ATI partners don't want to reduce prices

[chizow]

Originally posted by: Lonyo

Originally posted by: ExarKun333
Nvidia makes money, AMD loses money...enough said about wafer size and profit margins.

Q2 FY09 (May - July 2008) NVDA Net loss $120mn (most recent thing I found on the NV website)
Q2 FY09 (May - July 2008) AMD Net loss from graphics $38mn. (Direct comparison)
Q3 AMD $47mn profit, Q4 $10mn loss. (Subsequent quarters)

Let's talk more.

The Q2 FY09 loss was due to the 200M charge for notebook GPUs, but prior to that Nvidia had posted something like 10 straight quarters of record profits. AMD on the other hand was going in very much the opposite direction with a similar number of consecutive quarters in losses, a few of which were directly caused by impairments on ATI's book value.

Still its obvious his statement is more accurate than yours. Nvidia has 1 billion in cash sitting around collecting interest. AMD is hacking off bits and pieces and selling them off to stay afloat or divesting itself with foreign investors.

 

[SlowSpyder]

Originally posted by: Lonyo

Originally posted by: ExarKun333
Nvidia makes money, AMD loses money...enough said about wafer size and profit margins.

Q2 FY09 (May - July 2008) NVDA Net loss $120mn (most recent thing I found on the NV website)
Q2 FY09 (May - July 2008) AMD Net loss from graphics $38mn. (Direct comparison)
Q3 AMD $47mn profit, Q4 $10mn loss. (Subsequent quarters)

Let's talk more.

Nvidia didn't do so well last quarter.

Poor sales + expensive to produce cards + very bad economy.

For the record I don't care about financials, whether a company made money or not doesn't really have anything to do with their products being good or bad. But revenue being down 60% I think does say a little something about a companies product line up compared to the competition. Obviously selling price matters for that as well, but so does not being able to see at all.

 

[Leyawiin]

Originally posted by: Termie
I've been watching this pretty carefully. In the last 10 days, the HD4870 1GB (Sapphire) went from $225 to $250 to $190 to $200. In the past 5 days, the HD4870 512MB (Sapphire) went from $150 to $170 to $180 (it sold out immediately at $150, but is back in stock). This may all have been a reaction to the GTS250 pricing announcement last week and the eventual listings on the web...now that they are selling for up to $25 more than they were announced at (GTS 250 1GB is $150-$175 on Newegg), ATI's board partners probably thought "what the heck were we afraid of?" Until all of those GTS250s (aka 9800GTX) cards come down in price to what Nvidia announced them at, I don't think we'll see the HD4870 go down to $150. Most here would agree that the HD4870 should be priced about a GTS250.

Frankly, Nvidia really has to get its act together. There is no reason that a GTS250 1GB should be listed at $175 on Newegg when GTX260s are available for $160-170. The GTX260 is clearly a good deal, and the GTS250 clearly has no place in the market above $150. Same goes for the GTS250 512MB - why oh why are these listed on Newegg from $130-$155? Unfortunately, ATI partners probably took the hint, leading to what seems to be an increase in HD4850 prices over the past week, pushing into 4870 territory.

Advice to anyone looking to buy...wait until this all blows over...

The prices are oscillating pretty wildly at the moment. I just looked at the two cards you listed and there is a $30 to $50 difference between the GTX 260 (216) and the GTS 250 1GB (which is not "aka" the 9800 GTX - try 9800 GTX+). That's before any rebates. As someone said the difference between lower mid range cards and upper midrange cards is pretty small right now and the selection is huge. That category that the HD 4850 and 9800 GTX+/GTS 250 is slipping into irrelevancy. Either get a HD 4830 if you're on a tight budget or HD 4870 if you're not. That $30-40 more or less that the HD 4850 falls into doesn't make it very compelling anymore.

 

[josh6079]

Originally posted by: chizow

Originally posted by: WelshBloke
Apart from the fact that you pulled all those numbers from your arse. :roll:

No, not really. The only numbers I guessed on were the % per SKU...

"Guessed" is a nice word.

...I still based that off of distributions I've seen in various places like Steam survey or Yougamers/3DMark.

How could you when the Steam survey doesn't designate which 4800 derivative is being used?

How when the 4850 was on the list of Most Popular DX10 (SM4.0) Graphics Cards (all time), but the GTX 2** series wasn't? In fact, when looking within past half-year we see both the 4870 and the 4850 on the list: Most Popular Graphics Cards (last 6 months) Absent from both of the Yougamers results was any GTX 2** information.

And the "Popular 3Dmark 06 hardware" link they provide is missing any GTX 2** based cards. But, the 4850 is present yet again.

How could you "guess" any SKU % at all from those sources given that some didn't even have your claimed SKUs on the data list? The ones that did were 4800 series, 1/3 of which doesn't distinguish between derivatives and the other 2/3 that don't even encompass the 4830 in their findings - to which you also guessed yields of.

If you could, please retract those lies or provide the sources that led you to your SKU % guesses.

Number of dice per 300mm wafer are quoted from various websites presumably based on die size...

