You mean texture shimmering coming back in nv side?
Why don't you test a G80 yourself in various scenarios with and without various optimisations in the driver CP and tell me how much you gain?
The LAST R600 Rumours & Speculation Thread
- Thread starter Geo
- Start date
- Status
Why don't you test a G80 yourself in various scenarios with and without various optimisations in the driver CP and tell me how much you gain?
Yet it would make no difference to card in which they kept the ALU to ROP/texture unit balance.
chavvdarrr
Veteran
ATI actually scheduled an Editors' Day and then cancelled flight and hotel bookings that had already been made. That suggests two things:
1) They found out something bad.
2) Whatever it was they found out took them completely by surprise.
Whatever it was, just a couple of days ago they still believed they were going to launch on time; then something happened that forced them suddenly to change their minds. This leads me to believe that it is not simply a question of yields and bins - unless it's a really bad problem with yields and bins - because a problem like that is something they would have seen coming a week or two earlier. I think it has to be something more significant than that: an actual bug in the hardware that needs a respin, a major problem with the card or cooler design, or (less likely) a really intractable problem with the drivers that causes crashes and will take several weeks to fix.
I simply don't buy the "strategic" line; I think that's desperate backside-covering by the PR team. It's perfectly possible that AMD might decide to hold back the launch to coincide with middle- and low-range parts, or even to coincide with the launch of the K10 CPU; but that's not something they would suddenly change their minds about at the last minute!
This cancellation also suggests that we are not talking about a two-week delay. They wouldn't have cancelled everybody's flights over that - they would have held the meeting as planned and announced a slightly later launch. I think this points to a delay of at least a month.
1) They found out something bad.
2) Whatever it was they found out took them completely by surprise.
Whatever it was, just a couple of days ago they still believed they were going to launch on time; then something happened that forced them suddenly to change their minds. This leads me to believe that it is not simply a question of yields and bins - unless it's a really bad problem with yields and bins
- because a problem like that is something they would have seen coming a week or two earlier. I think it has to be something more significant than that: an actual bug in the hardware that needs a respin, a major problem with the card or cooler design, or (less likely) a really intractable problem with the drivers that causes crashes and will take several weeks to fix. I simply don't buy the "strategic" line; I think that's desperate backside-covering by the PR team. It's perfectly possible that AMD might decide to hold back the launch to coincide with middle- and low-range parts, or even to coincide with the launch of the K10 CPU; but that's not something they would suddenly change their minds about at the last minute!
This cancellation also suggests that we are not talking about a two-week delay. They wouldn't have cancelled everybody's flights over that - they would have held the meeting as planned and announced a slightly later launch. I think this points to a delay of at least a month.
Who made the call on the day for this abortive Editors' Day?
Is it the same group of people who canceled it?
More directly, was it the ATI-AMD folks who made the initial call and the AMD-AMD folks who cancelled, or is the graphics group still relatively independent?
Is it the same group of people who canceled it?
More directly, was it the ATI-AMD folks who made the initial call and the AMD-AMD folks who cancelled, or is the graphics group still relatively independent?
caboosemoose
Regular
In isolation this delay isn't all that big a deal. But when you factor in the R520 delay and the weakness of the X1600 at launch, the fact that ATI has essentially handed NVIDIA the advantage in mobile discreet graphics and of course the fact that the last thing ATI needs to do is screw up major product launches when its trying to streamline itself with AMD...well, taking a pretty cool headed view you still get a fairly worrying overall story.
Incidentally:
AMD desktop quad-core is late '07, not mid '08.
Incidentally:
AMD desktop quad-core is late '07, not mid '08.
I have to agree with nicolasb's post. I am just so disappointed. I expect R600 in the May/June timeframe as well. My prediction is there will be a refresh from Nvidia right after that comes out that will negate the x2900. Nvidia has been very aggressive and that wont surprise me at all. The nvidia card will prob have a gig of gddr4 for starters.
Edit:
Oh and another thing. Whatever advantage AMD/ATI might be claiming with their K10 and R600 simultaneous launch prob got its thunder stolen because of the Intel 45 nm Penryn announcement that they are launching it end of Q3 I believe. a Quad core 3 ghz 1600 mhz PSB is going to be their flagship.
Edit:
Oh and another thing. Whatever advantage AMD/ATI might be claiming with their K10 and R600 simultaneous launch prob got its thunder stolen because of the Intel 45 nm Penryn announcement that they are launching it end of Q3 I believe. a Quad core 3 ghz 1600 mhz PSB is going to be their flagship.
ATI actually scheduled an Editors' Day and then cancelled flight and hotel bookings that had already been made. That suggests two things:
1) They found out something bad.
2) Whatever it was they found out took them completely by surprise.
