GTX460 a green turd? or a wolf in sheeps clothing?

[Scali]

I'm not trying to. Additionally I said at that bracket of the market.

Now I quoted you because you came to the conclusion that claiming a site surely isn't biased against some brand means it is a way of making the world black and white.

My question was, why claiming a site isn't anti-ati is ok, not triggering a response from you, if that means people can present data to show that the 465 is great, even being faster than a 5850, but if you claim a site like guru3d isn't biased (or are you saying guru3d is pro-ATI and anti-NVIDIA?) and show their results which presents the 465 being slower than the 5850 and being on par with the 5830, prompts you to question why are all threads about brand vs brand and questioning the people culture (that makes them compare different competing products from different brands), while you repetitively do so, by comparing GPGPU from NVIDIA to AMD?

I just quoted you because I wanted to add to the discussion. Your post happened to be the last post on that issue at the time. I didn't disagree with anything you said, did I?
As for GPGPU, the brands aren't important. It's not really nVidia vs AMD, but rather Cuda vs OpenCL technology... with the side-notion that AMD doesn't really have its OpenCL sorted out yet. Which is simply a fact, I'm not judging AMD or nVidia on that. I'm just saying how it is. But apparently you are not allowed to speak your mind, or even mention a brandname, because people get all jumpy.
It's impossible to have a meaningful discussion this way, when people constantly rag on you, calling you a fanboy, or demand that you "give credit where credit's due".
I think moderators should work a LOT harder to shut these people up, and get some meaningful discussion going on technology, without the noise of kids who just cheer for their favourite brand without a clue.

 

[GaiaHunter]

I just quoted you because I wanted to add to the discussion. Your post happened to be the last post on that issue at the time. I didn't disagree with anything you said, did I?

When you quote me and then add that declaration of principles I got the impression that it was for me.

As for GPGPU, the brands aren't important. It's not really nVidia vs AMD, but rather Cuda vs OpenCL technology... with the side-notion that AMD doesn't really have its OpenCL sorted out yet. Which is simply a fact, I'm not judging AMD or nVidia on that. I'm just saying how it is. But apparently you are not allowed to speak your mind, or even mention a brandname, because people get all jumpy.

It is the same thing - competing products, so it is only fair to compare them.

It's impossible to have a meaningful discussion this way, when people constantly rag on you, calling you a fanboy, or demand that you "give credit where credit's due".
I think moderators should work a LOT harder to shut these people up, and get some meaningful discussion going on technology, without the noise of kids who just cheer for their favourite brand without a clue.

Well, while it is true there is loads of fanboyism, stating stuff like NVIDIA architecture is more efficient than ATI for gaming, for example as you did in another thread, because it is xxx cuda cores vs yyy stream cores isn't a bit fanboy speech or sounds like it? Why not die size, transistor count, power, price as metrics for efficiency?

Isn't saying that it is harder program to extract full performance of ATi architecture, the same?

 

[MagickMan]

Most likely the same old story: Acceptable performance but too hot, too expensive.

ditto

 

[Keysplayr]

Gaia Hunter and Scali. Find another thread that actually has to do with what you're talking about, or create one. I will not sift through another 10 pages of bickering micropoints that change focus every 11 words. If you two are going to start following each other around with chips on your shoulders, take it to PM's. Talk GTX460, or go.
Anandtech Moderator - Keysplayr

 

[edplayer]

thanks Keysplayr

I think the 5770 gpu is slightly bigger than the 4770. I assume that yields on the 5770 (now) are better than the 4770 at launch. Must be close to the same cost to make the 5770 compared to what the 4770 was at launch and ATi priced that at $99 msrp (though I believe it went up to $110 later).

ATi has a LOT more room to drop prices on their card versus Nvidia re the 460. I do hope that the 460 performs well and Nvidia is aggressive with their pricing. If its power consumption is reasonable (along with great pricing), I'd consider it.

 

[3DVagabond]

As far as nVidia pricing goes, assuming Fermi costs more to manufacture than Evergreen (I think that's pretty much a given) then it's to nVidia's best interest to charge as much as they can for their cards in reference to ATI's. The last thing nVidia can afford is a price war. It would be a blood bath for their bottom line.

