Nvidia crapping their pants over Lucid?

[SunnyD]

Originally posted by: chizow

Originally posted by: NicePants42

Originally posted by: chizow
I don't think Lucid had anything to do with it, their solution also relies on an add-in chip for mult-GPU functionality. From the various blurbs, it was probably due to OEMs not planning to use NF200 on X58 boards at all, which basically forced NV's hand into removing the artificial limitation on X58.

Wait, so you're saying that this decision by nVidia is based on real information concerning the size of their potential market in the next chipset refresh, and not speculation based on a closed tech demo and two vague AT articles?

*facepalm* Brilliant!

Yes it was so obvious that the OP and numerous others came to the conclusion SLI support on X58 was induced by Lucid. :roll:

Also, the bit about OEMs not planning to use NF200 at all is a new development as previous reports indicated board-makers would be implementing SLI via NF200 on X58 at some point.

I fail to understand how people think that Lucid had absolutely nothing in to do with this. Sure it's part may actually be fairly small - NVIDIA may have simply wizened up a bit with regards to income, but Hydra had to be a wake up call too. Up until Lucid proved Hydra works, NVIDIA indeed could say "HAH! No SLI for j00 with our cards!". But that's not the case anymore.

Sure Hydra isn't out yet, but it is disruptive to NVIDIA's plans. And you can bet that Intel told NVIDIA "Don't care if you don't want to give us SLI for free, we have Lucid waiting in the wings, so we're not going to pay you licensing we don't need anyway." So please, don't go saying Hydra had absolutely nothing to do with this decision, because you bet your ass it would make a dent in NVIDIA's wallet.

 

[Wreckage]

Originally posted by: SunnyD



I fail to understand how people think that Lucid had absolutely nothing in to do with this.

Because Lucid's "tech demo" is probably never going to see the light of day.

This is just a deal brokered between Intel and NVIDIA. I doubt they gave 2 shizzles about Lucid.

 

[chizow]

Originally posted by: SunnyD

Originally posted by: chizow

Originally posted by: NicePants42

Originally posted by: chizow
I don't think Lucid had anything to do with it, their solution also relies on an add-in chip for mult-GPU functionality. From the various blurbs, it was probably due to OEMs not planning to use NF200 on X58 boards at all, which basically forced NV's hand into removing the artificial limitation on X58.

Wait, so you're saying that this decision by nVidia is based on real information concerning the size of their potential market in the next chipset refresh, and not speculation based on a closed tech demo and two vague AT articles?

*facepalm* Brilliant!

Yes it was so obvious that the OP and numerous others came to the conclusion SLI support on X58 was induced by Lucid. :roll:

Also, the bit about OEMs not planning to use NF200 at all is a new development as previous reports indicated board-makers would be implementing SLI via NF200 on X58 at some point.

I fail to understand how people think that Lucid had absolutely nothing in to do with this. Sure it's part may actually be fairly small - NVIDIA may have simply wizened up a bit with regards to income, but Hydra had to be a wake up call too. Up until Lucid proved Hydra works, NVIDIA indeed could say "HAH! No SLI for j00 with our cards!". But that's not the case anymore.

Sure Hydra isn't out yet, but it is disruptive to NVIDIA's plans. And you can bet that Intel told NVIDIA "Don't care if you don't want to give us SLI for free, we have Lucid waiting in the wings, so we're not going to pay you licensing we don't need anyway." So please, don't go saying Hydra had absolutely nothing to do with this decision, because you bet your ass it would make a dent in NVIDIA's wallet.

How so? You do realize Hydra 100 is an add-in chip, just like NF200. And since NF200 is really just a glorified PCIE switch, Hydra 100 probably costs more to boot. You're trading in one solution that costs more and requires board space for another, so how would that worry NV who already has its foot in the door with board partners?

Hydra is also years off, Nehalem launches in 2 months. Personally I think Hydra is Intel's solution to make Larrabee a competent GPU since its doubtful it will be competitive on its own merit.