Can we have links?

...and verified by counting dice on press pictures of complete wafers.

So you're judging which GPUs are going to translate into a 4830/50/70 based on press pictures?

 

[ronnn]

Makes sense to me. Losing money to gain market share is last year. Prices will go up as sales go down, but the weak will go away.

 

[chizow]

Originally posted by: josh6079
"Guessed" is a nice word.

Yes, "guessed" because I don't have actual sales data for both Nvidia and AMD.....

...I still based that off of distributions I've seen in various places like Steam survey or Yougamers/3DMark.

How could you when the Steam survey doesn't designate which 4800 derivative is being used?

How when the 4850 was on the list of Most Popular DX10 (SM4.0) Graphics Cards (all time), but the GTX 2** series wasn't? In fact, when looking within past half-year we see both the 4870 and the 4850 on the list: Most Popular Graphics Cards (last 6 months) Absent from both of the Yougamers results was any GTX 2** information.

And the "Popular 3Dmark 06 hardware" link they provide is missing any GTX 2** based cards. But, the 4850 is present yet again.

LMAO, I mentioned the Steam survey because it explicitly shows the distribution between GTX 260 and 280 under the Video Card Description Header....which happen to closely mirror the distributions I mentioned: 0.91% for 280 (40%) and 1.41% for the 260 (60%).

Also, I'm not sure why you keep comparing the 4850 and 4870 to the GTX 200 series....none of the distributions I mentioned directly compare the different vendors, I'm strictly comparing parts that use the same chips from the same wafers.

But here's the Youmark figures I was referring to, since its obvious you weren't able to find them yourself:

YouGamers 3DMark06 Last 6 months

Again, 260 at 5.53% (62%) and 280 at 3.35% (38%) and 4850 at 8.34% (52%) and 4870 at 7.79% (48%). So ya, my guess was off the 4850 and 4870 are closer to 1:1 with no account for the 4830 which would presumably be selling at far higher volume given its price to performance positioning.

How could you "guess" any SKU % at all from those sources given that some didn't even have your claimed SKUs on the data list? The ones that did were 4800 series, 1/3 of which doesn't distinguish between derivatives and the other 2/3 that don't even encompass the 4830 in their findings - to which you also guessed yields of.

If you could, please retract those lies or provide the sources that led you to your SKU % guesses.

Here's a not-so-friendly tip. You may want to verify and exhaust any potential sources before running your mouth and demanding retractions at the risk of coming off like an incompetent ass. I understand you're probably butthurt from the various verbal ass kickings I've given you over the past few months, but you only exacerbate the problem when you show you're too incompetent to follow the obvious leads I've provided, then have the audacity to demand a retraction while calling me a liar.

Can we have links?

No, someone else can if they ask though. Of course anyone who knows how to use Google probably would've found it already.... (along with the reviews that actually count the number of die on a wafer)

 

[thilanliyan]

nm

 

[Zstream]

Originally posted by: chizow

What are you talking about? Die size relates to the amount of chips per wafer. They do not mix and match a wafer with different GPU's. One wafer is a set of 4870's while the other is something else.

I still do not understand your logic. The science behind it is that if all yields are equal and the die size is 30% less (ATI) then they will generate more GPU's per wafer. Make more GPU's per process thus able to keep cost down.

 

[MarcVenice]

Chizow, you 'look' like you're making a lot of sense, but your economic thinking is WAY off. AMD doesn't decide on a price for a gpu, the one that is used in the HD4850 for example, and then decides to just 'use' 60% of the waver to make gpu's for the HD4850. No, they make a wafer, and try to cut as many chips out of it, and try to rate as many at the highest speeds possible, the HD4870. Then, and only then, do they set a price. Obvisouly both gpu's cost the same to produce, so AMD would like to sell as many with the biggest markup, the HD4870.

Also, you're comparing 1024mb gddr5 to 512mb gddr3, but you should be comparing 896mb gddr3 to 1024mb gddr5. It's probably a given that gddr5 is more expensive though, but we've both no clue if the 256bit memorybus manages to offset that or not. Also, we've no clue what it costs to produce 1 waver for AMD, and what it costs to produce 1 wafer for Nvidia. In the end it's hard to say which is cheaper to produce. What we do know is, that people more knowledgeable, and with better sources have come up with numbers, based on boardpartners who actually have to buy the gpu, and who do so in qauntities of 1000's or more. We also know what nvidia tried to sell their gpu's for, and what amd asked for their HD4870/HD4850 at introduction. If AMD's aren't much cheaper to produce, why would AMD undercut Nvidia by such a big margin? They could have undercut just a little instead?

Speculation etc ...

 

[josh6079]

Originally posted by: chizow
I mentioned the Steam survey because it explicitly shows the distribution between GTX 260 and 280 under the Video Card Description Header....

But not the 4830 from the 4850 to the 4870 nor even the GTX 295 - all of which you guessed % of per wafer.