 

[cbn]

I thought this was too hilarious of a news/rumor headline to take away from the Hardocp site by not checking it there., so for the complete story, go to the link

heres a cutout from that link


- Kyle


http://hardocp.com/news/2010/06/09/nvidias_immediate_roadmap_rumors

Isn't this chip supposed to compete against 5770 with its 256 CUDA cores?

I thought 465 GTX was considerably stronger?

 

[Dark Shroud]

Isn't this chip supposed to compete against 5770 with its 256 CUDA cores?

I thought 465 GTX was considerably stronger?

The 465 is stronger than the 5770, but the 5770 is a lot cheaper. The 465 supposed to be compeats with the 5850 and the 5830.

http://www.anandtech.com/show/3745/nvidias-geforce-gtx-465

As we stated earlier, this is NVIDIA’s shot at the sub-$300 market, which is currently dominated by the Radeon HD 5850 at $289 and up, and the Radeon HD 5830 at $225 and up.

NVIDIA has built a product to slot in between AMD’s cards in terms of performance rather than taking AMD head-on, and the pricing reflects this.

 

[GaiaHunter]

Isn't this chip supposed to compete against 5770 with its 256 CUDA cores?

I thought 465 GTX was considerably stronger?

Although rumours put the GTX 460 at 240 instead of 256.

http://www.guru3d.com/news/nvidia-gf104-gf106-and-gf108-roadmap-surfaces/

In mid-July we will be seeing a "460" or "GF104" GPU from NVIDIA. It should MSRP around $230 and have very good overclocking potential. The 450 and 455, "GF106" series GPU will hitting the ground in mid-August and will be retailing at $129 - $179 price points, these GPUs will be replacing the 250 series. The 240 will be replaced in August or September by a "GF108."

The most interesting product will of course be the GeForce GTX 460. The specifications as rumored thuis far show that the GPU will make use of all the components available on the GF104, which include 240 CUDA cores, and will feature 768 MB of GDDR5 memory across a 192-bit wide memory interface. That's pretty much everything available.

Sincerely that number is very odd as it would mean half a cluster (?) disabled. Isn't GF 104 exactly GF100 in half? Or are there some significant differences?

Even if it is 256 cuda cores that is like 75% of the 465 shader power - even considering faster clocks should/could end slower than the 5830 and cost more.

So no price drops for us it seems...

 

[3DVagabond]

Although rumours put the GTX 460 at 240 instead of 256.

http://www.guru3d.com/news/nvidia-gf104-gf106-and-gf108-roadmap-surfaces/



Sincerely that number is very odd as it would mean half a cluster (?) disabled. Isn't GF 104 exactly GF100 in half? Or are there some significant differences?

Even if it is 256 cuda cores that is like 75% of the 465 shader power - even considering faster clocks should/could end slower than the 5830 and cost more.

So no price drops for us it seems...

If Hilbert can ascertain the power usage of a video card by subtracting system power usage at idle from system power usage while stressed then he can come up with shader counts by dividing the GTX-480 in half. It doesn't matter that neither of these methods will work.

Seriously, math isn't his strong suit.

 

[evolucion8]

As far as nVidia pricing goes, assuming Fermi costs more to manufacture than Evergreen (I think that's pretty much a given) then it's to nVidia's best interest to charge as much as they can for their cards in reference to ATI's. The last thing nVidia can afford is a price war. It would be a blood bath for their bottom line.

I think that nVidia can do it once again with Fermi, they did it with the GTX 2x0 series and managed to survive pretty well. But shareholders might think different like the card makers.

 

[Seero]

Unlike ATI's evergreen, Nvidia's Fermi doesn't scale down properly. cutting down SMs sounds good, but that is assuming the electricity used will go down too, that isn't the case it seems on 465. The power consumption of 465 is almost the same as 470, so why not simply get a 470?

I believe 460 is not about speed, but power consumption. If the can fix the power problem so it goes down as the number SMs decreases, then it will be a kick. Putting 2x470 for one card is good, but putting 4x460 under a card is also possible.

