6870 VS GTX460 Die sizes and price/performance

[n0x1ous]

We'll see when CS6 releases.

What will we see? AMD current Firepro cards barely beat rv770 based ones in pro applications and get beat by gt200 based cards.

Amd simply doesn't invest time/money/resources in their drivers for this market and thats why they have < 15&#37; share in it.

CS6 won't change that. Like i said, NV already supports open cl/gl better than AMD anyway.

 

[psoomah]

I have read other accounts that the die size of g104 is smaller than what that site 'measured'. I don't have a link handy, also Nvidia has not officially stated (believe) what the actual die size is.
Gpu-z lists its as 332m.

If Nvidia didn't release the die size, GPU-Z is guessing based on ... what?

The machined off heat sink exposes actual irrefutable physical evidence of the die size.

 

[GaiaHunter]

In terms of what? X1900XT/X consumed more power but it was the better performing card and had better image quality. I recommended X1900 series over 7900 every day of the week despite the much larger die size. Die sizes don't sell videocards to consumers. :awe:

In terms of performance/mm^2 and performance/watt. Wasn't what you were speaking when you said ATI had the lead in that since you remember?

And what card did NVIDIA launched after the tiny G71? G80, only one of the most successful cards.

What card did ATI/AMD launched after the gigantic X1900 that was only a bit faster than competition? X2900.

What card did AMD launched after the tiny 3800? 4800 series.

No die sizes don't sell cards. But it seems better for the company that makes them to have better performance/mm^2 than performance crown at the expense of performance/mm^2.

And I don't know why it is that the ATI division operating profits for last quarter was only $1million, but doesn't seem to be because GF100 and GF104 are eating AMD sales.

Curious thing is that this month in the steam survey, NVIDIA GTX460 and up all increased DX11 market share but barely at the expense of 5700 and 5800 market share - -0.22% and -0.02% respectively.

 

[tviceman]

My URL showed the die with the heatsink machined off. That's pretty definitive. The heat sink of the GF 104 is far larger than Cypress'

What's your source?

http://www.semiaccurate.com/2010/07/12/gtx-460-cash-cow-or-market-share-tourniquet/
http://www.brightsideofnews.com/news/2010/8/9/nvidia-fermi-geforce-die-sizes-exposed.aspx

I can keep linking you with more goodness, but I figure GPUz is probably more accurate than Charlie's old estimations - and even his coworkers don't use his figures in their articles. Hmmmmmmmmm.

 

[tviceman]

http://www.amazon.com/HIS-Eyefinity-...7525569&amp;sr=8-1

$275 for a 6870, which is really just the 5770 replacement, is a freaking rip off IMO. Nvidia has no reason to lower prices now as the gtx460 actually delivers the same performance per dollar which is important in this level of card.

Wow $275? If that is really what the MSRP is going to be for vanilla 6870's.... *IF* it is... AMD is just shooting themselves in the foot by leaving the door wide open for Nvidia to release the 384 shader GF104 at $0-15 premium and really put the frames per second per dollar metric in their corner. AMD should be pricing these cards aggressively if they want to continue to turn Nvidia's business upside down. Instead, all they've done since GF104 has come out is leave the door open for customers to go with the better price/performance option.

 

[badb0y]

Wow $275? If that is really what the MSRP is going to be for vanilla 6870's.... *IF* it is... AMD is just shooting themselves in the foot by leaving the door wide open for Nvidia to release the 384 shader GF104 at $0-15 premium and really put the frames per second per dollar metric in their corner. AMD should be pricing these cards aggressively if they want to continue to turn Nvidia's business upside down. Instead, all they've done since GF104 has come out is leave the door open for customers to go with the better price/performance option.

There is a 475? When is the release date?

 

[Silverforce11]

AMD doesn't care to "drive NV to the ground", as long as their market share is dominant, they need to make as much $$ as they can.

 

[Arkaign]

AMD doesn't care to "drive NV to the ground", as long as their market share is dominant, they need to make as much $$ as they can.

