open letter to ATi on gamers-depot
- Thread starter chavvdarrr
- Start date
- Status
Apparently you didn't read anything that I wrote.
I never said there wasn't a market it for it . I said that its uninteresting . Why ? Because of the problems with it and the way its implemented as a dual board configuration adn that the two on the same board were much more interesting because it solves some of the problems of the dual board set up .
And yes it isn't interesting in the same way that a faster athlon 64 isn't interesting .
Yes the are both faster . But whats interesting would be an nv50 , or an r520 . Not sli that has been done twice before by two other companys in various formats . One had almost no problems (3dfx) one had a ton of problems (ati rage maxx) and the newest one has a few problems (nvidia)
Please everyone learn to read .
GrapeApe said:
Whats the matter with you? I accepted that I siad something wrong... don't need to talk like an immature brat.
I think you're the one that has to look at the facts about the x800 and all of its derivities, and where they came from. How can you compare a gigabyte 3d1 to a volari duo? like to hear that one. I'm not talking about a technical leap, when did I say anything about that? Very arrogent of you if you assume that I think the x800's are technically backwards.
I said going to a difference process probably won't change the yield issue, even with the power gate thats not going to help them. .11 might improve thier yeilds for the x800xl because they can produce more chips per wafer.
When I say shift to low k the core techonolgy is very similiar through out the entire r300-480 lines. Might have missed the replay to whql.
.15 still will be cheaper then low -k. Low-k costs quite a bit more per wafer even though its .13 approximitly 2 times more expensive. Now .15 to .13 how much space do you save? Is it 50%? I'm thinking they would save around 25%. They should make the 9800's on .13 with no low-k or .11, but then again thats the x700's isn't it, and the x700xt had issues with going past 475 on the core....
ATi has to do more then just changing the fab. They have to change thier architecture it has hit the limit. Great cores but something has to change.
on the x800se, just checked pricegabber, only 3 companys have them. They might be more available in your neck of the woods but not everywhere.
http://www.pricegrabber.com/search_getprod.php/masterid=4773789/search=x800se
and yahoo has 2 one of them was in the pricegrabber.com search
http://shopping.yahoo.com/search;_y...1BHNlYwNzZWFyY2g-?is=1&p=x800+se&did=
Razor1 said:
What “logic†do you base that on? (even though you managed to contradict yourself in the same paragraph) Yields usually improve the long a process / chip is used.
Where are you pulling these calculation from?
What does that matter? 9800 didn’t go anywhere near those speeds.
whql said:
Ok I didn't contradict myself, since the x800xl is on a different process .11 then the others that one alone will have improved yeilds possibly.
I'm guessing on the % save if you know better let me know or if someone knows please answer.
Are you saying from going .15 to .11 should only give a 63 mhz bump on the core? Is that whole lot? I would expect alot more wouldn't you?
the 9800xt was clocked at 412 the x700xt has trouble at 475, so lets say 460 for safe margins. Not much for two process drops, 48 mhz. At 475 the x700xt was just a bit faster then the 6600gt. They need to change something. Now if the x800xl does come out with good yields thats thier answer, but then they need to finish up with thier bridge chip and get some AGP boards out.
A. the new topic/reply icons are PURRRRDY!
B. regarding R480, R430, and yields. a lot of people apparently think R480 won't help yields because it's clocked higher and uses the same process. remember the X800 Pro VIVOs and how they could pretty much all be softmodded into X800 XTPEs? this is why the R430 will help yields. they don't have to bin them anymore. congratulations, you are now seeing available X850 XT PE SE LE CEs or whatever they're called.
C. Razor, please, stop making silly guesses about fabs. like whoa.
B. regarding R480, R430, and yields. a lot of people apparently think R480 won't help yields because it's clocked higher and uses the same process. remember the X800 Pro VIVOs and how they could pretty much all be softmodded into X800 XTPEs? this is why the R430 will help yields. they don't have to bin them anymore. congratulations, you are now seeing available X850 XT PE SE LE CEs or whatever they're called.