Furthermore, the steam survey doesn't give us the GPU's population like Yougamers does. It only gives us percentages. Those percentages represent their corresponding GPU's occupancy among all video cards in the survey. Thus, it won't tell you jack when trying to estimate GTX 2** wafer distributions, nor the 4800 distributions.

You didn't use the steam survey to help you formulate those numbers because it won't tell you what you need to know, yet you said you did.

Just because you designate the same source in bold doesn't make it show what it needs to.

...which happen to closely mirror the distributions I mentioned: 0.91% for 280 (40%) and 1.41% for the 260 (60%).

40% for the 280? Your past guess was 20%.

This is an example of what you're doing wrong explained further below.

Also, I'm not sure why you keep comparing the 4850 and 4870 to the GTX 200 series....none of the distributions I mentioned directly compare the different vendors...

Because you "guessed" which 4800 SKU percentages come from a wafer when the source didn't even specify the different SKUs. As far as the steam survey goes, the 4830,50, and 70 are all lumped together. Even if they weren't, it wouldn't tell you their specific populations. So I don't know how you can even use that as a source for your wafer-SKU breakdown.

And no where was *I* directly comparing the two vendors together. I was highlighting when all of the SKUs you made guesses about were or were not present in your sources. Absent from the steam survey results (and all others) were any GTX 295 and 4830 figures which you also made % guesses.

I'm strictly comparing parts that use the same chips from the same wafers.

No, you're not. You're comparing yields from different vendors in hopes of arguing a GTX 260 is the same or cheaper to produce than a 4870.

You're doing so by giving a breakdown of SKU % per wafer:

30% 4830 @ <$100
40% 4850 @ <$150
20% 4870 @ <$200
10% 4870X2 @ <$225 (prorated per core)


In comparison, a 300mm GT200b wafer at 55nm with ~120 dice might be distributed like:

60% GTX 260 @ <$200
20% GTX 285 @ <$350
20% GTX 295 @ <$250 (prorated per core)

while simultaneously providing sources that don't even include figures for all of the SKUs you guessed on. Therefore your guesses stem from two cards per vendor - 4850 and 4870 and the GTX 260 and 280. Beings how you throw three cards into your percentages means you render those calculations void. Might as well have pulled them from an orifice.

But here's the Youmark figures I was referring to, since its obvious you weren't able to find them yourself:

YouGamers 3DMark06 Last 6 months

Again, 260 at 5.53% (62%) and 280 at 3.35% (38%) and 4850 at 8.34% (52%) and 4870 at 7.79% (48%). So ya, my guess was off the 4850 and 4870 are closer to 1:1 with no account for the 4830 which would presumably be selling at far higher volume given its price to performance positioning.

Again, this is why your numbers are off. There are no GTX 295 or HD 4830 results in the samples you chose from. Any calculations you do to discover the % distribution are going to be void once you account for a third SKU.

For instance, if those percentages are to represent the % of SKUs a wafer is divided into, the 4830 wouldn't even exist. According to your numbers, the 4850 would occupy 52% of a wafer while the 4870 would occupy the other 48%. But, your guesses show a 4830 using more of each wafer than a 4870, despite your own evidence lacking any information on it.

The sample populations of the steam survey and Yougamer/3Dmark 06 results are not enough to tell us wafer distribution - for either vendor.

Here's a not-so-friendly tip. You may want to verify and exhaust any potential sources before running your mouth and demanding retractions at the risk of coming off like an incompetent ass. I understand you're probably butthurt from the various verbal ass kickings I've given you over the past few months, but you only exacerbate the problem when you show you're too incompetent to follow the obvious leads I've provided, then have the audacity to demand a retraction while calling me a liar.

You can try and make this a personal issue if your ego demands it, but that would just increase your derailment history.

 

[dreddfunk]

Aren't there really two discussions here? The cost of the GPU itself, and the cost of the completed card? Both of which seem to involve enough factors to make detailed speculations dubious at best, and it seems likely that any arm-chair analysis of GT2xx costs is lacking. I'm sure that there was a reasonable amount of wiggle room left in the margins for Nvidia to drop prices and still make a profit.

If that's Chizow's point, I'd be inclined to grant it.


The only hesitation I have is this: there is some relation between the initial msrp of a product, and the manufacturing cost. The initial pricing of the GT2xx series seems to suggest a higher manufacturing cost than the HD48xx.

Yet that is hardly conclusive. There are, in fact, several reasonable options that could explain the initial pricing disparity between the GT2xx and HD48xx:

1) GT2xx costs were not high, but without competitive products from AMD, Nvidia sought to maximize profits by releasing at an inflated initial msrp, believing the products to be worth the inflated price in the market.

2) GT2xx costs were high, and the initial msrp is a direct reflection of that fact.

3) AMD was simply willing to accept lower initial (and final) margins with the HD48xx series because of market share pressure.

It would be hard to speculate among the three reasons, when it could be, in fact, a combination of all three.

 

[chizow]

Originally posted by: josh6079
But not the 4830 from the 4850 to the 4870 nor even the GTX 295 - all of which you guessed % of per wafer.