Hah, this is exactly what I pointed out earlier in the thread. AMD basically told us in advance that they didn't care to go bananas with a price war, as the naming of their new midrange cards as 68xx means they want to charge 58xx prices for them, while probably getting ready to charge around the $500+ level for their top-end single GPUs in the 69xx range.

I can't blame them at all from a corporate standpoint. The 68xx sounds like it will be relatively cheap to produce, they've already done the bulk of the R&D during the 58xx design, so charging premium for this new midrange product will load their coffers well.

For the consumer, I'm disappointed that there's not really going to be a compelling answer to GTX460, as it's fairly obvious that whatever they finally bring out under the 67xx range won't be a barn-burner and worthy of taking the $175-$200 price point (IMHO, could be wrong, but doubt AMD will canabalize 6850 with something that performs nearly as well for less $).

I'd like to see GTX460 level performance for $125 or so (MSRP, not rebate nonsense), that'd be awesome, but it's just not going to happen for quite some time yet.

 

[OCGuy]

Hah, this is exactly what I pointed out earlier in the thread. AMD basically told us in advance that they didn't care to go bananas with a price war, as the naming of their new midrange cards as 68xx means they want to charge 58xx prices for them, while probably getting ready to charge around the $500+ level for their top-end single GPUs in the 69xx range.

I can't blame them at all from a corporate standpoint. The 68xx sounds like it will be relatively cheap to produce, they've already done the bulk of the R&D during the 58xx design, so charging premium for this new midrange product will load their coffers well.

ATI is selling 90% of the DX11 market, and they can't turn a profit. It isn't that they dont want to lower prices, they really can't. Stockholders come before customers.

 

[GaiaHunter]

For the consumer, I'm disappointed that there's not really going to be a compelling answer to GTX460, as it's fairly obvious that whatever they finally bring out under the 67xx range won't be a barn-burner and worthy of taking the $175-$200 price point (IMHO, could be wrong, but doubt AMD will canabalize 6850 with something that performs nearly as well for less $).

I'd like to see GTX460 level performance for $125 or so (MSRP, not rebate nonsense), that'd be awesome, but it's just not going to happen for quite some time yet.

Have you guys checked the recent Asus TOP GTX 460 1GB review? http://www.hardocp.com/article/2010/10/18/asus_engtx460_top_1gb_video_card_review

Going by this review numbers the 5850, 26.5&#37; 😕 faster than stock GTX 460, at $230-240 is actually price competitive with the GTX 460!

Any insight?

 

[Arkaign]

Have you guys checked the recent Asus TOP GTX 460 1GB review? http://www.hardocp.com/article/2010/10/18/asus_engtx460_top_1gb_video_card_review

Going by this review numbers the 5850, 26.5% 😕 faster than stock GTX 460, at $230-240 is actually price competitive with the GTX 460!

Any insight?

That's interesting, and actually by far the worst I've seen a GTX460 in testing vs. 5850. I'm not gonna call bias, but compare to Anand test :

http://www.anandtech.com/show/3809/nvidias-geforce-gtx-460-the-200-king/18

Unoverclocked, stock 1GB GTX460 (not overpriced Asus TOP card) is very close to stock 5850 performance. Once overclocked, it's solidly between HD5850 and HD5870 stock performance. Yes, you can also overclock 5850s as well, and probably beyond the reasonable range for a GTX460 to match, but the price difference is definitely there.

The article brings up a kind of apples/oranges comparison too, by mentioning rebates on one card but not the other. The 5850 is a great card, but it's soon EOL and it's still a noticeable amount more expensive unless someone flat out overpays for the 460.

 

[thilanliyan]

nm

 

[GaiaHunter]

That's interesting, and actually by far the worst I've seen a GTX460 in testing vs. 5850. I'm not gonna call bias, but compare to Anand test :

http://www.anandtech.com/show/3809/nvidias-geforce-gtx-460-the-200-king/18

Unoverclocked, stock 1GB GTX460 (not overpriced Asus TOP card) is very close to stock 5850 performance. Once overclocked, it's solidly between HD5850 and HD5870 stock performance. Yes, you can also overclock 5850s as well, and probably beyond the reasonable range for a GTX460 to match, but the price difference is definitely there.