C. Razor, please, stop making silly guesses about fabs. like whoa.
The Baron said:
Learn how to speak to people first before you post.... Your name might be the baron but you talk like you have something stuck up your ass
. Sorry for putting right straight and blunt about it. Then what is the die size save? .15 microns to .13 microns is a drop of .2 microns on the LxW since transistors are only on one level. so by going from .15 to .13 I'm guessing .13x.13/.15x15 = 75% the resdiual is the saved space.Answer me this the Vivo's were unlockable yes, but what where thier clocks? I remember alot of people saying they couldn't go past 500, also people who did clock up to the xt pe level burning out.....
What is the reasoning behind "this is why the r430 will improve yields".
You said “I said going to a difference process probably won't change the yield issue†and then said “ .11 might improve thier yeilds for the x800xl because they can produce more chips per wafer.†– that’s a contraction!
Yields will also often increase the longer the process is used and refined.
From what to what? You have no apples to apples comparison since x700 is not 9800, they are different parts with different elements.
What trouble does x700 xt have at 475? There we’re reports of stability issues at 475mhz. There has evidently been a shift away from x700 xt itself, but that doesn’t mean there was trouble running at those speeds.
Here you are comparing a high volume mid end part to a very low volume high end part.
whql said:
Ok I have to get back to work but
where is this high volume mid end part? was it ever released? what was ATi's reasoning behind dropping a good performer for the mid range segment, I would think a 8 pipeline part that performed a bit faster then the 6600gt at the same price would be a likely candidate to stick with? Didn't they say the clocks of 475 was hard to reach on this part?
There were no yield issues for the 9800 lines if there were its no where near as bad as whats happening now with the x800 or x700 on what every process is being used at the desired clocks.
You are right about as the process matures, but making a part in the first place that will have bad yield issues is not the way ATi has worked in the past. They never had issues using .13 low k on the 9600xt did they? They never had this kind of problem with yields in the past also with other chips using new processes but not its happening, with thier high end chips on a tried and tested .13 with low k, and also now on thier higher end mid range chip which never was released on .11.
asicnewbie
Newcomer
Razor1 said:
Actually, TSMC 's 0.15u and 0.18u digital logic process share the same transistor-sizes and design-rules. You are correct in saying the 0.15u process offers higher transistor density, but the correct reason is due to finer metal (interconnect) pitch, and not smaller (L x W) transistor-devices.
TSMC and UMC's 0.15u/0.11u lines were envisioned, developed, and marketed as 'cost-reduced' production lines. To achieve this goal, they limit the level of miniaturization to the metal-layers only, allowing the half-node processes to re-use the foundry's investment in already existing/deployed processing-equipment. I'm sure "true" half-node (0.15u, 0.11u) production lines exist elsewhere in the semiconductor industry (like the dedicated DRAM foundries.)
Razor1 said:
the x800/x800xl are clocked much lower(400mhz) at those speeds it may be easyer to get good yields then at 475mhz.
and the x800 mem is clocked at 700mhz, wich will enable the OEM`s to use cheaper DDR1 memory, instead of 2.0ns GDDR3 wich will lower the production costs for the OEM, and still give better preformance then the 6600GT`s.
Razor1 said:
I accepted that you said many things wrong but I'm not going to ignore it in a direct reply to me, I addressed each point you presented to me that I felt like responding to. If you have a problem with me pointing it out I'd say you're the one being a petulant child.
Why? My statements are correct and describe how ATI transitioned their parts unlike your version, which is the only reason I bother going into any detail. Regardless of your admission, you base your arguments on those statements, the facts contradict those so I'd say it also invalidates your arguments unless you have something else to base them on.
Simple, your saying X=Y allows me to say A=Z, while it may be an extreme, it's quite similar. The 3D1 uses two core, the Volari Duo uses two cores, but they are only slightly different (both are graphics chips, not a graphics chip versus an ethernet controller). You want to discount the differences, then I will too.
How is anyone supposed to decipher what you're talking about? Technical leap, you mention the simplicity of the design and how it isn't a challenge and then stress how entirely "new" the 6800 core is. That show the arrogance or ignorance of your statements. I particullarly loved the magical comment, it best exemplifies that to me.