Whatever it was, just a couple of days ago they still believed they were going to launch on time; then something happened that forced them suddenly to change their minds. This leads me to believe that it is not simply a question of yields and bins - unless it's a really bad problem with yields and bins
I simply don't buy the "strategic" line; I think that's desperate backside-covering by the PR team. It's perfectly possible that AMD might decide to hold back the launch to coincide with middle- and low-range parts, or even to coincide with the launch of the K10 CPU; but that's not something they would suddenly change their minds about at the last minute!
This cancellation also suggests that we are not talking about a two-week delay. They wouldn't have cancelled everybody's flights over that - they would have held the meeting as planned and announced a slightly later launch. I think this points to a delay of at least a month.
My guess is that they found out that 8300/8600 are superior to their 2300/2600 offerings and decided to cover this fact by launching 2900/2600/2300 together, so that most headlines would point to the fact that 2900 beats 8800 and kind of skipping the other results in the headlines and in the minds of most customers.
That doesn't make sense either.
It would steal the limelight from the hypothetical R600 performance advantage, compared to the G80.
Meaning, the reviews would be dispersed by too many different products, and the initial marketing impact of R600 would be smaller.
Nvidia was able to do that for G71/G72/G73 because the first one of these was a refresh of the top end without any new performance/IQ features other than clockspeed, not a completely new product like the mainstream and low end variants.
This time, R600 is not out yet, so it needs "some time alone" in the market, until the public perception of the brand new technology gains some "roots".
Of course, i'm no marketing expert...
frameavenger
Newcomer
vertex_shader
Banned
Why do i get the feeling the Inquirer is printing everything they read here, even if it's nothing more than wild guesses ?
Even the 512bit (note the 512Mbit bus, they wish...
When you have a large chip, it's easier to have a wide memory bus, not more difficult.
The funny thing is,vrzone using inq as source.
Next time on inqurier (season6-episode91) "AMD skip 80nm process with R600 and go for 65nm process", "AMD skip R600 and go straight for R700" "AMD skip the 2007 year and go straight for 2009"
That doesn't say much for their current drivers if they do. NV30 took about 8 months for them to get their shader compiler going while ATI had theirs right out of the gate almost.
vertex_shader
Banned
HKEPC joined for the r600 rumor squad
"According to Taiwan's consumer mapping industry revealed AMD将 mapping industry in the end of February to provide samples of the latest version of R600 mapping results, CeBIT in Germany and the General Assembly on public display, but this is not an official release date R600. Therefore, the General Assembly and test results would not be made public R600, AMD tentatively scheduled for March 30 will be formally announced."
"In addition, the R600 will be times when the two lower models, including R600XT and R600XL. R600XT code Catseye (102-B006) will be set up XTX version of the same PCB Subtotal, core connections will decline 10-15% over the XTX version, the memory will be reduced to 512MB. and the switch to a lower speed GDDR3 particles cards for the same 9.5-inch long, the maximum power the same as pumped. expected to be released on April 19."
"According to mapping industry, R600 doubt on the effectiveness of the GeForce 8800 family can be overwhelming. However, the cost and price and more than able to compete with rivals into doubt R600 because of the cost of memory is more expensive than rivals."
http://www.google.com/translate?u=http%3A%2F%2Fwww.hkepc.com%2Fbbs%2Fnews.php%3Ftid%3D746258%26starttime%3D0%26endtime%3D0&langpair=zh%7Cen&hl=en&ie=UTF8
This "news" dates 2007-02-22, so HKPEC news writers sleep well yesterday, or someone made a bad joke about the whole delay story.
The infos in the link sounds like "Fuad inq infos"
"According to Taiwan's consumer mapping industry revealed AMD将 mapping industry in the end of February to provide samples of the latest version of R600 mapping results, CeBIT in Germany and the General Assembly on public display, but this is not an official release date R600. Therefore, the General Assembly and test results would not be made public R600, AMD tentatively scheduled for March 30 will be formally announced."
"In addition, the R600 will be times when the two lower models, including R600XT and R600XL. R600XT code Catseye (102-B006) will be set up XTX version of the same PCB Subtotal, core connections will decline 10-15% over the XTX version, the memory will be reduced to 512MB. and the switch to a lower speed GDDR3 particles cards for the same 9.5-inch long, the maximum power the same as pumped. expected to be released on April 19."
"According to mapping industry, R600 doubt on the effectiveness of the GeForce 8800 family can be overwhelming. However, the cost and price and more than able to compete with rivals into doubt R600 because of the cost of memory is more expensive than rivals."
http://www.google.com/translate?u=http%3A%2F%2Fwww.hkepc.com%2Fbbs%2Fnews.php%3Ftid%3D746258%26starttime%3D0%26endtime%3D0&langpair=zh%7Cen&hl=en&ie=UTF8
This "news" dates 2007-02-22, so HKPEC news writers sleep well yesterday, or someone made a bad joke about the whole delay story.
The infos in the link sounds like "Fuad inq infos"
They also admitted at the release of the NV40 that they really borked the shader instruction set, making it exceedingly hard to develop a good compiler. Note that NV40 had its shader compiler up and running well right out of the gate, too.
- Status