Nvidia's wake up call was learning from their OEM board partners that there was absolutely no movement on plans to integrate an NF200 chip on X58 boards in the 2 months or so since they announced NF200 + X58 SLI compatibility. That leaves them no SLI option on Intel's next-gen CPU platform, which certainly wasn't going to sit well with them as SLI would have no upgrade path.

 

[ajaidevsingh]

I think its 100% possible that SLi via software "X58" will be lower performance as compared to say the next Core i7 Nvidia M/B...

Nvidia ios not going to sit by and let its M/B business go to waste "The Nforce 1 and 2 platforms were gr8 in the old time but the 7xx series is utter cr*p"

 

[Nemesis 1]

Originally posted by: taltamir
um, the hydra chip DOES NOT EXIST YET! it is in development, and it is not ready.
It does not compete with with the NF200 chip, and it will not compete with it this generation, and it certainly has nothing to do with THIS ARTICLE you posted.

This article is about nvidia opening SLI to the X58 natively due to mobo makers not buying thier NF200 chips. Got nothing to do with lucid developing a product that is years away if ever. (although, since intel is funding them, i doubt it is vaporware)

Yes it does exist. It was shown . Its not for sale yet . But according to the IDF report . M/B makers and others will have them befor 2009. For release in 1st, second qt. of O9.

This chip isn't worth debating. If it works its a big deal . If it doesn't no harm no faul.

If it does work . an say you upgrade to Nehalem in Dec . This chip comes out in Feb. Your going to be pissed.

 

[aka1nas]

Originally posted by: ajaidevsingh
I think its 100% possible that SLi via software "X58" will be lower performance as compared to say the next Core i7 Nvidia M/B...

Nvidia ios not going to sit by and let its M/B business go to waste "The Nforce 1 and 2 platforms were gr8 in the old time but the 7xx series is utter cr*p"

The whole point of this discussion is that there won't be I7 nvidia boards because Intel won't license Quickpath to them.

 

[Acanthus]

Originally posted by: chizow

Originally posted by: SunnyD

Originally posted by: chizow

Originally posted by: NicePants42

Originally posted by: chizow
I don't think Lucid had anything to do with it, their solution also relies on an add-in chip for mult-GPU functionality. From the various blurbs, it was probably due to OEMs not planning to use NF200 on X58 boards at all, which basically forced NV's hand into removing the artificial limitation on X58.

Wait, so you're saying that this decision by nVidia is based on real information concerning the size of their potential market in the next chipset refresh, and not speculation based on a closed tech demo and two vague AT articles?

*facepalm* Brilliant!

Yes it was so obvious that the OP and numerous others came to the conclusion SLI support on X58 was induced by Lucid. :roll:

Also, the bit about OEMs not planning to use NF200 at all is a new development as previous reports indicated board-makers would be implementing SLI via NF200 on X58 at some point.

I fail to understand how people think that Lucid had absolutely nothing in to do with this. Sure it's part may actually be fairly small - NVIDIA may have simply wizened up a bit with regards to income, but Hydra had to be a wake up call too. Up until Lucid proved Hydra works, NVIDIA indeed could say "HAH! No SLI for j00 with our cards!". But that's not the case anymore.

Sure Hydra isn't out yet, but it is disruptive to NVIDIA's plans. And you can bet that Intel told NVIDIA "Don't care if you don't want to give us SLI for free, we have Lucid waiting in the wings, so we're not going to pay you licensing we don't need anyway." So please, don't go saying Hydra had absolutely nothing to do with this decision, because you bet your ass it would make a dent in NVIDIA's wallet.

How so? You do realize Hydra 100 is an add-in chip, just like NF200. And since NF200 is really just a glorified PCIE switch, Hydra 100 probably costs more to boot. You're trading in one solution that costs more and requires board space for another, so how would that worry NV who already has its foot in the door with board partners?

