Personally I think Nvidia made a mistake by going with 128 bit DDR2 instead of 256 bit DDR. Then again, I'm not sure how the costs work out, but eventually both companies are going to end up with 256 bit DDR2. I guess it's a moot point as to which method to go. Although ATi at any time could go to DDR2, I guess you could also say that Nvidia at any time could go to a 256 bit bus. On the other hand it seems like it would take a little bit more redesigning to go to 256 bit but in the end it's probably a wash. Considering the delays of the GF FX already, maybe it's just as well Nvidia didn't also elect to try to move to a new bus width too.
NVnews fun....
- Thread starter WaltC
- Start date
antlers4 said:
I think you are right but the matter is that the original DDR-2 was developed with nvidias spec and no input from ATi was requested AFAIK. Further there was some talk of creating a proprietary DDR-2 in favor of a particular IHV. Then ATi comes out with a better spec that turns out to be faster and renames it GDDR-3. Also with this spec there was one particular IHV attempting proprietary measures. But I think the real problem began when ATi was left out of the process, pure speculation on my behalf but I think this is what spurred this whole matter on.
DemoCoder,
Something being independently "100% fact" does not mean it is not spin. In fact, that is a major tenet of political speaking. Your professed ignorance of this simple observation leaves us with a statement that to me is "spin", and to you is proven to not be spin since obviously: wording choice cannot influence a message as long as the elements of the message are factual. I'm sure politicians would love manipulating your professed gullibility.
I don't know where to begin to address such a blatantly (and purposefully it seems to me) naive statement, and I don't know that you've ever given reason me to believe it is other than a waste of time to do so in your case, as unless it can be mathematically defined your ability to deny is demonstrably infinite (atleast, I assume you'd yield to mathematical proof).
Instead, I'll let the statements stand...let those who agree, agree, and those who don't, don't...
Something being independently "100% fact" does not mean it is not spin. In fact, that is a major tenet of political speaking. Your professed ignorance of this simple observation leaves us with a statement that to me is "spin", and to you is proven to not be spin since obviously: wording choice cannot influence a message as long as the elements of the message are factual. I'm sure politicians would love manipulating your professed gullibility.
I don't know where to begin to address such a blatantly (and purposefully it seems to me) naive statement, and I don't know that you've ever given reason me to believe it is other than a waste of time to do so in your case, as unless it can be mathematically defined your ability to deny is demonstrably infinite (atleast, I assume you'd yield to mathematical proof).
Instead, I'll let the statements stand...let those who agree, agree, and those who don't, don't...
Sabastian,
GDDR-III had input from nVidia IIRC, as well as tentative support (inasmuch as they don't steal the thunder of their own DDR-II marketing). I'm not aware of anything stopping nVidia from using it in anything when it is available (if they need the higher clock speeds and nothing else better is more cost efficient), and since I thought DDR-II as used on the nv30 was dubbed "GDDR-II" at some point, I think dropping that emphasis allows incorporating GDDR-III into future marketpseak smoothly for them.
This would assume to indicate a dedication to a 128-bit bus...they don't need much more in the way of RAM clock speed moving forward if they do otherwise.
GDDR-III had input from nVidia IIRC, as well as tentative support (inasmuch as they don't steal the thunder of their own DDR-II marketing). I'm not aware of anything stopping nVidia from using it in anything when it is available (if they need the higher clock speeds and nothing else better is more cost efficient), and since I thought DDR-II as used on the nv30 was dubbed "GDDR-II" at some point, I think dropping that emphasis allows incorporating GDDR-III into future marketpseak smoothly for them.
This would assume to indicate a dedication to a 128-bit bus...they don't need much more in the way of RAM clock speed moving forward if they do otherwise.
Here we are, I don't know how credible their sources are but......
