R520 benchmarks - Hardware Analysis
- Thread starter Sxotty
- Start date
- Status
I am still happy he published it, even if it is crazy, we need good rumors to moan about don't we? And the inquirer is too boring anyway 
You should have done as you just posited though Sassen.
Simply showed the graphs, and said I got these from a source, I did not see or touch the harwdware, nor can I vouch for the accuracy of the results.
If you left it at that I think the storm would be much less.
You should have done as you just posited though Sassen.
Simply showed the graphs, and said I got these from a source, I did not see or touch the harwdware, nor can I vouch for the accuracy of the results.
If you left it at that I think the storm would be much less.
BruteForce170
Newcomer
ati has made misleading and/or completely inaccurate statements before, why does everyone assume they are accurate this time?digitalwanderer said:ATi has said it's bullshit, ATi has said no partners have production grade boards to bench with; that is nothing?
Man, tough crowd here.
what else do you expect them to say, "sorry folks, our high end products just aren't competitive with nvidia's"?
while i find it interesting that a week or two away from launch AIB's have no production grade boards, so what? there was never a claim of using a production based board. I find it really interesting that this article has spawned such passion across most of the h/w based forums
as to the accuracy of the #'s, we can all argue back and forth, but the reality is we really can't substantiate or refute them at this time can we? we'll just have to wait a few weeks. seems to me then we can either burn sander at the stake, or thank him for giving us some "insider info", but to do so now with nothing to substantiate either way seems a bit premature....
i also don't believe (axe to grind or not) that he would post completely false information, as that would most certainly be an actionable (as in leaving him open to a slander suit by ati) offense, and why would he do that? doesn't make sense...
It would be the first step towards supporting your basis if you'd disclose or otherwise make available those "benchmark scripts" for public consumption.ssassen said:
If you're already done so, forgive me. But by providing those, you take some amount of accountability for the legitimacy of the tests/results shown. It also creates accountability for future capabilities to backtrack over these results (for R5xxx and other compared hardware in the tests).
Cheers
ssassen said:
OK, but since you were not permitted any degree of verification by your source then what was it exactly that tempted you to think that such scores as you received from that source might have been genuine? I would have seen any attempt by an outside source to get me to publish numbers I could not verify as a direct attempt at manipulating me and I would have rejected such an attempt out of hand. But again, that's just me...
As you mention FUDO at the INQ, are you now naming him as a source of some kind? If so, then you know that what he publishes are also unverified rumors, which hardly would inspire my confidence. Apparently, though, you aren't claiming even direct contact with FUDO, but rather simply that you read something FUDO wrote and thought it applicable to your own account. That leads me to suggest that maybe the source of your test reports had also read FUDO's rumors and built upon them.What you failed to do is precisely what FUDO fails to do most of the time--and that is to verify the content of whatever rumors you may receive which you yourself are unable to personally verify. In an earlier response you mentioned that Hook had your cell-phone number. This leads me to believe that you also have his, which in turn leads me to suggest that it might have been prudent for you to simply use that number prior to publishing your report in an attempt to verify whether the info you had received was even close to credible. I mean, unless Hook is telepathic there really was no way or reason for him to have contacted you about information he did not know you had received and did not know that you planned to publish, it seems to me. But you certainly had ample opportunity to contact him--especially if, as you say, you have each other's cell numbers handy. So the question for me is why you did not do that. Taking it for granted that you also are not telepathic you could not have known in advance what he would tell you in response to such a query.
As to what you "should have done," since you bring it up, I think you should have attempted to verify the information you received before you published it and, failing that, declined to publish it.
Also, I'd really like to know what it is that ATi does, exactly, which might put a journalist "in a bind"...
Are you referring to the fact that probably no one at ATi is telepathic and so cannot know what journalists are going to write before they write it so that they can call these journalists and set them straight ahead of time? Look, in all seriousness, fact is fact and rumor is rumor and ne'er the twain shall meet...
A competent journalist would never become confused between the two, right?
WaltC,
I hear you, and I wholeheartedly (lovely word) agree. The thing is that once I got these numbers in me and ATI were no longer on speaking terms, hence calling Chris was not an option. I did type up an email to him, stating that I was going to post preliminary scores and the reason for that was all due to the personal prejudice of Andrzej and the situation resulting from that. That email ended up in my trashcan prior to being sent as honestly what would his reply be? Either (a) go ahead we don't care what you post, as they said earlier (b) holdon lets talk, you don't have to go to these lengths, or (c) if you do that we'll get nasty and clip your toenails with a beltsander or (d) no reply. I guessed that whatever their reply it would come off as me being an arrogant prick basically blackmailing them to please work with me. Considering the fact that they stated 'you can write what you like' and 'don't expect to hear from us in the next 6 months' I just went ahead and posted it.