Hydra is also years off, Nehalem launches in 2 months. Personally I think Hydra is Intel's solution to make Larrabee a competent GPU since its doubtful it will be competitive on its own merit.

Nvidia's wake up call was learning from their OEM board partners that there was absolutely no movement on plans to integrate an NF200 chip on X58 boards in the 2 months or so since they announced NF200 + X58 SLI compatibility. That leaves them no SLI option on Intel's next-gen CPU platform, which certainly wasn't going to sit well with them as SLI would have no upgrade path.

The Hydra 100 means that nvidia is no longer the exclusive deal for multi GPU with nvidia graphics cards. The Hydra 100 has near 100% scaling with an (seemingly) unlimited number of GPUs. It has been shown to work AND scale better than Nvidia SLI via hardware. It is no coincidence that Intel has a large stake in Lucid... either. Hydra is speculated to go retail in Q2 of '09 which is not "years off".

 

[SunnyD]

Originally posted by: chizow

Originally posted by: SunnyD
I fail to understand how people think that Lucid had absolutely nothing in to do with this. Sure it's part may actually be fairly small - NVIDIA may have simply wizened up a bit with regards to income, but Hydra had to be a wake up call too. Up until Lucid proved Hydra works, NVIDIA indeed could say "HAH! No SLI for j00 with our cards!". But that's not the case anymore.

Sure Hydra isn't out yet, but it is disruptive to NVIDIA's plans. And you can bet that Intel told NVIDIA "Don't care if you don't want to give us SLI for free, we have Lucid waiting in the wings, so we're not going to pay you licensing we don't need anyway." So please, don't go saying Hydra had absolutely nothing to do with this decision, because you bet your ass it would make a dent in NVIDIA's wallet.

How so? You do realize Hydra 100 is an add-in chip, just like NF200. And since NF200 is really just a glorified PCIE switch, Hydra 100 probably costs more to boot. You're trading in one solution that costs more and requires board space for another, so how would that worry NV who already has its foot in the door with board partners?

Hydra is also years off, Nehalem launches in 2 months. Personally I think Hydra is Intel's solution to make Larrabee a competent GPU since its doubtful it will be competitive on its own merit.

Nvidia's wake up call was learning from their OEM board partners that there was absolutely no movement on plans to integrate an NF200 chip on X58 boards in the 2 months or so since they announced NF200 + X58 SLI compatibility. That leaves them no SLI option on Intel's next-gen CPU platform, which certainly wasn't going to sit well with them as SLI would have no upgrade path.

You just don't get it do you. Yes, the Hydra is an add-in chip like the NF200, but given Intel's relationship with Lucid, you can absolutely bet that Hydra would cost Intel and hence Intel OEM's a hell of a lot less than NVIDIA wants for their NV200 chip (isn't the NV200 something on the order of $47 - talk about price gouging). Not to mention Hydra would likely also mean that boards wouldn't have to go through the SLI certification process, saving yet another undisclosed cost to board manufacturers. I am certain that Intel would leverage its relationship with Lucid to bring Hydra at an insanely low cost to hit NVIDIA where it hurts.

Also, Hydra is hardly "years off". It has already been demoed successfully, and while it won't be in time for Nehealm's launch, I would venture a guess at this stage that it will be out within a year. Sure, that's not tomorrow, but it's also not "years off".

I do agree with you though, NVIDIA did get slapped in the face by the OEM's which helped this SLI issue along some. But if SLI was the best way, and there was no other options, NVIDIA being typically stubborn would have waited it out a bit longer instead of relenting so easily. As I said, it came down to the fact that NVIDIA has everything to loose after all the cards were laid on the table. No one other than NVIDIA's executives can really say what prompted them to allow SLI on the X58 in the end, and as history indicates they would likely not even give you the real reasons anyway. But in the end, it's about $$$, no CSI/QPI license, and Hydra.