"On the sidelines of the JEDEX conference here, memory industry insiders complained about growing pressure from graphics chip suppliers that want to tailor upcoming DRAM to meet the needs of their high-end, low-volume graphics chips. In some cases, sources said, the top two graphics chip suppliers have tried to get manufacturers to include proprietary features in their memories that would preclude competitors from using the chips, further splintering an increasingly divided memory market."
http://www.eetimes.com/semi/news/OEG20021024S0043
"On the sidelines of the JEDEX conference here, memory industry insiders complained about growing pressure from graphics chip suppliers that want to tailor upcoming DRAM to meet the needs of their high-end, low-volume graphics chips. In some cases, sources said, the top two graphics chip suppliers have tried to get manufacturers to include proprietary features in their memories that would preclude competitors from using the chips, further splintering an increasingly divided memory market."
http://www.eetimes.com/semi/news/OEG20021024S0043
OT:
Look in your driver's manual: it will tell you what octane gas to use. If your car's engine was designed for 87 octane fuel--pretty likely unless you have an old (pre fuel-injection) car, or a high-performance one--you won't see any difference with 93 octane except for wasted money. Wait, that's not true: you may actually get worse gas mileage and damage your catalytic converter and the environment, because you're going to be sending unburned gasoline through your emissions system.
If your car *was* designed for high-octane gas you will not only see a performance difference but will probably get better mileage, partially offsetting the higher cost. It all depends on what your fuel-injection system is expecting.
Incidentally, this brings up an interesting observation on fanchildism, namely that it often occurs over issues where there actually is an objectively correct answer, and that people who know a bit about the question at hand tend to find the ignorant fanchildren more than a little silly. You've probably felt this way quite often about the vast majority of 3d-related discussion that occurs on the Internet. (BTW, I don't know anything about car engines, but had been tipped off re: the octane thing by someone who does, and filled in the blanks for this post with some googling.)
Of course that rarely stops the experts from being fankids themselves, only about slightly more complicated issues. But it does at least mean that there's some potential good to be had from message boards, even if they are all filled with Them-Who-Shall-Not-Be-Named. Assuming, at least, that the posters are knowledgable and reasonable enough to actually get at some of the objective answers underlying many of the issues argued over.
Sage said:
Look in your driver's manual: it will tell you what octane gas to use. If your car's engine was designed for 87 octane fuel--pretty likely unless you have an old (pre fuel-injection) car, or a high-performance one--you won't see any difference with 93 octane except for wasted money. Wait, that's not true: you may actually get worse gas mileage and damage your catalytic converter and the environment, because you're going to be sending unburned gasoline through your emissions system.
If your car *was* designed for high-octane gas you will not only see a performance difference but will probably get better mileage, partially offsetting the higher cost. It all depends on what your fuel-injection system is expecting.
Incidentally, this brings up an interesting observation on fanchildism, namely that it often occurs over issues where there actually is an objectively correct answer, and that people who know a bit about the question at hand tend to find the ignorant fanchildren more than a little silly. You've probably felt this way quite often about the vast majority of 3d-related discussion that occurs on the Internet. (BTW, I don't know anything about car engines, but had been tipped off re: the octane thing by someone who does, and filled in the blanks for this post with some googling.)
Of course that rarely stops the experts from being fankids themselves, only about slightly more complicated issues. But it does at least mean that there's some potential good to be had from message boards, even if they are all filled with Them-Who-Shall-Not-Be-Named. Assuming, at least, that the posters are knowledgable and reasonable enough to actually get at some of the objective answers underlying many of the issues argued over.
demalion, I can only find one word that seems to displease you in the PR statement, the world "cannot", it seems that if the statement had been worded
"in situations where you need more bandwidth, you can choose to increase clock rate, bus width, or increase efficiency" instead of
"in situations where you cannot increase clock rate, or efficiency, you can ..."
But I find your twisting this statement into some monumental anti-ATI PR spin to be ludicrous. I think I know a bit more about PR than you do, given that I actually hired a PR firm when I ran my company ($20,000/mo) and was coached by a PR handler before each press interview about what to say. I know exactly how it works, and PR statements are not that subtle. Perhaps for you, the NVidia guy made a poor choice of words. I highly doubt he was implying that ATI could not increase ram speed since ATI can obviously buy any RAM that NVidia can.