Furthermore, the steam survey doesn't give us the GPU's population like Yougamers does. It only gives us percentages. Those percentages represent their corresponding GPU's occupancy among all video cards in the survey. Thus, it won't tell you jack when trying to estimate GTX 2** wafer distributions, nor the 4800 distributions.

You didn't use the steam survey to help you formulate those numbers because it won't tell you what you need to know, yet you said you did.

Just because you designate the same source in bold doesn't make it show what it needs to.

LMAO. No, nowhere did I claim it had exact distributions for every SKU or claim Steam had figures for both AMD and Nvidia. You just demonstrate how incompetant you are with each line of text. Remember that whole part about the distributions being a "guess"?

40% for the 280? Your past guess was 20%.

This is an example of what you're doing wrong explained further below.

No, its an accurate demonstration of high vs. low-end distribution based on pricing, yields, and demand. Obviously high-end will cannibilize from the high-end and low-end will cannibilize on the low-end. Obviously a 40% figure for the 280 or 285 will not remain static as its share is cannibilized by other competing SKUs. Is this making any sense to you? Do we need to break out the pie shaped building blocks to help you understand?

Because you "guessed" which 4800 SKU percentages come from a wafer when the source didn't even specify the different SKUs. As far as the steam survey goes, the 4830,50, and 70 are all lumped together. Even if they weren't, it wouldn't tell you their specific populations. So I don't know how you can even use that as a source for your wafer-SKU breakdown.

No I didn't, you're just too stupid to comprehend what I wrote. At no point did I claim Steam had a distribution break out for ATI parts, try again.

And no where was *I* directly comparing the two vendors together. I was highlighting when all of the SKUs you made guesses about were or were not present in your sources. Absent from the steam survey results (and all others) were any GTX 295 and 4830 figures which you also made % guesses.

Sure you did, you kept claiming that the 4850 and 4870 were all-time greats this and that and how they appeared on some list and the GT2xx didn't or whatever when that was not only incorrect, it was irrelevant.

No, you're not. You're comparing yields from different vendors in hopes of arguing a GTX 260 is the same or cheaper to produce than a 4870.

You're doing so by giving a breakdown of SKU % per wafer:

while simultaneously providing sources that don't even include figures for all of the SKUs you guessed on. Therefore your guesses stem from two cards per vendor - 4850 and 4870 and the GTX 260 and 280. Beings how you throw three cards into your percentages means you render those calculations void. Might as well have pulled them from an orifice.

LMAO, again, I'm not comparing yields from different vendors, they're only relevant to the SKUs they share wafer space with so comparing yields between vendors doesn't matter.

As for guessing on SKUs that aren't included explicitly, of course I did as any of those surveys are simply a snapshot in time and will obviously lag behind current market data. The 4830, GTX 285, and GTX 295 were all released significantly later than the 260, 280, 4850 and 4870. Not only would they need to catch up to become statistically significant, you're relying on the surveys themselves to distinguish between SKUs or vendor ID strings.

Going back to my earlier point, its obvious that certain parts in certain price and performance segments would cannibilize the share of other parts. Its also obvious the lower priced parts will sell more than the high-end parts. Do you honestly think the 4870 at $190-200 is selling more than the $65 4830? These are simple marketing and sales concepts.

Again, this is why your numbers are off. There are no GTX 295 or HD 4830 results in the samples you chose from. Any calculations you do to discover the % distribution are going to be void once you account for a third SKU.

For instance, if those percentages are to represent the % of SKUs a wafer is divided into, the 4830 wouldn't even exist. According to your numbers, the 4850 would occupy 52% of a wafer while the 4870 would occupy the other 48%. But, your guesses show a 4830 using more of each wafer than a 4870, despite your own evidence lacking any information on it.

The sample populations of the steam survey and Yougamer/3Dmark 06 results are not enough to tell us wafer distribution - for either vendor.

ROFL, no, that's where the "guessing" comes in. Its obvious the 295 and 4830 are selling more than 0%, and as such, they would be cannibilizing shares from the other SKUs. I don't see how this isn't obvious. Your arguments about the 4830 not existing because a small sample of 3DMark results don't list it is not only comical, but borderline retarded.

Once again, as I stated at the very beginning, the distributions are simply a guess based on market segments, performance, pricing, various polling sources and common sense lol. Only a pedantic idiot like yourself would attempt to try and draw such literal ties to what was already acknowledged as a guess.

You can try and make this a personal issue if your ego demands it, but that would just increase your derailment history.

Oh its nothing personal, when a person repeatedly demonstrates themselves to be incompetent, as you have, that's how I treat them.

 

[chizow]

Originally posted by: MarcVenice
Chizow, you 'look' like you're making a lot of sense, but your economic thinking is WAY off. AMD doesn't decide on a price for a gpu, the one that is used in the HD4850 for example, and then decides to just 'use' 60% of the waver to make gpu's for the HD4850. No, they make a wafer, and try to cut as many chips out of it, and try to rate as many at the highest speeds possible, the HD4870. Then, and only then, do they set a price. Obvisouly both gpu's cost the same to produce, so AMD would like to sell as many with the biggest markup, the HD4870.