As for the benchmarks (you did read the 2nd post in this thread did you) the source is not Fudo, but I talk to Fudo regularly by IM so this has been discussed between us at length. He however did not vindicate my scores, but said exactly what he wrote on TheINQ. And as I outlined before I had no way of verifying due to the physical distance (me close to Amsterdam in Europe, them somewhere in Taiwan), but I believe my contact at the AIB would not deliberately feed me false scores. Again, I outlined all of this in the 2nd post already.
When I refer to what you call 'in a bind' it is simply the fact that they say they want to engage on every level, but if I put it in black and white they will happily use against you any comments you make about ATI when for example you're just having beers, or to other journos, or simply in a critical article, that are then used as a leverage later when determining exactly who will be elligible for an early sample. So many journalists try to get into their good grace as that would guarantee such early access, although I say this with caution, as these are my views and my experience, I do get similar stories from others though, but I'll leave up to them to comment on that.
Regards,
Sander Sassen
http://www.hardwareanalysis.com
I hear you, and I wholeheartedly (lovely word) agree. The thing is that once I got these numbers in me and ATI were no longer on speaking terms, hence calling Chris was not an option. I did type up an email to him, stating that I was going to post preliminary scores and the reason for that was all due to the personal prejudice of Andrzej and the situation resulting from that. That email ended up in my trashcan prior to being sent as honestly what would his reply be? Either (a) go ahead we don't care what you post, as they said earlier (b) holdon lets talk, you don't have to go to these lengths, or (c) if you do that we'll get nasty and clip your toenails with a beltsander or (d) no reply. I guessed that whatever their reply it would come off as me being an arrogant prick basically blackmailing them to please work with me. Considering the fact that they stated 'you can write what you like' and 'don't expect to hear from us in the next 6 months' I just went ahead and posted it.
As for the benchmarks (you did read the 2nd post in this thread did you) the source is not Fudo, but I talk to Fudo regularly by IM so this has been discussed between us at length. He however did not vindicate my scores, but said exactly what he wrote on TheINQ. And as I outlined before I had no way of verifying due to the physical distance (me close to Amsterdam in Europe, them somewhere in Taiwan), but I believe my contact at the AIB would not deliberately feed me false scores. Again, I outlined all of this in the 2nd post already.
When I refer to what you call 'in a bind' it is simply the fact that they say they want to engage on every level, but if I put it in black and white they will happily use against you any comments you make about ATI when for example you're just having beers, or to other journos, or simply in a critical article, that are then used as a leverage later when determining exactly who will be elligible for an early sample. So many journalists try to get into their good grace as that would guarantee such early access, although I say this with caution, as these are my views and my experience, I do get similar stories from others though, but I'll leave up to them to comment on that.
Regards,
Sander Sassen
http://www.hardwareanalysis.com
Hellbinder
Banned
Enough of this...
I started reading this thread but its a waste of time.
Bottom line ATi does not feel the need to work with someone who intentionally tries to smear them. Someone who is totally biased towards their major competitor. Its not rocket science. I have read several of the articles on your site and they show a pattern of Nvidia Bias and Anti Ati rhetoric. Where is the example of Equal outrage at any of the misconducts of Nvidia over the past several years?
Then on top of that ATi has made a public statement that these numbers aren’t even close. They estimate that the numbers are the result of an over clocked X850. I don’t frikking care what you think the numbers are, or what you suppose they were run on.
When you are as Biased an immature as you show yourself to be why would *any* serious public company want to work with you? Other than the obvious reasons of Blind Bias for one hardware over another, and the side you play favorites to using you as their pawn.
When the Final numbers are released you are going to look like a fool. Bank on it. I am sure that wont have any actual affect on you though you will likely just write another slanderously biased and misguided article.
I started reading this thread but its a waste of time.
Bottom line ATi does not feel the need to work with someone who intentionally tries to smear them. Someone who is totally biased towards their major competitor. Its not rocket science. I have read several of the articles on your site and they show a pattern of Nvidia Bias and Anti Ati rhetoric. Where is the example of Equal outrage at any of the misconducts of Nvidia over the past several years?
Then on top of that ATi has made a public statement that these numbers aren’t even close. They estimate that the numbers are the result of an over clocked X850. I don’t frikking care what you think the numbers are, or what you suppose they were run on.