 

[Nemesis 1]

This is a bit shocking to myself. Why wouldn't anyone want Lucid not to succeed? IF this chip does what it promises. Its a hugh step forward. Only insane fanbois would not want HYDRA.

Cost of hydra chip. Don't know. But I would pay $100+ for a M/B with it on . Its worth that in added value easily.

 

[Creig]

While Hydra does seem intruiging, it still has a long way to go before it proves itself the successor to either SLI or Crossfire. The data we've seen so far is EXTREMELY limited and there's no guarantee that Hydra will ever make it to the production line, much less dominate the multi-GPU compositing category.

 

[chizow]

Originally posted by: Acanthus
The Hydra 100 means that nvidia is no longer the exclusive deal for multi GPU with nvidia graphics cards. The Hydra 100 has near 100% scaling with an (seemingly) unlimited number of GPUs. It has been shown to work AND scale better than Nvidia SLI via hardware. It is no coincidence that Intel has a large stake in Lucid... either. Hydra is speculated to go retail in Q2 of '09 which is not "years off".

But the point is NV isn't worried about exclusivity, they're worried about exclusion from Nehalem. The whole point of clinging to SLI and tying it to their chipset business was to sell more of both. It was made clear to them months ago that they would not get a chipset compatible with Bloomfield and made clear to them recently no one was putting NF200 on their X58 boards to enable SLI. At that point it became clear the only way to ensure SLI's future on the industry-leading platform going forward would be to remove artificial SLI restrictions.

And yes I was mistaken on Hydra's release, Q1 2009, so 4-10 months away compared to 2 months away for Nehalem......Do you really think they were worried about a part that might not even be relevant until Nehalem's refresh, Westmere, when they didn't have a Nehalem SLI solution which is going to be released in a few months?

Originally posted by: SunnyD
You just don't get it do you. Yes, the Hydra is an add-in chip like the NF200, but given Intel's relationship with Lucid, you can absolutely bet that Hydra would cost Intel and hence Intel OEM's a hell of a lot less than NVIDIA wants for their NV200 chip (isn't the NV200 something on the order of $47 - talk about price gouging). Not to mention Hydra would likely also mean that boards wouldn't have to go through the SLI certification process, saving yet another undisclosed cost to board manufacturers. I am certain that Intel would leverage its relationship with Lucid to bring Hydra at an insanely low cost to hit NVIDIA where it hurts.

Also, Hydra is hardly "years off". It has already been demoed successfully, and while it won't be in time for Nehealm's launch, I would venture a guess at this stage that it will be out within a year. Sure, that's not tomorrow, but it's also not "years off".

I do agree with you though, NVIDIA did get slapped in the face by the OEM's which helped this SLI issue along some. But if SLI was the best way, and there was no other options, NVIDIA being typically stubborn would have waited it out a bit longer instead of relenting so easily. As I said, it came down to the fact that NVIDIA has everything to loose after all the cards were laid on the table. No one other than NVIDIA's executives can really say what prompted them to allow SLI on the X58 in the end, and as history indicates they would likely not even give you the real reasons anyway. But in the end, it's about $$$, no CSI/QPI license, and Hydra.

Get what? Honestly this is about as close to a slam dunk as you get....various blurbs about NF200 priced it at $30, but given the propensity for people here to grossly inflate NV chip prices ($150-200 on GT200 anyone?), let's just call it $100. :roll: I'd expect anything produced by Lucid to cost as much as NF200 given the fact its proprietary tech (and not just a PCIE switch) and the result of years of R&D.

Again, its just trading one add-in solution for another, except SLI is native and NV already has strong ties to board makers. There is no need for the NF200 chip now and while NV may have to validate SLI Nehalem boards, they certainly would've anyways even if there was an NF200 chip attached. So really, do you think NV is worried about an unproven part some 4-9 months down the pipe when they didn't have a solution for their SLI flagship on Intel's high-end platform launch only 2 months away?

