Nvidia ready to counter AMD`s Mantle

[Piroko]

Sad days, but the way of the future, and turbos do have some good characteristics. BUT, give me a screaming naturally aspirated engine for my "Sunday" driver.

Imho, a small car with a small turbo engine can be a huge laugh to drive if it's done properly, screaming and screeching around corners all while still remaining at legal speeds. Much more fun than the small NAs or hybrids I drove at least. And the larger engines are just overkill for DDs


Anyways, Gsync caught my interest, at least moreso than improving the performance of OpenGL (not ES?). Not that the latter one is something bad, it's just not very catchy either, even with the prospect of OpenGL powered Steam boxes.

 

[desprado]

Why you keep posting whatever those guys say when they're on Nv's payroll?

AMD GOD
Johan Andresson was also there he had also commented.

 

[tential]

Yes. I've often stewed about something. Why, after all the years AMD and Nvidia, or AMD and Intel have been designing silicon, has not one of the companies ran away with performance? You'd think that over a period of a few decades, one company would vastly overshadow another, not ALWAYS be ridiculously neck and neck especially when AMD for example has next to no funds. They can STILL offer a Titan competitor, when Nvidia has relatively unlimited funds to utterly wipe the floor with AMD. Same with Intel. AMD does not offer a faster CPU than Intels fastest, BUT is STILL able to keep up a few tiers lower with next to no money for R&D and such. When I say next to no money, I'm talking in relative terms compared to Intel or Nvidia.

Seems quite impossible to me and the only explanation that I can wrap my mind around is that they actually are all working together, despite public appearances.
Something isn't right. Hasn't been for many many many years. Only in the last few have I suspected anything like collusion between these companies.

I'm sick (Not really. Just sick of it) over the whole industry. My views have changed pretty radically over the last few years. Anyone ever get this feeling when they see year after year, gen after gen, that one company does not continually walk away from the other? Does not make sense. Even two turtles racing over a 100 mile distance will cross a given finish line at completely different times. Could be a week apart, could be a month, or a year. Best analogy I can come up with on a half a cup of coffee.

Good morning to all, and have a great day!

I think the reason you don't walk super fast the other competitor is that you destroy your own products.
Imagine if Haswell had been 2x Sandybridge performance.
What would the MAJORITY Of people do? Buy a processor 1/2 the price of their original purchase point.

Problem is, if you make TOO far of a performance jump at the current pricing scheme, you end up overshadowing all your OWN products.
If AMD's R series video cards had shown a 2x performance jump over the 7 series I guarantee you the MAINSTREAM customer isn't going to go "Wow, I can get a $400 dollar vid card and it will destroy nvidia!" They would go "Oh, I can now spend $200 and get $400 performance." Also, if the R290x was twice as fast as the HD 7970 what are you going to use it to play really? 4K gaming? Even if it was fast enoguh to 60FPS 4K gaming right now I still as a MAINSTREAM customer wouldn't get it because I can't afford a 4K panel at the moment.

Software isn't moving fast enough at all really for there to be major differences between the companies in performance without destroying the whole idea of a "product line up"

Edit: That said, if there was MAINSTREAM software came out tomorrow that needed 4x the performance we ahve today, you can guarantee AMD and NVidia and Intel would start nuclear wars to be able to get the processing speed necessary to run it and offer a "typical product lineup of slowest to fastest" to do it.

Probably mvoe heaven and earth and see the necessary processing equipment within 4-5 months. Engineers would be given a diet of coffee and adderall.

 

[bystander36]

I think the reason you don't walk super fast the other competitor is that you destroy your own products.
Imagine if Haswell had been 2x Sandybridge performance.
What would the MAJORITY Of people do? Buy a processor 1/2 the price of their original purchase point.