How is this different than what I wrote? They set out to produce as many flawless wafers as possible and from there, downgrade often perfectly functional chips to meet market demands. This is why you see 4850s overclocking to near 4870 speeds, Phenom II X3's magically unlocking into Phenom II X4's, vanilla Core 2 and i7s overclocking as well as the Extreme speeds etc. etc.

They do set a price btw, they set the price for the end-product SKU. Market demand and relative perfomrance determines what % of each wafer goes to which SKU, not necessarily any binning or defects.

Also, you're comparing 1024mb gddr5 to 512mb gddr3, but you should be comparing 896mb gddr3 to 1024mb gddr5. It's probably a given that gddr5 is more expensive though, but we've both no clue if the 256bit memorybus manages to offset that or not. Also, we've no clue what it costs to produce 1 waver for AMD, and what it costs to produce 1 wafer for Nvidia. In the end it's hard to say which is cheaper to produce. What we do know is, that people more knowledgeable, and with better sources have come up with numbers, based on boardpartners who actually have to buy the gpu, and who do so in qauntities of 1000's or more. We also know what nvidia tried to sell their gpu's for, and what amd asked for their HD4870/HD4850 at introduction. If AMD's aren't much cheaper to produce, why would AMD undercut Nvidia by such a big margin? They could have undercut just a little instead?

Speculation etc ...

Yes its obvious GDDR5 is more expensive than GDDR3. Sure a 512-bit board is more expensive than a 256-bit bus. Of course a beefy Vitec inductor is more expensive than no-name analog inductors. When it comes to wafer cost, I'd say volume pricing would certainly favor Nvidia, as they sell about 2x as many GPUs as AMD. Now that they're on the same 55nm process there shouldn't be any conspiracy theories about 65nm being so much more expensive (because they use more dirt-cheap glass and metal). Of course those 65nm fabs are all busy (@25% utilization) producing those really expensive, high-margin parts like Nvidia Ions and Intel Atoms. :laugh:

Sure its a lot of speculation but I think its clearly obvious pointing to die size and claiming one part costs more than another is clearly inaccurate.

 

[chizow]

Originally posted by: Zstream
What are you talking about? Die size relates to the amount of chips per wafer. They do not mix and match a wafer with different GPU's. One wafer is a set of 4870's while the other is something else.

I still do not understand your logic. The science behind it is that if all yields are equal and the die size is 30% less (ATI) then they will generate more GPU's per wafer. Make more GPU's per process thus able to keep cost down.

That's so far from accurate I don't know where to begin, so I won't even bother. Do you really think engineers and manufacturers are telling electrons to work only 70% as hard to produce certain parts? Or that they work really, really hard on 65% of the die, then just mail it in for the rest? LOL.

 

[chizow]

Originally posted by: dreddfunk
Aren't there really two discussions here? The cost of the GPU itself, and the cost of the completed card? Both of which seem to involve enough factors to make detailed speculations dubious at best, and it seems likely that any arm-chair analysis of GT2xx costs is lacking. I'm sure that there was a reasonable amount of wiggle room left in the margins for Nvidia to drop prices and still make a profit.

If that's Chizow's point, I'd be inclined to grant it.


The only hesitation I have is this: there is some relation between the initial msrp of a product, and the manufacturing cost. The initial pricing of the GT2xx series seems to suggest a higher manufacturing cost than the HD48xx.

Yet that is hardly conclusive. There are, in fact, several reasonable options that could explain the initial pricing disparity between the GT2xx and HD48xx:

1) GT2xx costs were not high, but without competitive products from AMD, Nvidia sought to maximize profits by releasing at an inflated initial msrp, believing the products to be worth the inflated price in the market.

2) GT2xx costs were high, and the initial msrp is a direct reflection of that fact.

3) AMD was simply willing to accept lower initial (and final) margins with the HD48xx series because of market share pressure.

It would be hard to speculate among the three reasons, when it could be, in fact, a combination of all three.

Good assessment, I'd say all 3 were certainly in play to varying degrees.

1) There's little doubt Nvidia maximized margins while AMD wasn't competitive, with record profits and margins in that 2 year period before RV770. Their margins, revenue and profit have all declined in the 2 quarters since.

2) Nvidia parts are certainly going to be more expensive to produce, but again, I think there's massive margins built-in to these parts to compensate for low volume. What you see now is Nvidia sacrificing margins for volume in an attempt to increase revenue to try and reach overall profitability.

3) I've always felt this was a mistake on AMD's part and its really hurting them now as their retail price has to be pushing close to their break-even point on the low-end only 9 months after launch. They've essentially priced the 4850 out of the market at $129 when compared to a $149 512MB 4870 with a superior cooler or a sub-$100 4830 that offers 80-90% of its performance.

 

[thilanliyan]

Originally posted by: chizow
Only a pedantic idiot like yourself would attempt to try and draw such literal ties to what was already acknowledged as a guess.

Are personal attacks allowed in the TOS? You can still make your point without resorting to that kind of thing.