So now what's the end-game? When Hydra launches it can enable multi-GPU configurations on....boards that support multi-GPU configurations? Wow. Sounds awful redundant doesn't it? The only way it will be worthwhile is if it performs better than native multi-GPU configurations.

Originally posted by: Nemesis 1
This is a bit shocking to myself. Why wouldn't anyone want Lucid not to succeed? IF this chip does what it promises. Its a hugh step forward. Only insane fanbois would not want HYDRA.

Cost of hydra chip. Don't know. But I would pay $100+ for a M/B with it on . Its worth that in added value easily.

Who's saying they don't want Lucid to succeed? This thread is about why NV enabled SLI on X58. The question for Lucid is if they can deliver as promised without any serious issues to justify the cost, as their Hydra 100 just became the new NF200.

 

[nitromullet]

Originally posted by: Nemesis 1
This is a bit shocking to myself. Why wouldn't anyone want Lucid not to succeed? IF this chip does what it promises. Its a hugh step forward. Only insane fanbois would not want HYDRA.

Cost of hydra chip. Don't know. But I would pay $100+ for a M/B with it on . Its worth that in added value easily.

This is tough to say. A potential issue I see with having a Hydra chip on the motherboard (provided that it works as advertised), is that it claims to basically intercept the DirectX instructions before they every reach the video driver and apply a load balancing algorithm. So, I'm thinking that the Hydra chip has to be compatible with the same level of DirectX as the video cards. I don't know about most people, but I upgrade and replace my video cards generally more often then I replace my motherboard. If I buy a motherboard with a Hydra chip when it launches with DX10 support, will it still work with future revisions of DX10 or DX11 that should be part of Windows 7?

 

[aka1nas]

Originally posted by: nitromullet

Originally posted by: Nemesis 1
This is a bit shocking to myself. Why wouldn't anyone want Lucid not to succeed? IF this chip does what it promises. Its a hugh step forward. Only insane fanbois would not want HYDRA.

Cost of hydra chip. Don't know. But I would pay $100+ for a M/B with it on . Its worth that in added value easily.

This is tough to say. A potential issue I see with having a Hydra chip on the motherboard (provided that it works as advertised), is that it claims to basically intercept the DirectX instructions before they every reach the video driver and apply a load balancing algorithm. So, I'm thinking that the Hydra chip has to be compatible with the same level of DirectX as the video cards. I don't know about most people, but I upgrade and replace my video cards generally more often then I replace my motherboard. If I buy a motherboard with a Hydra chip when it launches with DX10 support, will it still work with future revisions of DX10 or DX11 that should be part of Windows 7?

For a device like Hydra to actually work with non-matching cards from different generations and vendors, it's going to have to treat them as generic DirectX devices, without any of the vendor-specific stuff like the custom AA modes. Otherwise your frames wouldn't match up as the ISVs implement a lot of features differently. I don't even want to think of what they would do with non-matching DX generations.

 

[nitromullet]

Originally posted by: aka1nas

Originally posted by: nitromullet

Originally posted by: Nemesis 1
This is a bit shocking to myself. Why wouldn't anyone want Lucid not to succeed? IF this chip does what it promises. Its a hugh step forward. Only insane fanbois would not want HYDRA.

Cost of hydra chip. Don't know. But I would pay $100+ for a M/B with it on . Its worth that in added value easily.

This is tough to say. A potential issue I see with having a Hydra chip on the motherboard (provided that it works as advertised), is that it claims to basically intercept the DirectX instructions before they every reach the video driver and apply a load balancing algorithm. So, I'm thinking that the Hydra chip has to be compatible with the same level of DirectX as the video cards. I don't know about most people, but I upgrade and replace my video cards generally more often then I replace my motherboard. If I buy a motherboard with a Hydra chip when it launches with DX10 support, will it still work with future revisions of DX10 or DX11 that should be part of Windows 7?