 

[slayernine]

So the gtx 465 is 15% faster then a 5830?

Is this website anti - ATI? Review done May 31st
They have the gtx 465 even with the 5850.

http://www.pcper.com/article.php?aid=928&type=expert&pid=4

The games they used are some of the same Anand used and they look almost identical?

http://www.anandtech.com/show/3745/nvidias-geforce-gtx-465/11

Conclusion from the Anand review..

Quote"

"Unless you’re going to be gaming with a 30” LCD, NVIDIA has done an appropriate job of pricing the GTX 465 on a pure performance basis. For $280 you can have a GTX 465, or for 8% more you can have a card that performs 8% faster (the 5850). If that’s all you care about, stop here and figure out how much you wish to spend and you’ll be able to figure out which card you want."

The GTX 465 does not measure up to the 5850 at any resolution. So even if you are living with a small monitor the 5850 will offer lower power consumption and depending on the cooler, less noise. If you do have a large monitor you will see 20-25% performance gains in many games. As far as this website goes, I don't consider it strongly opinionated towards ATI or nVidia. People will jump on whatever bandwagon has the best cards overall. That includes: Price, Performance, Drivers, Power Consumption, Heat output, Reliability, and Maturity. So yes I would have to say that most people seem to be putting their money on ATI cards at the moment because they have the edge. But go back a year or so and this was the complete opposite.

 

[evolucion8]

The GTX 465 does not measure up to the 5850 at any resolution. So even if you are living with a small monitor the 5850 will offer lower power consumption and depending on the cooler, less noise. If you do have a large monitor you will see 20-25% performance gains in many games. As far as this website goes, I don't consider it strongly opinionated towards ATI or nVidia. People will jump on whatever bandwagon has the best cards overall. That includes: Price, Performance, Drivers, Power Consumption, Heat output, Reliability, and Maturity. So yes I would have to say that most people seem to be putting their money on ATI cards at the moment because they have the edge. But go back a year or so and this was the complete opposite.

At 2008, ATi had the same position with the HD 4x00 series that has currently with their latest GPU. The only difference is that GTX series consumed slighly less power than ATi's counterpart, and that HD 4870 was able to match quite often the GTX 280, something that the HD 5870 does a little less often compared to the GTX 480. At 2009, it stayed the same with the GTX 285/275 vs HD 4890.

 

[taltamir]

computex is a priest, it casts heals on me!

BTW... green turn is hilarious!

 

[Grooveriding]

At 2008, ATi had the same position with the HD 4x00 series that has currently with their latest GPU. The only difference is that GTX series consumed slighly less power than ATi's counterpart, and that HD 4870 was able to match quite often the GTX 280, something that the HD 5870 does a little less often compared to the GTX 480. At 2009, it stayed the same with the GTX 285/275 vs HD 4890.

Of important note if you want to try this comparison:

Radeon 4870 release date: July 8th 2008
GTX 280 release date: June 17 2008

Radeon 5870 release date: September 21st 2009
GTX 480 release date: April 12 2010

I don't believe you can put the 5870 and 480 into the same release schedule/architecture lifecycle. 5870 released 1 year and 3 months after 4870, 480 released 1 year and 10 months after 280, a seven month disparity.

At this point, nvidia is behind ATI in terms of performance and technology development. With what is normally 1 and a half times a gpu's lifecycle where normally we expect to see a 75% gain over the past generation, nvidia gave a 50% gain. Whereas ATI made the same improvement in 2/3 the time nvidia did.

ATI is currently the leader and nvidia needs to catch up and release a card that matches ATI's next gpu no less than 4 weeks from it's release to regain their lead.

As far as I can remember this is one of the longest stretches where ATI has had the fastest video card on the market and looks to maintain that position with no sign of a dual gpu card from NV on the horizon, beyond rumour.

 

[Scali]

At this point, nvidia is behind ATI in terms of performance and technology development.