When you learn PR, you prepare typically 3 anchor points that you want to get across to the interviewer before hand, and you repeat them as many times in as many ways as possible during the interview. You never answer in the negative (e.g. "does you product do X?" answer: "No" is bad), you always refer to the positive aspects of your product/company instead of reflecting on the negatives. You have to be direct, not indirect. That's why politicians directly attack each other by name and by their record.
Democrats, for example, would never attack Bush during a campaign by not mentioning his name or his policies, but issuing vague statements like "some people's healthcare plans decrease care for some, and increase it for others." In marketing, you've got people with incredibly short attention spans and who may have no knowledge at all of the context, so you must be direct and unequivocal.
The idea that this guy is trying to attack ATI, but not mention them by name, and using incredibly subtle language is ludicrous. It's a complete waste of his time, since the vast majority of the people won't even know what's hes talking about. So if this is "spin", it's the most incompetent spin I've ever seen.
Want to see what real spin looks like, here's a spin like PR statement (written by me just now):
Question: Why did you choose faster RAM instead of a 256-bit bus?
"The GeForceFX uses state of the art DDR2 memory at 1Ghz to achieve breakthrough bandwidth and price/performance. Utilizing advanced compression and DDR2 memory, the GeForceFX makes Cinematic Shading a reality."
Accentuate the positive, ignore the real meaning of the question.
If pressed, you could say
"The GeForceFX is the most advanced chip we have designed to date. With over $50 million spent on development at our state-of-the-art design center, the design is a revolutionary .13um micron device with over 110 million transistors"
If pressed further, you would say
"Our Academy Award winnning engineers looked at all different ways of increasing bandwidth, and we decided that the 256-bit bus is overrated. We felt that by going with Revolutionary DDR2 memory, we could simplify our design and lower costs. The 256-bit bit bus is a great idea for the future, but we can achieve Cinematic Shading today without it."
That's real PR. Playing hidden meaning games with innuendo? That's for deconstructionists to crap around with on campus.
Sorry if I seem to be "condescending" again, but the fact of the matter is, I have been through PR training, and I just don't see the blantant spin here. When I here Derek Perez or David Kirk speak, it's almost painful (especially to see a smart guy like Kirk be forced to do it). Sorry, but this latest statement just doesn't set off my "super PR alert" alarm.
"in situations where you need more bandwidth, you can choose to increase clock rate, bus width, or increase efficiency" instead of
"in situations where you cannot increase clock rate, or efficiency, you can ..."
But I find your twisting this statement into some monumental anti-ATI PR spin to be ludicrous. I think I know a bit more about PR than you do, given that I actually hired a PR firm when I ran my company ($20,000/mo) and was coached by a PR handler before each press interview about what to say. I know exactly how it works, and PR statements are not that subtle. Perhaps for you, the NVidia guy made a poor choice of words. I highly doubt he was implying that ATI could not increase ram speed since ATI can obviously buy any RAM that NVidia can.
When you learn PR, you prepare typically 3 anchor points that you want to get across to the interviewer before hand, and you repeat them as many times in as many ways as possible during the interview. You never answer in the negative (e.g. "does you product do X?" answer: "No" is bad), you always refer to the positive aspects of your product/company instead of reflecting on the negatives. You have to be direct, not indirect. That's why politicians directly attack each other by name and by their record.
Democrats, for example, would never attack Bush during a campaign by not mentioning his name or his policies, but issuing vague statements like "some people's healthcare plans decrease care for some, and increase it for others." In marketing, you've got people with incredibly short attention spans and who may have no knowledge at all of the context, so you must be direct and unequivocal.
The idea that this guy is trying to attack ATI, but not mention them by name, and using incredibly subtle language is ludicrous. It's a complete waste of his time, since the vast majority of the people won't even know what's hes talking about. So if this is "spin", it's the most incompetent spin I've ever seen.