For a device like Hydra to actually work with non-matching cards from different generations and vendors, it's going to have to treat them as generic DirectX devices, without any of the vendor-specific stuff like the custom AA modes. Otherwise your frames wouldn't match up as the ISVs implement a lot of features differently. I don't even want to think of what they would do with non-matching DX generations.

IIRC, Hydra doesn't work with mis-matched vendors exactly for the reasons you described. They did claim that different generation card would work, but they didn't say at what level of compatibility. If you run say an X1950XT and an HD 4850, you might be limited to DX9 which is the highest common denominator, although I don't think the Hydra team commented on that. They might not know how to handle that yet, as IIRC Hydra so far only works with DX9 and they are working on DX10.

The main point of my comment that IMO a chip like Hydra would probably be more attractive as part of a multi-gpu video card that has the support of the card's vendor for Hydra+gpus, as opposed to being a chip that was present on the motherboard. Plus, that way if you decided to go with a single gpu you wouldn't need to pay for the Hydra chip.

 

[taltamir]

anyone has some links for hydra? I mean, i am hearing now that the chip is ready, demoed, and slated for early in Q1 09.
Since the whole premise of hydra is one chip that can do 100% scaling on both nvidia and AMD...

BTW, nvidia ceo said it is impossible, beacause some tasks have to be done in all GPUs...
What if the hydra chip does them, and then just forwards other, simpler tasks to the various cards? Basically it will be a GPU with the ability to offload some specific tasks to other GPUs, nvidia's and AMDs... (that is just a guess i just thought about, not what i heard)

 

[SSChevy2001]

Originally posted by: taltamir
anyone has some links for hydra? I mean, i am hearing now that the chip is ready, demoed, and slated for early in Q1 09.
Since the whole premise of hydra is one chip that can do 100% scaling on both nvidia and AMD...

BTW, nvidia ceo said it is impossible, beacause some tasks have to be done in all GPUs...
What if the hydra chip does them, and then just forwards other, simpler tasks to the various cards? Basically it will be a GPU with the ability to offload some specific tasks to other GPUs, nvidia's and AMDs... (that is just a guess i just thought about, not what i heard)

It's not ready it and is only being demo on DX9. Hopefully it should be done soon.

While we can't give specific performance numbers, since the private demonstration we were given didn't include a benchmarking session, we did observe a consistent 58-60 fps in Crysis on a pair of NVIDIA GeForce 9800GTX cards, with all the game's graphics settings pegged at ?high' under DirectX 9. So the HYDRA Engine works, delivering nearly double the single-card performance figures - and we've seen it with our own eyes.

http://techgage.com/article/lu...multi-gpu_technology/1

 

[taltamir]

ok so the people on this thread that keep on saying it is ready, and being integrated into mobos as we speak are full of it. Thanks SSChevy.

But a working tech demo, even on DX9, is impressive. Especially considering the nvidia CEO says it is IMPOSSIBLE to make such a chip (although, the issues he cited might be DX10 limited)

 

[Acanthus]

Originally posted by: taltamir
ok so the people on this thread that keep on saying it is ready, and being integrated into mobos as we speak are full of it. Thanks SSChevy.

But a working tech demo, even on DX9, is impressive. Especially considering the nvidia CEO says it is IMPOSSIBLE to make such a chip (although, the issues he cited might be DX10 limited)

I dont believe anyone said the chip is being integrated now.

It is still alpha hardware, but is rapidly approaching beta and an eventual launch in 1H '09. (some sources say Q1, others Q2)

 

[Nemesis 1]

Here is a link. I am also going to bump the Hydra thread back to front page.


http://www.pcper.com/article.p...=607&type=expert&pid=1

 

[emilyek]

I'm just waiting for Nehalem plus 2x Larrabee cards via Hydra, on which I can play Project Offset.

And yeah, Intel Capital helps funds Lucid.

 

[wwswimming]

nVidia will probably end up selling more video cards, and their
profits will go up. they may even tell us this was their idea in
the first place.

buy nVidia so nVidian's can work !