 

[chizow]

Originally posted by: thilan29
Are personal attacks allowed in the TOS? You can still make your point without resorting to that kind of thing.

Its not a personal attack, its an observation based on repeated behavior and comments.

 

[Zstream]

Originally posted by: chizow

Originally posted by: Zstream
What are you talking about? Die size relates to the amount of chips per wafer. They do not mix and match a wafer with different GPU's. One wafer is a set of 4870's while the other is something else.

I still do not understand your logic. The science behind it is that if all yields are equal and the die size is 30% less (ATI) then they will generate more GPU's per wafer. Make more GPU's per process thus able to keep cost down.

That's so far from accurate I don't know where to begin, so I won't even bother. Do you really think engineers and manufacturers are telling electrons to work only 70% as hard to produce certain parts? Or that they work really, really hard on 65% of the die, then just mail it in for the rest? LOL.

Huh? The wafer produced is aimed at and meant for 4870 specs. When part of the wafer (single die) they move it over to a 4850. The other wafer is 4830's and they downgrade from that spec. They do this on all parts and chips...

 

[chizow]

Originally posted by: Zstream
Huh? The wafer produced is aimed at and meant for 4870 specs. When part of the wafer (single die) they move it over to a 4850. The other wafer is 4830's and they downgrade from that spec. They do this on all parts and chips...

That's no different than what I wrote, but is very different from your reply.

 

[josh6079]

Originally posted by: chizow
LMAO. No, nowhere did I claim it had exact distributions for every SKU or claim Steam had figures for both AMD and Nvidia.

I'm not blaming you for not having exact distributions.

I'm saying your guesses have a great margin of error.

You tried to defend them when WelshBloke said, "you pulled all those numbers from your arse" by saying they were based off of the Steam survey - a source that tells you absolutely nothing concerning SKU percentages per wafer:

...The only numbers I guessed on were the % per SKU, but I still based that off of distributions I've seen in various places like Steam survey...

Yes. You did not claim Steam had figures for both AMD and nVidia. But you claimed it was one of your sources for displaying such data. Beings how the Steam survey doesn't tell you jack crap about what your data shows, it's clear you were lying or throwing out hardware survey models at a whim to artificially strengthen your "guesses".

You didn't get those numbers from Steam at all. I know exactly where you got them.

You just demonstrate how incompetant you are with each line of text. Remember that whole part about the distributions being a "guess"?

Exactly, so why take issue with WelshBloke's accurate comment?

No, its an accurate demonstration of high vs. low-end distribution based on pricing, yields, and demand. Obviously high-end will cannibilize from the high-end and low-end will cannibilize on the low-end. Obviously a 40% figure for the 280 or 285 will not remain static as its share is cannibilized by other competing SKUs. Is this making any sense to you? Do we need to break out the pie shaped building blocks to help you understand?

No, you don't need to go to great lengths to show your scewed logic.

No I didn't, you're just too stupid to comprehend what I wrote. At no point did I claim Steam had a distribution break out for ATI parts, try again.

No, you didn't claim that. Instead you said your "guesses" were supposedly based in part off of the Steam survey, and in those guesses was a distribution breakdown for ATi parts.

Resorting to ad hominem won't help you at this moment.

Sure you did, you kept claiming that the 4850 and 4870 were all-time greats this and that and how they appeared on some list and the GT2xx didn't or whatever when that was not only incorrect, it was irrelevant.

Where did *I* say they were? I claimed the site had those cards on those particular lists when I checked them. I was merely indicating cards that were both on your SKU-wafter breakdown and their lists.

I'll break it down for you:

Your list had seven cards: 4830, 4850, 4870, 4870x2 and GTX 260, 285, 295:

30% 4830 @ <$100
40% 4850 @ <$150
20% 4870 @ <$200
10% 4870X2 @ <$225 (prorated per core)


In comparison, a 300mm GT200b wafer at 55nm with ~120 dice might be distributed like:

60% GTX 260 @ <$200
20% GTX 285 @ <$350
20% GTX 295 @ <$250 (prorated per core)

Upon going to the sites you based the above figures on:

Yougamers.com "Popular Hardware" section - "Most Popular DX10 (SM4.0) Graphics Cards (all time)"
Yougamers.com "Popular Hardware" section - "Most Popular DX10 (SM4.0) Graphics Cards (last 6 months)"
Note: "This data has been collected from users of the Game-o-Meter tool. Game-o-Meter uses this data to estimate a system's gaming performance."

Again, that is why I mentioned the cards in those contexts. They either were or were not present on both the relevant lists on Yougamers website and our SKU-wafter breakdown.

I then checked their 3Dmark 06 section. When sorting by "All time" you get these results:

"Most Popular DX10 (SM4.0) Graphics Cards (All time)
Note: The data, these statistics are based on, has been collected from all submits of Futuremark's industry standard benchmark 3DMark06.

It is only when sorting to the last 6 months do you even get 4/7 of your cards present from the same source.

"Most Popular DX10 (SM4.0) Graphics Cards (last 6 months).

Well, this is the best it gets, so lets do some chizow-math shall we?