Want to see what real spin looks like, here's a spin like PR statement (written by me just now):
Question: Why did you choose faster RAM instead of a 256-bit bus?
"The GeForceFX uses state of the art DDR2 memory at 1Ghz to achieve breakthrough bandwidth and price/performance. Utilizing advanced compression and DDR2 memory, the GeForceFX makes Cinematic Shading a reality."
Accentuate the positive, ignore the real meaning of the question.
If pressed, you could say
"The GeForceFX is the most advanced chip we have designed to date. With over $50 million spent on development at our state-of-the-art design center, the design is a revolutionary .13um micron device with over 110 million transistors"
If pressed further, you would say
"Our Academy Award winnning engineers looked at all different ways of increasing bandwidth, and we decided that the 256-bit bus is overrated. We felt that by going with Revolutionary DDR2 memory, we could simplify our design and lower costs. The 256-bit bit bus is a great idea for the future, but we can achieve Cinematic Shading today without it."
That's real PR. Playing hidden meaning games with innuendo? That's for deconstructionists to crap around with on campus.
Sorry if I seem to be "condescending" again, but the fact of the matter is, I have been through PR training, and I just don't see the blantant spin here. When I here Derek Perez or David Kirk speak, it's almost painful (especially to see a smart guy like Kirk be forced to do it). Sorry, but this latest statement just doesn't set off my "super PR alert" alarm.
Hellbinder
Banned
I could have sworn i had a post in this thread.... Didn't I?????
I have heard NVidia justify 128 bit wide DDR II by saying its a balanced design between the GPU and the memory controller to RAM. They also said 256 bit wide DDR II would either be 1) over kill 2) too expensive or 3) not really deliver much more because it would unbalance the design.
Let's take those sentiments at face value for a moment. It poses some interesting questions.
If you had moved to either 128MB or 256MB of 256 bit wide RAM I supposed you'd have 32 GB/sec raw bandwidth. Would we expect NV30's performance to lift by up to 100% by merely doubling the memory bandwidth - or would the gain be only 10% - 20% more?
Can anyone say whether and in what situations the NV30 will be losing out heavily because it is memory starved? Is it fast enough to run 1,600 * 1,200 with 8 * AA and 16 * AF in the latest games?
With the current design where will NV30 hit the wall - what are its limiting factors? I seem to hear everyone say sure the GPU is faster but it may not have the memory bandwidth to do great image quality improvement techniques. Too in these forums many folk have said (at least in benchmarks like 3d Mark 2001 SE) that todays highest end video cards are at least 50% CPU limited in the tests.
Does anyone have a good link towards how top end video cards scale vs CPU? I'd love to see a shader based game like Aquanox benchmarked in high resolution (1,600 * 1,200 in 32 bit colour) on a GF4 Ti4600 then a 9700 PRO on a systems platform ranging from a 1 GHz CPU in say 500 MHz increments up to 3 GHz machine.
Let's take those sentiments at face value for a moment. It poses some interesting questions.
If you had moved to either 128MB or 256MB of 256 bit wide RAM I supposed you'd have 32 GB/sec raw bandwidth. Would we expect NV30's performance to lift by up to 100% by merely doubling the memory bandwidth - or would the gain be only 10% - 20% more?
Can anyone say whether and in what situations the NV30 will be losing out heavily because it is memory starved? Is it fast enough to run 1,600 * 1,200 with 8 * AA and 16 * AF in the latest games?
With the current design where will NV30 hit the wall - what are its limiting factors? I seem to hear everyone say sure the GPU is faster but it may not have the memory bandwidth to do great image quality improvement techniques. Too in these forums many folk have said (at least in benchmarks like 3d Mark 2001 SE) that todays highest end video cards are at least 50% CPU limited in the tests.
Does anyone have a good link towards how top end video cards scale vs CPU? I'd love to see a shader based game like Aquanox benchmarked in high resolution (1,600 * 1,200 in 32 bit colour) on a GF4 Ti4600 then a 9700 PRO on a systems platform ranging from a 1 GHz CPU in say 500 MHz increments up to 3 GHz machine.
