R600 nice read

[josh6079]

Never underestimate a womans power...

 

[Spoelie]

Really? How many watt does your wife consume?

 

[apoppin]

Originally posted by: Makaveli
She agreed that its sexist, or that you are correct?

If the latter, apoppin u just got pwned by that guys wife, and she wasn't even on the computer LOL.

actually i find it rather interesting that he chose to ask his wife --instead of his girlfriend.
:Q



look at the line above ... he agreed with me ... i said it was sexist ... and she agrees

logical


actually i can't imagine anyone sneaking a peak at what the guests are wearing before arriving late to to be impressively dressed.
just ask the host/hostess what is appropriate attire if you are

 

[Zenoth]

It's just marketing (related to the date) in my opinion.

It's obvious that AMD (or AMi or whatever) is awaiting the release of Vista before releasing R600, or else I'm sure we'd have it by now, or at least a paper launch. It's just too good of a coincidence if so.

But Jan 30th is still much sooner than I expected nonetheless.

 

[Nelsieus]

I just want to remind everyone of the infamous 2005 summer where we all sat on the edge of our seats expecting another 9700 hit from ATI with R520, that was going to "blow G70 out the water."

As much as I accuse theINQ of being ATI cheerleaders, the one thing they do to contradict that claim is basically putting so much excitement into a future product that by the time it releases, and was never bound to meet those unrealistic standards, people think its a dissapointment

And to back that up, just look at G80. Look at how many articles Fuad wrote how it was just going to be G70 with some DX10 stapled on, or I believe "the old way" is how he described it. And then it realeased, completely different of what he, and most others had "predicted" and now we look at it as nVidia's own 9700 hit.

I'm not saying R600 isn't going to be truly amazing, because I really have that faith in ATI. But I hope we all realize theINQ usually likes to embellish facts, especially when it creates excitement for the red team.

Anywho, now to shift subjects. The advantage of a scaler unit is that it has much higher utilisation and could be argued as more versatile. One Vec4 shader is more powerful than onescaler, but 64 Vec4 vs. 128 Scaler is going to be a pretty even match, granted that rumor is true. What I dislike is, though, is Theo's misrepresentation that one is good for "complexity" and the other is for "simplicity." Basically, in other words, Theo's confused and wanted to try to use a good analogy, which unfourtunately is inaccurate (but atleast it's exciting to read *shrugs)

Also, I don't doubt nVidia has some more GPUs to launch this Feb., as Anandtech has been reporting. And for everyone who thinks that totally impossible, then obviously you missed out this last generation when nVidia launched the 7800GTX 512MB in Nov. to take the lead spot, and then the 7950GX2 in June to regain that lead spot again.

Nelsieus

 

[redbox]

Originally posted by: Wreckage

Originally posted by: BlizzardOne

Originally posted by: Cookie Monster
The thing is that nVIDIA is preparing a total of nine G80 SKUs. 5 of them are found in the drivers i think.

Sure it was nine G80 SKU's, and not G8x's? Being spoiled for choice is one thing, but nine derivatives of essentially the same thing is just nonsense.

Well that could include mobile and integrated chips. G8950GX2 FTW!

I know you really like the idea of a G80 GX2, but just how do you think Nvidia will be able to do it? The heat and power limitations alone would keep them from doing it.

 

[AmdInside]

The Inq is just like terrorists. They just have to get it right one time for them to feel like they are succesful. Nevermind that most of the time they get information wrong. There are so many informative websites out there that focus on getting the facts straight, I don't know why people keep quoting a website such as the INQ. I guess its the same people who like to read The National Enquirer and Star at supermarkets.

 

[redbox]

Originally posted by: Nelsieus


Also, I don't doubt nVidia has some more GPUs to launch this Feb., as Anandtech has been reporting. And for everyone who thinks that totally impossible, then obviously you missed out this last generation when nVidia launched the 7800GTX 512MB in Nov. to take the lead spot, and then the 7950GX2 in June to regain that lead spot again.

Nelsieus

I agree on most parts except this. The 7800gtx 512 came 6 months after the first G70. To expect them to have enough yeild in their chips especially one as complex as G80 in 4 months is a little optomistic. I don't doubt they will be producing one, I just don't see them producing one in that short of a time table.

As far as the 7950GX2, they where only able to produce that card because of G71's lower power consumption along with lower heat out put. True we could see another die shrink for G80, but not even close to the time table of Feb. The 7950GX2 was a year after the original G70's. The only thing I can see Nvidia putting together in time for R600 is an 8800gt i.e. another cut down of the gtx.

redbox

 

[ronnn]

Originally posted by: Nelsieus
I just want to remind everyone of the infamous 2005 summer where we all sat on the edge of our seats expecting another 9700 hit from ATI with R520, that was going to "blow G70 out the water."