102,470 own a GTX 260 and 61,815 own a GTX 280. So, out of that sample population 164,285 own a GTX 2** card. What are the percents?

102,470 / 164,285 = .6237 x 100 = 62.37% or 62% for the GTX 260.

61,815 / 164,285 = .3763 x 100 = 37.63% or 38% roughly 40% for the GTX 280.

But wait, there's a GTX 295 to consider...

How should I break 62% and 38% up to accommodate? Guess I could just split the GTX 280 in half and give the extra 20% to the GTX 295. Yeah, that looks nice.

Now onto the ATi parts. 154,118 own a 4850 and 144,078 own a 4870. Out of that sample population, what are the percents?

154,118 / 298,196 = .5168 x 100 = 51.68% for the 4850

144,078 / 298,196 = .4831 x 100 = 48.31% for the 4870.

But wait, there's a 4830 to consider...and a 4870x2...

How should I break 51.68% and 48.31% up to accommodate? Guess I could just throw some numbers in the air and see where they land. /toss. Look, 10% on the 4870x2 and 30% on the 4830 leaving 20% on the 4870. Yeah, that looks good...

Yeah, looks good..."Apart from the fact that you pulled all those numbers from your arse."

As for guessing on SKUs that aren't included explicitly, of course I did as any of those surveys are simply a snapshot in time and will obviously lag behind current market data.

Yep, and not only are they a snapshot in time but they are a snapshot in time of a sample population that doesn't reflect the true dispersion of the SKUs. Let alone the percentage those SKUs take up on a single wafer, which is what you were claiming.

The 4830, GTX 285, and GTX 295 were all released significantly later than the 260, 280, 4850 and 4870. Not only would they need to catch up to become statistically significant, you're relying on the surveys themselves to distinguish between SKUs or vendor ID strings.

That doesn't mean that your rearrangement of numbers is valid.

Do you honestly think the 4870 at $190-200 is selling more than the $65 4830? These are simple marketing and sales concepts.

Going by the "Most popular" information on your source, yes.

You can't pick and chose the information you want from Yougamers to fit your "supported" guess.

ROFL, no, that's where the "guessing" comes in. Its obvious the 295 and 4830 are selling more than 0%, and as such, they would be cannibilizing shares from the other SKUs. I don't see how this isn't obvious.

F'n Duh. The point is the page you used to do your SKU-wafer percentages doesn't list any information regarding that SKU. So tampering with its value distorts any calculations you've done.

Your arguments about the 4830 not existing because a small sample of 3DMark results don't list it is not only comical, but borderline retarded.

It's retarded that your reading comprehension led you to believe I said that.

Originally posted by: josh6079
For instance, if those percentages are to represent the % of SKUs a wafer is divided into, the 4830 wouldn't even exist. According to your numbers, the 4850 would occupy 52% of a wafer while the 4870 would occupy the other 48%. But, your guesses show a 4830 using more of each wafer than a 4870, despite your own evidence lacking any information on it.

Of course it exists. Just not in the sample population you used to conjure up SKU-wafer percentages. Hence, any figure you designate to it is completely bogus.

Originally posted by: chizow
Once again, as I stated at the very beginning, the distributions are simply a guess based on market segments, performance, pricing, various polling sources and common sense lol.

Yeah, it was a guess, from 1 page of your supposed "sources" to which you even had to adjust the numbers using nothing but your bias.

This is why WhelshBloke's statement was spot on and your defense of it is nothing more than a defense of your own ego.

Only a pedantic idiot like yourself would attempt to try and draw such literal ties to what was already acknowledged as a guess.

So why did you defend it so? Why disagree with WelshBloke's assertion when you did in fact make up numbers?

 

[AmberClad]

Chizow, you've crossed the line numerous times in this thread with the personal attacks. It's time to calm yourself down, or you're finished in here.

Here's a not-so-friendly tip. You may want to verify and exhaust any potential sources before running your mouth and demanding retractions at the risk of coming off like an incompetent ass. I understand you're probably butthurt from the various verbal ass kickings I've given you over the past few months, but you only exacerbate the problem when you show you're too incompetent to follow the obvious leads I've provided, then have the audacity to demand a retraction while calling me a liar.

You just demonstrate how incompetant you are with each line of text.

No I didn't, you're just too stupid to comprehend what I wrote.

Only a pedantic idiot like yourself would attempt to try and draw such literal ties to what was already acknowledged as a guess.

Oh its nothing personal, when a person repeatedly demonstrates themselves to be incompetent, as you have, that's how I treat them.

AmberClad
Video Moderator

 

[chizow]

Originally posted by: josh6079
I'm not blaming you for not having exact distributions.

I'm saying your guesses have a great margin of error.

You tried to defend them when WelshBloke said, "you pulled all those numbers from your arse" by saying they were based off of the Steam survey - a source that tells you absolutely nothing concerning SKU percentages per wafer:

...The only numbers I guessed on were the % per SKU, but I still based that off of distributions I've seen in various places like Steam survey...