DemoCoder said:
Hmm...is dropping the comments about compression supposed to be subtle, DemoCoder?
Actually, DemoCoder, if your comments had been limited to an evaluation of whether the PR spin was "monumental" or not I would not have commented as I had. Just as if your comments concerning the "dustbuster" had been restricted to whether many of the comments were useful or not I would actually have come into that thread agreeing with you.
But, as usual, your comments are not like that.
Let me revisit my comment in a "dustbuster" thread and see if it applies:
[url=http://www.beyond3d.com/forum/viewtopic.php?p=59615&#59615 said:
Hmm...oh wait, you said "take lithium", not "you're a fanboi". My parallel must be flawed, eh?
Also present in my post are comments about how, instead of sticking to a rational argument about the merits of an opposing viewpoint, you prefer to try to marginalize the very idea of disagreeing with your viewpoint by ignoring simple truth staring you in the face, usually by attacking the idea that the person is qualified to hold and opinion (in the case of that thread, it was inconvenient how some holding the viewpoint you attacked actually had some "thermal engineer" related backgrounds, eh?).
Then, it was, for example, ignoring the reports that indicated the fan was noisy, and marginalizing the concern that it was a radical new design that might have problems.
In this, it is simply trying to ignore the fact that the statement is spin, pure and simple....or atleast that was what you were doing before you apparently admitted it amongst the BS in this post.
It is almost like you were trained in the art of BS... Interesting that you couldn't let the statements stand and had to redefine your stance to finally admit that it is spin, and pretend we (you and I, that is) were arguing over whether it is "monumental" or not (your recent comments about "childish hyperbole" already strike me as bitterly ironic, you don't have to make them moreso).
Your entire post is about "I know more than you, and even though it is spin, you aren't qualified to point it out". This is so far afield from actual coherency, and so far into your constant obsession with substituting ego comparison for logic that I am again truly amazed by your argumentative "prowess".
In case there is any doubt, I don't mean in a good way.
I don't know, "English must not be my first language" (that comment of yours still amazes me
), so I must just be failing to understand the subtleties of "Playing hidden meaning games with innuendo", which you seem to finally admit is what this is, and how that is not spin. I guess if the BS is only knee high it doesn't count, and it only warrants recognition if you need a snorkle? So that's where you argument structure came from!Psst...what about the compression comment? A bit too inconvenient to mention that when my comment stated "with the same wording and the same inaccurate implication regarding a competing product"?
Doesn't fit that interview at all, right?
Are you arguing for me, or against me?So this was not direct because they didn't say "ATI". Saying "256-bit bus" does not qualify? I' find it odd that you paid people $20,000/mo (wow, that's a big number, you must be right) for such a simple definition of PR and spin. It could be you are selectively remembering...but what basis would I have for believing that? Hmmm...
Hey, but that would be attacking the person, not the point.
To address your point, I simply ask you to visit here and tell me everything there is based on being direct (to me, it looks like it is not the case).
I found out about Edward Bernays, the so-called "Father of Spin".
From that, I also went and perused the online readable pages of "The Father of Spin" and observed how early examples of his persuasion were indeed not "direct".
I finally found an online museum of PR and noticed that the examples I persued also were not direct. But were still spin. Hmm...
In all cases, your simple mandate does not seem to be reflected, even though you propose it as all encompassing because...what...you paid $20,000/month for it?
I don't know, I personally believe you are just heaping on the BS.
Hmm...I really could spend a lot of text doing this to the rest of your post, but I'm not going to. If you think there is some part of your post I couldn't have done that to, simply repeat it. As it stands, I think I've addressed your comments and don't think doing that again for every time you repeat the same basic idea in your post serves a purpose, so I'll skip ahead to your last paragraph.
Being consistently condescending can be abrasive to people's estimation of you. When redeeming characteristics of your post are absent, I propose you shouldn't be surprised if people react to you in a certain way.