When you are as Biased an immature as you show yourself to be why would *any* serious public company want to work with you? Other than the obvious reasons of Blind Bias for one hardware over another, and the side you play favorites to using you as their pawn.
When the Final numbers are released you are going to look like a fool. Bank on it. I am sure that wont have any actual affect on you though you will likely just write another slanderously biased and misguided article.
Hellbinder said:
Oh great, we're back to square one again, if you had refrained to post more of that nonsense for the umpteenth time we might have actually got to read an interesting post here, but alas you felt compelled to add your 2 cents. I'm sorry, but your post reflects that which we all agreed on several posts back, and I'm sorry again, but I'm going to have to repeat on myself here:
Enough said I guess, but feel free to make a comment that DOES add to the discussion, instead of setting us back a few pages.
Regards,
Sander Sassen
http://www.hardwareanalysis.com
Instead of ping-ponging the same words back and forth, why not apply all this energy towards analyzing the results at HardwareAnalysis.com to see if they jive with anything that has been said?
For example, let's go with the "overclocked X850XT" theory.
Exhibit A: Doom 3 benchmarks.
Using a naive prediction method, the relative scores achieved would imply an X850XT-PE running at over 900MHz.(!!!!!)
Logic:
X850XT-PE: 540MHz core / 1180 MHz memory
X1800XT: 650MHz core / 1400 MHz memory
Let's disregard memory and assume that the "overclocked X850 XT-PE" has the same relative memory bandwidth as the stock model (as difficult as that is to believe).
Let's use the score for the resolution 1024*768@ 4x FSAA & 8x AF as this shows the biggest delta between X850XT-PE and the X1800XT (which you will remember is claimed to be a cleverly disguised X850XT-PE by boosting its clocks).
At stock speeds, as shown on HardwareAnalysis, the X850XT-PE scores an average of 30.6 fps at the aforementioned settings using a custom demo (to which we don't have access). The X1800XT scores 53.8 fps in the same test.
Using a naive method, let's see how overclocked that X850XT-PE would have to be to reach that score, given that it is not imbued with magical powers at a certain core frequency that makes performance increase more than the clock speed.
540MHz/30.6 fps = 17.65 MHz/fps
17.65MHz/fps * 53.8fps = 949.57 MHz (let's call it 950 MHz)
I will add that I think the scores for the Doom 3 looks very low for all ATI cards when compared to Nvidia. Perhaps Sander Sassen can clarify what makes this custom benchmark custom so we can accept a larger discrepancy than what is the reported norm. I would also like to add that if we assume this relative bandwidth, we are working with an overclocked X850XT-PE running at 950 core and a whopping 2076 MHz (effective) memory clock. That's some card! Now, it may be that memory bandwidth is not so important and this is why I said we should disregard the memory clocks (it could be that you get significantly better results with the X850 using only a higher core clock).
At any rate, this result seems outlandish, to say the least, and should make one seriously question the theory that this test was conducted on overclocked X850XT-PE hardware. Now, if any of you can believe that the hypothesis of an overclocked X850XT-PE is reasonable for the result shown in the Doom 3 benchmark, now is the time to chime in. I am sure we can pick apart the bechmarks one by one.
For example, let's go with the "overclocked X850XT" theory.
Exhibit A: Doom 3 benchmarks.
Using a naive prediction method, the relative scores achieved would imply an X850XT-PE running at over 900MHz.(!!!!!)
Logic:
X850XT-PE: 540MHz core / 1180 MHz memory
X1800XT: 650MHz core / 1400 MHz memory
Let's disregard memory and assume that the "overclocked X850 XT-PE" has the same relative memory bandwidth as the stock model (as difficult as that is to believe).
Let's use the score for the resolution 1024*768@ 4x FSAA & 8x AF as this shows the biggest delta between X850XT-PE and the X1800XT (which you will remember is claimed to be a cleverly disguised X850XT-PE by boosting its clocks).
At stock speeds, as shown on HardwareAnalysis, the X850XT-PE scores an average of 30.6 fps at the aforementioned settings using a custom demo (to which we don't have access). The X1800XT scores 53.8 fps in the same test.
Using a naive method, let's see how overclocked that X850XT-PE would have to be to reach that score, given that it is not imbued with magical powers at a certain core frequency that makes performance increase more than the clock speed.
