely different.

 

[PatrickL]

Too much wasted time with this guy, let him rest in pace until 2006

 

[HaLDoL]

But he was right about the X1800XT actual date on the market.

 

[Neeyik]

I would like to point out that the original thread was created to allow both sides, ATI and Sassen, to present their own comments and views about the R520 article. However, unless either party choose to continue the debate in this thread fairly soon, then it will be closed - nothing ever good comes of sniping about other people and websites.

 

[digitalwanderer]

nothing ever good comes of sniping about other people and websites.

What about intraweb drama?

 

[Neeyik]

What about intraweb drama?

Never liked soap operas myself. Seriously though, I personally dislike threads about other websites where the content is generally very negative - that's not to say that I oppose them outright; it's just that in this particular case, the threads remains for people to see if the results from Sassen's article were justified or not, and then to allow the appropriate parties to comment on this. If this doesn't happen, then the thread will get closed because we've already had such a similar debate.

 

[digitalwanderer]

I was just being a bit silly Neeyik, I agree with your reasoning.

I figured one thread about it to check to see what everything thought after the fact wouldn't be bad, but I don't think a bash Sander thread is necessary either.

 

[Pete]

He did show the same clocks as were used with the GPGPU R520 sample for their ClawHMMer paper, so apparently at least one card with those clocks existed (contrary to what ATI said, IIRC).

 

[Nite_Hawk]

He did show the same clocks as were used with the GPGPU R520 sample for their ClawHMMer paper, so apparently at least one card with those clocks existed (contrary to what ATI said, IIRC).

That, is an interesting tidbit of information. Thanks.

Nite_Hawk

 

[neliz]

That, is an interesting tidbit of information. Thanks.

Nite_Hawk

Now all someone needs to do is clock an xt down to 600mhz and see if sander is/was right, but only 75mhz accounting for a 40% speed deficit still doesn't sound believeable...

 

[Razor1]

Now all someone needs to do is clock an xt down to 600mhz and see if sander is/was right, but only 75mhz accounting for a 40% speed deficit still doesn't sound believeable...

Then there is Rys and PC prespective benchmarks, 3/4 of them are very close to what Sander's got, (% difference), What ever it is this is a never ending arguement.

 

[Tim]

Then there is Rys and PC prespective benchmarks, 3/4 of them are very close to what Sander's got, (% difference), What ever it is this is a never ending arguement.

Compared to Hexus.net the relative performance is atleast 15% of in Farcry and still 35-40% off in Splinter Cell - so no Rys´ numbers are not close the Sanders. Sander is is only off by little more than 10% in comparison with PC perspective (both Farcry and Halflife 2), PC perspective did however use a GTX at 450MHz.

And yes it is a never ending argument because you don't check you facts before you post and someone has to correct your BS. You make a post where you claim Anandtechs and Techreports numbers are close to Sanders - the fact is that they are way off. Then you make a new post where you once again makes similar just with different reviews, but once again the numbers does not show what you claim. It just seems like you refering to random reviews hoping that they by pure chance are matching Sanders numbers.

 

[Tim]

Now all someone needs to do is clock an xt down to 600mhz and see if sander is/was right, but only 75mhz accounting for a 40% speed deficit still doesn't sound believeable...

25Mhz = 4.2% higher clockspeed.

 

[neliz]

25Mhz = 4.2% higher clockspeed.

yes, I'm trying to indicate that the marginal difference in clockspeed can never account for a 40% difference in a benchmark..

 

[Razor1]

Compared to Hexus.net the relative performance is atleast 15% of in Farcry and still 35-40% off in Splinter Cell - so no Rys´ numbers are not close the Sanders. Sander is is only off by little more than 10% in comparison with PC perspective (both Farcry and Halflife 2), PC perspective did however use a GTX at 450MHz.

And yes it is a never ending argument because you don't check you facts before you post and someone has to correct your BS. You make a post where you claim Anandtechs and Techreports numbers are close to Sanders - the fact is that they are way off. Then you make a new post where you once again makes similar just with different reviews, but once again the numbers does not show what you claim. It just seems like you refering to random reviews hoping that they by pure chance are matching Sanders numbers.

oh Tim grow up, down clock the current xt cores to 600 and 1400 on the mem, you will end up with those numbers. I know its a bit too much math for you to comprehend, 3 out of the 4, I wasn't including SC:CT.

What if they used a 450 mhz GTX, guess what most GTX manufacturers don't use default clocks, most of them are 450+. Also we don't know driver version, that 10% differential doesn't seem much when you downclock. its down to what 2-3% after that. So whats you point. Is Sander that good of a fortunate teller where he can tell you the benchmarks from 3 different games from 2-3%? I don't think so. Because everyone believed ATi had double z I would think the Doom 3 benchmarks would have been alot closer if he took that into concideration of faking the benchmarks. Which never happend. He is an ASS but not a liar.

 

[Rys]

oh Tim grow up, down clock the current xt cores to 600 and 1400 on the mem, you will end up with those numbers. I know its a bit too much math for you to comprehend, 3 out of the 4, I wasn't including SC:CT.

No, downclocking the XT to those levels does not give you the results you want. Locking this thread to prevent more stupidity or hostility on what's already a too-trodden subject with no useful merit.