Then again, you may really think "I paid $20,000 a month" makes your argument strong enough to be a redeeming characteristic.
So if it doesn't go the "heights" of direct and incorrect statements about a "competitor's method" it doesn't count? Wait a minute...did he (EDIT: David Kirk, I believe) even "name" ATI in that interview (EDIT: the one with the incorrect statements I refer to, which I think you recall)? If he didn't, I guess you and I both don't have any case against him.
I don't know what is more sad- that quote from NVIDIA or the reponses from the known sympathizers for the company.
The spin on that quote is pretty obvious to see. The feeling is:
"If you are totally inept at making hardware and designing cutting edge compression technologies, you are forced to go to the evil, convulted and dreaded 256 bit."
This is from the same folks that previously dictated the same ethic several years ago that large videocards were "bad" and having an external power connector was similar technological ineptness.
I only hope that ATI will return fire in similar style, although I doubt they will. It would make for some interesting dialogue on these forums to hear the sympathizers talk about evil ATI's comments concerning how a 128 bit is bandwidth stiffling mhz to mhz and how resorting to a multiple slot card with thermal conduction chamber and fan assembly the size of a Volkswagon is a necessity to get a clockspeed up to be equivalent to a smaller, lower clocked 256 bit bus.
The spin on that quote is pretty obvious to see. The feeling is:
"If you are totally inept at making hardware and designing cutting edge compression technologies, you are forced to go to the evil, convulted and dreaded 256 bit."
This is from the same folks that previously dictated the same ethic several years ago that large videocards were "bad" and having an external power connector was similar technological ineptness.
I only hope that ATI will return fire in similar style, although I doubt they will. It would make for some interesting dialogue on these forums to hear the sympathizers talk about evil ATI's comments concerning how a 128 bit is bandwidth stiffling mhz to mhz and how resorting to a multiple slot card with thermal conduction chamber and fan assembly the size of a Volkswagon is a necessity to get a clockspeed up to be equivalent to a smaller, lower clocked 256 bit bus.
Diespinnerz
Newcomer
The sad thing is you actually believe the crap you spout sharkfood.
If NVIDIA would just shut up, or even cut down the crap by 50%, I wouldn't have a single substantial beef with them as a company. Smacks of the big lie, when they don't have to resort to that, their products are actually quite good. I think about buying some NVIDIA card, then I run into some heap of steaming PR and I get repulsed and look elsewhere.
nVidia PR, IMO, is in a really bad position right now.
On one hand, they want the GFFX to sell well, so they got to say not having a 256-bit bus is not all that important. And they've absolutely got to focus on what the card can do ( huge shader programs, cheap adaptive Aniso )
On another hand, they don't have to make the 256-bit bus look too bad because they're gonna have one real soon. Probably NV35 ( but it might have to wait for a NV35 refresh, who knows )
So, in barely a few months they'll get a bandwidth monster. And at the same time, they got to defend their stand on their card not being bandwidth limited.
Either way, one of their goal won't be realized correctly.
I pity you, nVidia PR
Uttar
On one hand, they want the GFFX to sell well, so they got to say not having a 256-bit bus is not all that important. And they've absolutely got to focus on what the card can do ( huge shader programs, cheap adaptive Aniso )
On another hand, they don't have to make the 256-bit bus look too bad because they're gonna have one real soon. Probably NV35 ( but it might have to wait for a NV35 refresh, who knows )
So, in barely a few months they'll get a bandwidth monster. And at the same time, they got to defend their stand on their card not being bandwidth limited.
Either way, one of their goal won't be realized correctly.
I pity you, nVidia PR
Uttar
Sabastian,
Here is an nVidia comment provided by ben that, in my interpretation, leaves the option of GDDR III open to nVidia if they need it. But I can see how it could be reasonably interpreted to support what you are saying as well.
Here is an nVidia comment provided by ben that, in my interpretation, leaves the option of GDDR III open to nVidia if they need it. But I can see how it could be reasonably interpreted to support what you are saying as well.