And this is the crux of my disagreement with you. WTF does the above statement have to do with finding a part to fill the upper part of the mid-range segment? If they are 'great cores' then what's the problem with putting them in that segment? A change in architecture won't do that, and while it may not have the same checkbox features, the different architectures compete pretty evenly in the market place. But changing the chip being used and thus potentially changing the overall production and yields has a great impact on the discussion at hand, because it offers additional benifits. Moving the segment towards R420 cast offs instead of any R3xx cores gives you production benifits, if not necessarily cost savings, and like I mentioned ATI might have to execute this plan at a loss just like the R9500pro. But at least there might be additional benifits with their potentially being more higher end products as well, whereas the Rialto solution has a different impact. These issues relate to the discussion, unlike whether the marchitecture has hit a was or not.
Which means they are still out there, and not a mythical beast as you claim. I don't say that they aren't rarer than many other cards, however they are not unavailable, so your argument once again is an empty one.
If you want to put forth an argument that is based on something other than a combination of misinterpretations, redirections and just plain falsehoods, then there'd be something further to discuss. However, if all you want to do is complain about my dismissing your statements out of hand and without the appropriate level of respect then forget it, I've got better things to do, especially with the airports being snowed in out east and flying tomorrow.
DOGMA1138 said:
Exactly. While both ATI and nV seem to feel obligated to push each of their "top of the fab" parts to extremes, the X800XL and X800 sit at a comfortable position where there are far better and far worse performing chips out there, so they can be clocked at what is likely determined to be the best yield level for a change. Pushing the X700XT as far as it could go is bound to have a negative impact on yields. And if ATI thinks they can't get enough chips at the speeds required, it makes far more sense to bring out a chip with better potential. Getting the best return is the most important, and if they can't produce/sell enough cards there's no point.
Yeah, as we saw with alot of other MFR using TSOP instead of BGA memory, they wil try and squeeze out efficiencies which may allow them to price 'Lite' cards within reach of other consumers. I just hope that people don't get confused/fooled into buying a downgraded version.
GrapeApe said:
dude grow up and get laid or something to get that extra energy out of your system..... good cause I'm not going to respond to your childish remarks directed at me anymore but I will correct your mistakes.
As I said maybe in your neck of the woods they are available but don't make it sound like everyone can get it if they go to the street corner or something.
Exactly like you felt you had to respond to. But your resposes are unexceptable. Sorry.
Razor1 said:
Yeah get laid, that's it. Can't actually progress your argument so it comes down to that, now that's mature of course. Did you wanna try peni$ size, bank accounts, or mothers next?
I don't see a single correction you've made, and definitely not what was to follow.
I didn't make it sound like they were available on every street corner (talk about twisting people's words), and that's not the point, under ATI's stated plan they aren't supposed to be out in retail in any significant number they were supposed to stay primarily OEM. With that consideration they are far more plentiful than not only you make them seem, but than you actually stated as if it were fact; "ATI never really had the x800se" . My pointing out even a single retail card shows that they REALLY have them, and they are actually being made in large enough numbers for partners to make packaging for them. So for this discussion that easily shows that your statement was just BS.
Mine are unacceptable, yours are inexplicable untenable. Even you're 'correction' was in error.
DOGMA1138 said:
This is a possibility, but having 8 more pipelines then the x700's would cause more heat, and drop the clocks down a bit thats what is hurting ATi's yeild, ability to push its core to the desired clocks. Just have to wait and see if thats going to effect the x800xl.
Also its targeted at the 6800 nu market not the 6600gt. 6800nu's are going for 250ish, and the 6800 gts are going for 350ish if the x800xl does show up as an AGP or if you can find a 6800nu pci-e for 300 dollars, so its right in the middle and will take a good deal of the low high end range and mid high end range of nV if they can get it out in good quantities before nV starts dropping prices or introduces another part (both highly unlikely).
Is this going to be an OEM product, or retail, or both? I was in the impression its more for retail and not so much for OEM.
Also the reference cards that were reviewed weren't great overclockers either so I have a doubt it will have similiar yeild issues as the x700xt.
- Status