As much as I accuse theINQ of being ATI cheerleaders, the one thing they do to contradict that claim is basically putting so much excitement into a future product that by the time it releases, and was never bound to meet those unrealistic standards, people think its a dissapointment

And to back that up, just look at G80. Look at how many articles Fuad wrote how it was just going to be G70 with some DX10 stapled on, or I believe "the old way" is how he described it. And then it realeased, completely different of what he, and most others had "predicted" and now we look at it as nVidia's own 9700 hit.

I'm not saying R600 isn't going to be truly amazing, because I really have that faith in ATI. But I hope we all realize theINQ usually likes to embellish facts, especially when it creates excitement for the red team.

Anywho, now to shift subjects. The advantage of a scaler unit is that it has much higher utilisation and could be argued as more versatile. One Vec4 shader is more powerful than onescaler, but 64 Vec4 vs. 128 Scaler is going to be a pretty even match, granted that rumor is true. What I dislike is, though, is Theo's misrepresentation that one is good for "complexity" and the other is for "simplicity." Basically, in other words, Theo's confused and wanted to try to use a good analogy, which unfourtunately is inaccurate (but atleast it's exciting to read *shrugs)

Also, I don't doubt nVidia has some more GPUs to launch this Feb., as Anandtech has been reporting. And for everyone who thinks that totally impossible, then obviously you missed out this last generation when nVidia launched the 7800GTX 512MB in Nov. to take the lead spot, and then the 7950GX2 in June to regain that lead spot again.

Nelsieus

Are you in this thread to discuss the r600 or bash ati and pump nvidia? If you are saying the inquirer is frequently wrong - what was your first clue? If you think the r600 sucks, just say why as a history lesson in the 7800gtx 512mb is about something I don't remember.

 

[Cookie Monster]

The moral of the story is that INQ is not a reliable source. I expect R600 to be pretty fast, heck maybe even faster than the G80. Although specualating is fun, INQ makes specualating look like a joke, because the writer has no clue whats going on. Heck she or he probably dont even know how a GPU pipeline works.

 

[Gstanfor]

Originally posted by: redbox

Originally posted by: Nelsieus


Also, I don't doubt nVidia has some more GPUs to launch this Feb., as Anandtech has been reporting. And for everyone who thinks that totally impossible, then obviously you missed out this last generation when nVidia launched the 7800GTX 512MB in Nov. to take the lead spot, and then the 7950GX2 in June to regain that lead spot again.

Nelsieus

I agree on most parts except this. The 7800gtx 512 came 6 months after the first G70. To expect them to have enough yeild in their chips especially one as complex as G80 in 4 months is a little optomistic. I don't doubt they will be producing one, I just don't see them producing one in that short of a time table.

As far as the 7950GX2, they where only able to produce that card because of G71's lower power consumption along with lower heat out put. True we could see another die shrink for G80, but not even close to the time table of Feb. The 7950GX2 was a year after the original G70's. The only thing I can see Nvidia putting together in time for R600 is an 8800gt i.e. another cut down of the gtx.

redbox

How to you know nvidia hasn't already taped out g81/5 and are stockpiling it? Its quite possible that this could have been done if the refresh is on 80nm. nvidia has had plenty of time to work at their leisure on the G8x family.

Note: to the clueless among us, I'm not saying this is definitely what nvidia has done, its just a possibility and one redbox appears to be ignoring.

 

[ronnn]

Originally posted by: Gstanfor

How to you know nvidia hasn't already taped out g81/5 and are stockpiling it? Its quite possible that this could have been done if the refresh is on 80nm. nvidia has had plenty of time to work at their leisure on the G8x family.

Note: to the clueless among us, I'm not saying this is definitely what nvidia has done, its just a possibility and one redbox appears to be ignoring.

So start a thread on Nvidia's fantastic new sku's and I would know to avoid this sales tripe.

 

[redbox]

Originally posted by: Gstanfor

Originally posted by: redbox

Originally posted by: Nelsieus


Also, I don't doubt nVidia has some more GPUs to launch this Feb., as Anandtech has been reporting. And for everyone who thinks that totally impossible, then obviously you missed out this last generation when nVidia launched the 7800GTX 512MB in Nov. to take the lead spot, and then the 7950GX2 in June to regain that lead spot again.