540MHz/30.6 fps = 17.65 MHz/fps
17.65MHz/fps * 53.8fps = 949.57 MHz (let's call it 950 MHz)
I will add that I think the scores for the Doom 3 looks very low for all ATI cards when compared to Nvidia. Perhaps Sander Sassen can clarify what makes this custom benchmark custom so we can accept a larger discrepancy than what is the reported norm. I would also like to add that if we assume this relative bandwidth, we are working with an overclocked X850XT-PE running at 950 core and a whopping 2076 MHz (effective) memory clock. That's some card! Now, it may be that memory bandwidth is not so important and this is why I said we should disregard the memory clocks (it could be that you get significantly better results with the X850 using only a higher core clock).
At any rate, this result seems outlandish, to say the least, and should make one seriously question the theory that this test was conducted on overclocked X850XT-PE hardware. Now, if any of you can believe that the hypothesis of an overclocked X850XT-PE is reasonable for the result shown in the Doom 3 benchmark, now is the time to chime in. I am sure we can pick apart the bechmarks one by one.
Last edited by a moderator:
Razor1 said:what if you factor in double z?
Why would I factor that in? We are testing the hypothesis that the X1800XT and the X850XT-PE are the same cards with the X1800XT being an overclocked X850XT-PE as postulated in one of the ATI replies at DriverHeaven (was it?).
Sure, go ahead and factor that in if you think we are dealing with numbers on new hardware, but that is not the hypothesis I wanted to test. We could just as easily assume the number was pulled out of thin air. However, I think by doing this simple arithmetic we can conclude that an overclocked X850XT-PE is not the answer.
Speaking of pot calling the kettle black, you're doing exactly the same thing by failing to publish anything regarding the benchmarks run. We have every reason to ignore your claims. You admit to having no information about the hardware, the scores don't add up in any way, and you seemingly refuse to release any benchmarking methodology so analysis (by people who are much smarter than you or I) could be done to see if such scores are feasible in any way. There's no information on feature set at all, no synthetic benchmarks (which would have solved everything instantly), and just the most basic of games used as benchmarks. Finally, you package it all as a substantial R520 preview and preface it with a statement saying how you are only releasing these because ATI wouldn't give you a R520 sample.ssassen said:
You act like we're all being totally unreasonable to think that you're a schmuck who has decided to stir up some controversy to raise his site's traffic, but what reason have you given to the contrary?
Also, wireframe, it's clear that ATI have implemented Ultrashadow 2 so it no longer loses in Doom 3.
But if we want to look at benchmarks, look at Pro versus XTPE. You really expect me to believe that despite lower fillrate and memory bandwidth it's 50% faster than an X850XTPE in Doom 3? Bullshit. Compare the B3D X850XTPE review from December 04 to the SLI preview from March with a near-identical test setup, where the X850XTPE scored within 10% of a 6800 Ultra. Or that the Pro is 20% faster in 1024x768 SC: CT (an unnamed version running an unnamed shader version, OF COURSE) but is less than 10% faster at 1280x1024? What the HELL. Or that while Far Cry is faster on the X850 XTPE, HL2 is slower?
These numbers were pulled out of somebody's ass.
Christ, Dave, make me a mod again so I don't have to read this soap operatic shit here. This thread is an absolute waste of bandwidth and space on the B3D server.
Actually, the frustration vented on this thread probably saved the life of the drunk guy who wandered in my room last night at 3:30.John Reynolds said:
wireframe said:
ATI is obviously at a huge disadvantage here. They speculated, and clearly identified it as such. I think it would be a mistake to draw any conclusions by "disproving" that off-the-cuff theory.
I think it much more likely that if this is a case of serious bad faith, then either those numbers are made up out of whole cloth with a calculator and liberal amounts of alcohol. . .or, and this one would be more subtle, using an X1800XL mis-labled ("oops!") as an XT. It is damn curious that there are no XL scores but are XT and PRO scores, given what we know about the release schedule. Mislabeling an XL (remember, my posit was "bad faith"), finesses things a bit. . .allows the slammie on ATI now as frustrated pro-ATIans wince (Remember Bowen? He went out and bought a 7800GTX based on these scores), positions for a "well, I wouldn't have made that mistake if I was under NDA and had support" later (which seems to have been his campaign all along, blown badly by grossly overplaying his hand in the narrative of the article), and still allows him to claim "scoop" victory later as well too, only slightly tarnished by the mislabel --as in this scenario it is still the first signficant X1800 benchies anywhere (but, I must note, would NOT qualify him for an apology from me
).If it isn't a case of serious bad faith, but still is an X1800XL mislabeled, then it is just the karma of trying to pull off a benching situation in these conditions, which many others have already pointed at as rife for miscommunication and impossible for the careful benchmarker to validate.
At any rate, personally I'll be taking a close look at XL benchies against these as well, assuming the XT benchies are not similar at all.
- Status