Problem is, if you make TOO far of a performance jump at the current pricing scheme, you end up overshadowing all your OWN products.
If AMD's R series video cards had shown a 2x performance jump over the 7 series I guarantee you the MAINSTREAM customer isn't going to go "Wow, I can get a $400 dollar vid card and it will destroy nvidia!" They would go "Oh, I can now spend $200 and get $400 performance." Also, if the R290x was twice as fast as the HD 7970 what are you going to use it to play really? 4K gaming? Even if it was fast enoguh to 60FPS 4K gaming right now I still as a MAINSTREAM customer wouldn't get it because I can't afford a 4K panel at the moment.

You do realize that 20 years ago, every 2 years, CPU's did double in performance. The reason they don't today is simply a matter not having as many avenues open to progress. If they could, they would. It worked back then, it would work today.

 

[2is]

You do realize that 20 years ago, every 2 years, CPU's did double in performance. The reason they don't today is simply a matter not having as many avenues open to progress. If they could, they would. It worked back then, it would work today.

There were also appreciable gains in many tasks/apps back then, not so much today.

 

[SammichPG]

I thought the exact same thing about cloudfire.

As for Carkmack, he is VERY pro OpenGL (All ID games are OpenGL). And for whatever reason, he has been very against Mantle. I am not sure if this is because he is working with nVidia, or because Mantle isn't OpenGL.

Carmak is a zombie, his software house hasn't produced anything ground breaking since quake 3.
It's not the 90s anymore.

 

[frozentundra123456]

This seems to be Nvidia's way of responding to Mantle, by getting Carmack and others to downplay it as much as possible.

Same as dice and amd touting mantle. Just natural to promote your own product and downplay your competetor's. It's all just talk until somebody backs it up with real performance data.

 

[UaVaj]

Yes. I've often stewed about something. Why, after all the years AMD and Nvidia, or AMD and Intel have been designing silicon, has not one of the companies ran away with performance? You'd think that over a period of a few decades, one company would vastly overshadow another, not ALWAYS be ridiculously neck and neck especially when AMD for example has next to no funds. They can STILL offer a Titan competitor, when Nvidia has relatively unlimited funds to utterly wipe the floor with AMD. Same with Intel. AMD does not offer a faster CPU than Intels fastest, BUT is STILL able to keep up a few tiers lower with next to no money for R&D and such. When I say next to no money, I'm talking in relative terms compared to Intel or Nvidia.

Seems quite impossible to me and the only explanation that I can wrap my mind around is that they actually are all working together, despite public appearances.
Something isn't right. Hasn't been for many many many years. Only in the last few have I suspected anything like collusion between these companies.

I'm sick (Not really. Just sick of it) over the whole industry. My views have changed pretty radically over the last few years. Anyone ever get this feeling when they see year after year, gen after gen, that one company does not continually walk away from the other? Does not make sense. Even two turtles racing over a 100 mile distance will cross a given finish line at completely different times. Could be a week apart, could be a month, or a year. Best analogy I can come up with on a half a cup of coffee.

Good morning to all, and have a great day!

well said. :thumbsup::thumbsup:

there definitely unproven collusion going on. that is simply part milking the system to the max.

best example. duct tape and gps. it was invented and available years ago. only made available not too long ago to the public.

willing to bet Kepler and Hawaii were available ages ago too. willing to bet Maxwell is just collecting dust waiting to launch only after until Kepler is totally milked dry. willing to bet NVidia is currently working on Volta as we speak.

 

[bgt]

turbos do have some good characteristics.

The failure rate of modern downsized petrol engines with turbo's is much higher than the engines without turbo'sD: look at the VAG group, PSA, BMW!

 

[HurleyBird]

Yes. I've often stewed about something. Why, after all the years AMD and Nvidia, or AMD and Intel have been designing silicon, has not one of the companies ran away with performance? You'd think that over a period of a few decades, one company would vastly overshadow another, not ALWAYS be ridiculously neck and neck especially when AMD for example has next to no funds. They can STILL offer a Titan competitor, when Nvidia has relatively unlimited funds to utterly wipe the floor with AMD. Same with Intel. AMD does not offer a faster CPU than Intels fastest, BUT is STILL able to keep up a few tiers lower with next to no money for R&D and such. When I say next to no money, I'm talking in relative terms compared to Intel or Nvidia.