Performance: yes.
Technology: not necessarily.
Fermi does have a quite innovative new parallel pipeline design for better tessellation performance.
Fermi also has improvements for GPGPU and capabilities of running multiple shader kernels simultaneously.
The architecture is quite rough around the edges at this point, but it could be that it comes into its own with a die shrink and a bit of optimization.
ATi on the other hand... their architecture is already quite refined... so if nVidia manages to polish up the architecture, ATi will have to go back to the drawing board and come up with something new to counter those extras that Fermi has.

Reminds me a bit of the early days of Pentium 4. The Willamette wasn't very competitive with the Athlon... but Northwood made a huge difference, even though they didn't change all that much to the basic design. Just a bit of polish and a die shrink. AMD needed the new Athlon 64 to beat it.

 

[Grooveriding]

Performance: yes.
Technology: not necessarily.
Fermi does have a quite innovative new parallel pipeline design for better tessellation performance.
Fermi also has improvements for GPGPU and capabilities of running multiple shader kernels simultaneously.
The architecture is quite rough around the edges at this point, but it could be that it comes into its own with a die shrink and a bit of optimization.
ATi on the other hand... their architecture is already quite refined... so if nVidia manages to polish up the architecture, ATi will have to go back to the drawing board and come up with something new to counter those extras that Fermi has.

Reminds me a bit of the early days of Pentium 4. The Willamette wasn't very competitive with the Athlon... but Northwood made a huge difference, even though they didn't change all that much to the basic design. Just a bit of polish and a die shrink. AMD needed the new Athlon 64 to beat it.

I agree with this. Fermi is impressive in what it can deliver in what you call a rough or unpolished state. They could likely pull off what Intel did with Core2 after fixing Fermi's issues.

It's obvious that Fermi bleeds power, is far too hot and should probably have been held back and worked on for another six months, but Nvidia was in no position to wait even longer. They needed to get whatever market share was left to them from what the 5XXX series had already gobbled up. Considering their corporate vision, it is very likely a case of over-ambition in wanting more than they could deliver in an amount of given time.

All we've heard is that ATI is doing something completely new with Northern Islands, we'll have to see how it contends with where Nvidia is at. You could call 5870 the culmination of the 3XXX series, perhaps they will have to repeat the process again with N.I., or maybe it will come out swinging. I'm sure they have an interest in remaining a dominant contender now that they've scored two wins with the 4 and 5 series.

I would say that potential or not, what is put to market is generally where judgment and comparison should fairly be lain. Hence my opinion on ATI having delivered beyond nvidia.

Should be fun to see what comes out next year though, it will be ATI with something totally new and nvidia refining what they currently have.

 

[taltamir]

It's obvious that Fermi bleeds power, is far too hot and should probably have been held back and worked on for another six months, but Nvidia was in no position to wait even longer. They needed to get whatever market share was left to them from what the 5XXX series had already gobbled up. Considering their corporate vision, it is very likely a case of over-ambition in wanting more than they could deliver in an amount of given time.

Pretty much, nvidia has created something truely impressive... but it was released a year early and a year too late at the same time.
A year too late because that is how late it was to the market.
A year too early because that is how much time they should have spent polishing it.

nVidia should have had a different architecture for this gen which is an evolutionary step from their previous gen. And fermi should have come out in 2011, polished and ready to replace it.

but, you know... hindsight is 20/20... and coulda/shoulda/woulda...

 

[Scali]

It's obvious that Fermi bleeds power, is far too hot and should probably have been held back and worked on for another six months, but Nvidia was in no position to wait even longer. They needed to get whatever market share was left to them from what the 5XXX series had already gobbled up. Considering their corporate vision, it is very likely a case of over-ambition in wanting more than they could deliver in an amount of given time.

Yup... I can't wait to see what they do with GTX460. A smaller die, and possibly some early optimizations, it could result in a significantly better power/performance balance.
If the GTX460 shows potential, then Fermi might have a chance with a die shrink.
But if the GTX460 doesn't improve the situation, perhaps it's just not going to happen for Fermi.

On another note, I received an email from the nVidia developers mailing list today... nVidia has OpenCL 1.1 candidate drivers available!