The article brings up a kind of apples/oranges comparison too, by mentioning rebates on one card but not the other. The 5850 is a great card, but it's soon EOL and it's still a noticeable amount more expensive unless someone flat out overpays for the 460.

Curiously I think it is the first time I see 5850 vs GTX 460 on [H] other than CF vs SLI where the GTX 460 SLI is a force to be reckoned with, so I'm not sure we can call [H] biased when we have those CF vs SLI articles.

On the other hand the 5850 always performed better in [H] reviews when pitted against a GTX 470 than in most other sites.

Can it be the QX9650 @3.66 GHz having an impact instead of a i7?

But if the 5850 is indeed 26.5% faster than a GTX 460 at stock, $230-250 is just in line with its performance.

 

[Arkaign]

^ Indeed, but it's the first time I've seen a huge gap in performance. Anand's test, along with a handful of others have all shown more like a 5-10&#37; gap, and it's always erased with OC (yes, apples / oranges comparing stock to OC, but it shows the range well).

 

[happy medium]

Have you guys checked the recent Asus TOP GTX 460 1GB review? http://www.hardocp.com/article/2010/10/18/asus_engtx460_top_1gb_video_card_review

Going by this review numbers the 5850, 26.5&#37; 😕 faster than stock GTX 460, at $230-240 is actually price competitive with the GTX 460!

Any insight?

The review is bias and sucks.?

 

[Skurge]

The review is bias and sucks.?

Wait, why is this review bias, but the 460SLI ones aren't?

 

[Arkaign]

Wait, why is this review bias, but the 460SLI ones aren't?

If I had to guess, it's because it wildly contradicts every other major 460 review that also includes 5850.

Do you trust Anand? Because AT review shows STOCK 460 1GB dangerously close to stock 5850, even winning a few benchmarks, and solidly ahead when overclocked.

Either Anand or [H] screwed the pooch, and you can guess whom I trust more.

 

[GaiaHunter]

If I had to guess, it's because it wildly contradicts every other major 460 review that also includes 5850.

Do you trust Anand? Because AT review shows STOCK 460 1GB dangerously close to stock 5850, even winning a few benchmarks, and solidly ahead when overclocked.

Either Anand or [H] screwed the pooch, and you can guess whom I trust more.

Testing methods are not the same though.

I wish more people adopted [H] style so we could have a second/third opinion on that test method - on one hand this result is completely off the charts, on the other hand [H] is one of the main reviewers and while their reviews sometimes disagree a bit with other sites, they seem to be consistent and their method, while having cons, also have merits.

 

[OCGuy]

Testing methods are not the same though.

I wish more people adopted [H] style so we could have a second/third opinion on that test method - on one hand this result is completely off the charts, on the other hand [H] is one of the main reviewers and while their reviews sometimes disagree a bit with other sites, they seem to be consistent and their method, while having cons, also have merits.

Which is why you throw out the lowest and the highest...I've seen comprehensive charts like that around launches.

The problem is claims of performance or "re-testing" after cards have been out a few weeks or months. Tech sites will do it to get page-clicks, and most people are too lazy/indifferent to call them on it if it goes against the previous held consensus.

 

[Mr. Pedantic]

The problem is claims of performance or "re-testing" after cards have been out a few weeks or months. Tech sites will do it to get page-clicks, and most people are too lazy/indifferent to call them on it if it goes against the previous held consensus.

Well, a lot can change after the card is released. c.f. 480. More overclocking potential, less heat, less power, better performance, better multi-GPU scaling, etc.