Nelsieus

I agree on most parts except this. The 7800gtx 512 came 6 months after the first G70. To expect them to have enough yeild in their chips especially one as complex as G80 in 4 months is a little optomistic. I don't doubt they will be producing one, I just don't see them producing one in that short of a time table.

As far as the 7950GX2, they where only able to produce that card because of G71's lower power consumption along with lower heat out put. True we could see another die shrink for G80, but not even close to the time table of Feb. The 7950GX2 was a year after the original G70's. The only thing I can see Nvidia putting together in time for R600 is an 8800gt i.e. another cut down of the gtx.

redbox

How to you know nvidia hasn't already taped out g81/5 and are stockpiling it? Its quite possible that this could have been done if the refresh is on 80nm. nvidia has had plenty of time to work at their leisure on the G8x family.

Note: to the clueless among us, I'm not saying this is definitely what nvidia has done, its just a possibility and one redbox appears to be ignoring.

Because the INQ hasn't told me that yet
No, you are right they could be doing just that. They stock piled the G70 because of ATI's lateness which is part of the reason they could do the hard launch the way they did. Even when they stock piled them the best they could do is 6 months and we all know how the availbilty not to mention price of the 7800gtx 512mb turned out.

I don't intend to sound like my opinion is the end all be all and that's just how it's going to be. I am just giving my opinion that we won't see an ultra gpu from Nvidia in four months. Your view could be just as true as mine.

redbox

 

[josh6079]

How to you know nvidia hasn't already taped out g81/5 and are stockpiling it? Its quite possible that this could have been done if the refresh is on 80nm. nvidia has had plenty of time to work at their leisure on the G8x family.

Note: to the clueless among us, I'm not saying this is definitely what nvidia has done, its just a possibility and one redbox appears to be ignoring.

I don't think he's so much "ignoring" that possibility, it's just that that possibility isn't likely. You very well may be right, but it sounds more like wishful thinking that practical. I agree with redbox that the GX2 technology won't be a factor of the G80 series for a while and that while Nvidia might be able to give another flagship product shortly, I don't think it's feasible in the R600 launch time period.

However, IMO even if the G80 doesn't get a refresh by the time R600 arrives, the G80 already has shown impressive performance and IQ all while having time to improve driver quality and fall in price. I think the greater appeals the G80 will hold with the R600 competition are those factors; it will most likely be cheaper, already offer great IQ and performance, and already have gotten a couple driver versions out of the way on its road for driver optimizations. The only thing that could hurt this is if the R600 flagship had a runner-up that was more of a GTX competitor than a GTS competitor (i.e. X1900XT vs. 7900GTX).

 

[Wreckage]

Originally posted by: redbox

Originally posted by: Wreckage

Originally posted by: BlizzardOne

Originally posted by: Cookie Monster
The thing is that nVIDIA is preparing a total of nine G80 SKUs. 5 of them are found in the drivers i think.

Sure it was nine G80 SKU's, and not G8x's? Being spoiled for choice is one thing, but nine derivatives of essentially the same thing is just nonsense.

Well that could include mobile and integrated chips. G8950GX2 FTW!

I know you really like the idea of a G80 GX2, but just how do you think Nvidia will be able to do it? The heat and power limitations alone would keep them from doing it.

The current GX2 is based off of NVIDIA's mobile chips. Are you saying there will not be a mobile version of the G80?

 

[josh6079]

The current GX2 is based off of NVIDIA's mobile chips. Are you saying there will not be a mobile version of the G80?

Heh, right now the whole thing would probably have to be made of copper.

I think I agree with you though, "G8950GX2 FTW!" not G8850GX2 FTW.

We'll see though. I think Nvidia's first priority is going to shrink the G80's die to make it even more efficient, maybe dress it up with a complete 512-bit / GDDR4 bandwidth and see what power savings can come out of the whole process. After that I wouldn't be surprised for them to release another GX2 pancake card.

 

[redbox]

Originally posted by: Wreckage

Originally posted by: redbox

Originally posted by: Wreckage

Originally posted by: BlizzardOne

Originally posted by: Cookie Monster
The thing is that nVIDIA is preparing a total of nine G80 SKUs. 5 of them are found in the drivers i think.

Sure it was nine G80 SKU's, and not G8x's? Being spoiled for choice is one thing, but nine derivatives of essentially the same thing is just nonsense.

Well that could include mobile and integrated chips. G8950GX2 FTW!

I know you really like the idea of a G80 GX2, but just how do you think Nvidia will be able to do it? The heat and power limitations alone would keep them from doing it.

The current GX2 is based off of NVIDIA's mobile chips. Are you saying there will not be a mobile version of the G80?