LMAO. Are you kidding me? You're an idiot, this is why no one is going to come to your aid when it may seem as if I'm using ad hominem or personal attacks. The words I use are accurate observations backed by continuous behavior and comments from you that justify the use of such strong verbage.

Didn't someone just get banned for similarly misquoting a source while trying to prove a point? Funny, I could've swore you were actively involved there as well, did you not learn a lesson? Fortunately for you there's no penalty for posting stupid.

To complete the quote:

Originally posted by: chizow

Originally posted by: WelshBloke
Apart from the fact that you pulled all those numbers from your arse. :roll:

No, not really. The only numbers I guessed on were the % per SKU, but I still based that off of distributions I've seen in various places like Steam survey or Yougamers/3DMark. Market prices are what they are and directly translate into margins. Number of dice per 300mm wafer are quoted from various websites presumably based on die size and verified by counting dice on press pictures of complete wafers.

There's no point in responding to the rest of your idiotic ramblings, you've already demonstrated you're too incompetent to understand simple concepts and are just focused on showing us all the limits of your pedantic fixation on the irrelevant. Only a true simpleton would try and draw such literal parallels to claims that were acknowledged to be guesses LMAO.

 

[thilanliyan]

nm

 

[Idontcare]

Originally posted by: chizow

Originally posted by: josh6079
I'm not blaming you for not having exact distributions.

I'm saying your guesses have a great margin of error.

You tried to defend them when WelshBloke said, "you pulled all those numbers from your arse" by saying they were based off of the Steam survey - a source that tells you absolutely nothing concerning SKU percentages per wafer:

<...The only numbers I guessed on were the % per SKU, but I still based that off of distributions I've seen in various places like Steam survey...

LMAO. Are you kidding me? You're an idiot, this is why no one is going to come to your aid when it may seem as if I'm using ad hominem or personal attacks. The words I use are accurate observations backed by continuous behavior and comments from you that justify the use of such strong verbage.

Didn't someone just get banned for similarly misquoting a source while trying to prove a point? Funny, I could've swore you were actively involved there as well, did you not learn a lesson? Fortunately for you there's no penalty for posting stupid.

To complete the quote:

Originally posted by: chizow

Originally posted by: WelshBloke
Apart from the fact that you pulled all those numbers from your arse. :roll:

No, not really. The only numbers I guessed on were the % per SKU, but I still based that off of distributions I've seen in various places like <Steam survey or Yougamers/3DMark. Market prices are what they are and directly translate into margins. Number of dice per 300mm wafer are quoted from various websites presumably based on die size and verified by counting dice on press pictures of complete wafers.

There's no point in responding to the rest of your idiotic ramblings, you've already demonstrated you're too incompetent to understand simple concepts and are just focused on showing us all the limits of your pedantic fixation on the irrelevant. Only a true simpleton would try and draw such literal parallels to claims that were acknowledged to be guesses LMAO.

chizow, dude, wow, I hate to be a debbie-downer and suggest taking a break on the cadence of this to self-destructive rage cycle but seriously man look at this post I'm quoting above and ask yourself in all honesty what value this is adding to the thread and the forum?

I opened this thread seeing thilan29 had started it and bam, mine eyez. Its chizow vs the whole damned forum. To what end dude? To what end? You aren't going to succeed in convincing a single forum goer here that you are right (even if you are) when you are lamblasting folks with this amplitude of rhetoric.

And if you aren't here to add value (because zero value-add is all that comes of posts like this) then why on earth are you wasting your time here? For your sake man, for your time and effort, what are you doing? Think about it.

 

[chizow]

Originally posted by: Idontcare
chizow, dude, wow, I hate to be a debbie-downer and suggest taking a break on the cadence of this to self-destructive rage cycle but seriously man look at this post I'm quoting above and ask yourself in all honesty what value this is adding to the thread and the forum?

I opened this thread seeing thilan29 had started it and bam, mine eyez. Its chizow vs the whole damned forum. To what end dude? To what end? You aren't going to succeed in convincing a single forum goer here that you are right (even if you are) when you are lamblasting folks with this amplitude of rhetoric.

And if you aren't here to add value (because zero value-add is all that comes of posts like this) then why on earth are you wasting your time here? For your sake man, for your time and effort, what are you doing? Think about it.

If the mods see my comments as over the line that's fine, they can ban me if they like but I'm not going to apologize for the comments or refrain from using valid English words that accurately describe the person in question when that person has repeatedly demonstrated my comments are justified.

As to what end, its simple, I'm not going to stifle my opinion or censor myself knowing there's 3-4 pedantic <insert whatever shiny happy term you like> waiting to try and find some insignificant fault in what I've written. If they do so repeatedly I'm not sure how you expect me to react.

In this case, its obvious certain individuals would rather focus on the irrelevant or even go as far as to fabricate quotes rather than focus on my actual points. The words I use may rub some folks with gentler sensitivities the wrong way, but they're certainly not unjustified or inaccurate.



OK we won't ban you but you do have a week off. Amberclad warned you above and you disregarded her warning.


esquared
Anandtech Senior Moderator