 

[psoomah]

http://www.semiaccurate.com/2010/07/12/gtx-460-cash-cow-or-market-share-tourniquet/
http://www.brightsideofnews.com/news/2010/8/9/nvidia-fermi-geforce-die-sizes-exposed.aspx

I can keep linking you with more goodness, but I figure GPUz is probably more accurate than Charlie's old estimations - and even his coworkers don't use his figures in their articles. Hmmmmmmmmm.

The S/A article was written on JULY 12 and said "an estimated die size of 320mm2".

The BSN article was written by THEO VALICH, the gold standard of poorly written, inaccurate tech articles.

So I guess you can keep linking with more outdated/3rd rate bullshit links that do not remotely refute an actual physical measurement of an actual 460 die.

Differentiation is a good thing. You might want to give it a try sometime.

Hmmmmmmmmm.

 

[Arkaign]

Well at least AMD and Nvidia both have excellent parts out now, I just wish the prices would fall more quickly.

 

[RussianSensation]

No die sizes don't sell cards. But it seems better for the company that makes them to have better performance/mm^2 than performance crown at the expense of performance/mm^2.

For the company, most definitely. For the consumers it's less clear cut, especially in the enthusiast segment, where outright performance is still important. But ultimately, you have to keep in mind performance/mm^2 or otherwise you'll end up with Prescott :thumbsdown:

Have you guys checked the recent Asus TOP GTX 460 1GB review? http://www.hardocp.com/article/2010/10/18/asus_engtx460_top_1gb_video_card_review

Going by this review numbers the 5850, 26.5&#37; 😕 faster than stock GTX 460, at $230-240 is actually price competitive with the GTX 460!

Any insight?

This review says otherwise. I am not saying one review is better than the other. Just 2 different data points.

At this point, it's fair to say HD6870 > HD5850. So GTX460 1GB has bigger things to worry about.

 

[GaiaHunter]

Which is why you throw out the lowest and the highest...I've seen comprehensive charts like that around launches.

The problem is claims of performance or "re-testing" after cards have been out a few weeks or months. Tech sites will do it to get page-clicks, and most people are too lazy/indifferent to call them on it if it goes against the previous held consensus.

The most annoying thing is that so far [H] only tested GTX 460 vs 5830 and 5770 and never against a 5850.

I made post on their forums about it. Lets see what they answer.

On the other hand I also remember that other article that generated some controversy by showing the 5850 on top of the GTX470 by benchmarking with real game play troughs instead of canned benchs.

And is the consensus right or it that consensus is just formed because most people test it in exactly the same way?

We've tons of sites doing canned benches. I wished a few more would test like [H] or do both - maybe not for release or not for so many games but just to establish control points.

We all compare canned benches results between our systems and canned benches results between tech sites. We rarely talk about gaming experience, except if related to microstutter and some stuff like that, or have the opportunity to play in different systems. And then you read these articles, from someone that most of the time is in line with the consensus ([H] considered GTX460 a great card) but sometimes stuff like this appear and we just have to discard it because of consensus or have to believe at face value with no peer control.

 

[RussianSensation]

And is the consensus right or it that consensus is just formed because most people test it in exactly the same way? We've tons of sites doing canned benches. I wished a few more would test like [H] or do both - maybe not for release or not for so many games but just to establish control points.

HardwareCanucks hates canned benchmarks unless they have run the bench both ways to see if there isn't a material difference. Most of their runs are manual runs.

You can see EVGA GTX460 @ 763mhz trades blows with the HD5850 depending on the game and resolution.
http://www.hardwarecanucks.com/foru...uperclocked-ee-external-exhaust-review-8.html

Actually GTX460 cards run Metro 2033 about 50&#37; slower than a GTX470 and BF:BC2 runs significantly faster on AMD cards. So GTX460 losing in those 2 games to the 5850 at HardOCP isn't a surprise.

The thing about HD5850 is that it was $250+ for the last 12 months. Now that MSI HD5850 is ~ $200, it's certainly a great alternative to the GTX460, if not a better deal. Remember when some GTX470s fell to $200-230 and some GTX480s could be bought for $350-380? Well those deals are long gone too. Price changes really affect what card can be recommended.