There will probably be a G80 mobile chip just not a G80 GX2. For one the Current GX2 uses moblile chips that compete real well with 7900gt chips. I don't see that happening with 8800gts chips. There will have to be alot of work done and I suspect performance droped before the G80 gets down to mobile weight. Did ATI launch a 1900 mobile chip? If so would they have created a card like the GX2? Who knows. Thats the situation Nvidia is in right now. The Big advantage the G71 had was it's heat and power consumption Nvidia exploited that and created the GX2. This time around they just don't have those luxuries.

 

[Gstanfor]

I doubt efficiency will be the driving force for a die shrink - I'm guessing earning margins per die will be weighing more heavily on Mr Huang's mind and thats the primary reason why you'll see nvidia implement a die shrink the instant they are capable of doing so.

 

[josh6079]

Originally posted by: Gstanfor
I doubt efficiency will be the driving force for a die shrink - I'm guessing earning margins per die will be weighing more heavily on Mr Huang's mind and thats the primary reason why you'll see nvidia implement a die shrink the instant they are capable of doing so.

True. Really though reducing die size generally is a win/win for manufacturers and consumers. We get cooler-running, higher overclockable cores and they get a few dollars saved.

 

[Munky]

While were all speculating, I might as well join the fun. It's very possible that NV is planning to do another gtx-512 style launch in time for the r600, which means higher clocks and most likely low availability. But, it's also possible that Ati is already expecting this kind of response from NV, and may clock the r600 high enough that it wont be beaten by any single card when it launches. Seeing how the current g80 tops out at about 650mhz, you can expect that the higher clocked version, if released, will be clocked somewhere close to those specs.

And the notion of Nvidia already stockpiling the g81 refresh gpu's for the r600 launch has about 99.9% chance of not happening. So the most I expect from NV is cherry-picked g80's, maybe along with new "special" drivers.

If the r600 is a derivative of the never released r400, and the r500, then Ati has been working on it for a long time, and I expect it to pull off some impressive numbers when it's launched. IMO, the 512-bit memory bus is a realistic expectation, along with 96 vec4 unified shaders.

 

[zephyrprime]

Originally posted by: josh6079True. Really though reducing die size generally is a win/win for manufacturers and consumers. We get cooler-running, higher overclockable cores and they get a LOT or dollars saved.

fixed. A 65nm process offers about 2x the transistor density of a 90nm process so the price savings is pretty darn big especially on a die as large as the G80.

 

[Italicized]

Yeah, 90nms to 65 is a pretty big step, more so than 90 to 80.

 

[sethk]

I think at this point we're going to see leapfrogging, as it benefits both companies giving them a short period of dominance during which they can get their launch cards sold at $600+

Won't revisit history, but at this point G80 is the fastest single card out there, the R600 will be the fastest when it is out (although it wont have the sort of lead the G80 has now,) and within a short period of time we'll see the G80 refresh which will in turn be faster, followed by the R600 refresh. Add another round of refreshes after that as well.

This follows the pattern started with the 7800GTX, and I dont see it changing much in this generation, unless ATI really screws up the R600, or somehow manages to very significantly outperform G80 past the point of where they can catch up with a very simple refresh. I find both of those scenarios unlikely.

Another way of stating it is you will always be only a few months away from a faster and / or cheaper card with these leapfrog releases.

 

[schneiderguy]

Originally posted by: redbox
Did ATI launch a 1900 mobile chip?

nope

 

[Cookie Monster]

Can ATi risk using 65nm? Im thinking 80nm more likely for the R600. In a business sense, launching card around december would be very idealistic and appropriate. But releasing this card around january as some sources are saying means its delayed due to several reasons, and one of them could be yields. nVIDIA and ATi does have different cycles, but if R600 is the derivative of R400 and R500, then there is no reason for it to be late.

96vec4 shaders sound impressive. R600 might probably have the edge in theorectical shader performance if that was true. But in a real performance sense, G80s scalar shaders is utlised much more than a vec4 shader. Although vec4s do much more work per cycle, we are talking about 100% utlisation of the scalar shaders in G80. You will need R600 to be clocked around 700mhz to match this. Then even so, the R600 shader design could be vec4+scalar found in the Xenos.

My guess at the R600 is

80nm
>500 million transistors
600~700mhz core clock
Unified Shader Architecture
64 shaders (or 96 shaders)
16 ROPs
16 TMUs (note that ATis current 16 TMUs for R5x0 is much more efficent than G7x 24 TMUs)
512bit memory interface
GDDR4 (or maybe GDDR3)
1GB of vram