As far as Intel goes, they're currently underachieving due to lack of competition. Intel could easily come out with mainstream six and eight core chips, but why would they when they can continue to sell quad core chips at the same price? Keeping AMD in the running also avoids nasty antitrust actions. You can argue that antitrust actions are less likely today with the popularity of ARM, but Intel's ultimate goal is to kill ARM entirely and get everything back on x86. If they ever realize that goal, they'll still need token competition from AMD.

How do they keep up with Nvidia? Better engineers. They beat Nvidia to every process node and architecturally seem to be ahead. That lead is shrinking fast though -- RV770 embarrassed GT200 in terms of perf/mm2, but GCN and Kepler are pretty close even if GCN seems to be a bit more forward looking. Where Nvidia has AMD beat is software and drivers, though that gap seems to be narrowing as well.

 

[ShintaiDK]

As far as Intel goes, they're currently underachieving due to lack of competition. Intel could easily come out with mainstream six and eight core chips, but why would they when they can continue to sell quad core chips at the same price? Keeping AMD in the running also avoids nasty antitrust actions. You can argue that antitrust actions are less likely today with the popularity of ARM, but Intel's ultimate goal is to kill ARM entirely and get everything back on x86. If they ever realize that goal, they'll still need token competition from AMD.

Thats a pile of utter nonsense. The demand for anything above 4 cores on the desktop/mobile segment is close to zero. People rather want faster IGPs and more efficient CPUs. AMDs future Steamroller chips is also 2M/4T. 6 and 8 cores dead there too because its not what the masses want.

 

[SlickR12345]

I want the GPU wars to start already so I can buy a GTX 760 for 150 euros.

 

[HurleyBird]

Thats a pile of utter nonsense. The demand for anything above 4 cores on the desktop/mobile segment is close to zero.

Don't confuse need with demand. You don't need more than two cores on a smart phone, but there is plenty of demand for quad core chips nonetheless. If Intel put out six and eight core chips at mainstream prices there would certainly be demand for them.

 

[ShintaiDK]

Don't confuse need with demand. You don't need more than two cores on a smart phone, but there is plenty of demand for quad core chips nonetheless. If Intel put out six and eight core chips at mainstream prices there would certainly be demand for them.

Dont confuse your demands with the demands of the masses.

The demand is so low, that AMD will only sell 2M/4T streamrollers.

 

[UaVaj]

Don't confuse need with demand. You don't need more than two cores on a smart phone, but there is plenty of demand for quad core chips nonetheless. If Intel put out six and eight core chips at mainstream prices there would certainly be demand for them.

qual core, hex core, octo core, duodec (12) core. with the exception of a few. all the mass need is a nice dual core, however bragging rights can draw up demand.

 

[DiogoDX]

kkkkkkkkkkkkkkkkkkkkkkkk

Now Camark don't like proprietary stuff. Says the guy that use CUDA to stream textures in Rage.

LOL too for the Epic guy that is using physX in Unreal.

This is too funny to be truth.

 

[bystander36]

There were also appreciable gains in many tasks/apps back then, not so much today.

A big part of the big gains in tasks/apps is the dev's knowing what was available, so they designed around the new power of PC's.

Currently, the CPU is a huge bottleneck in gaming. At least for those wanting 100+ FPS for ultra smooth gaming. Even 60 FPS all the time, in many games, does not happen because the CPU. Of course that probably wouldn't change, because as soon as the dev's have more CPU power to work with, they'll put it to use. They may not like multi-threading, due to the extreme work involved in making single processes work as multiple threads, but that isn't to say they have the power they want either.

 

[AnandThenMan]

well said. :thumbsup::thumbsup:

there definitely unproven collusion going on. that is simply part milking the system to the max.

best example. duct tape and gps. it was invented and available years ago. only made available not too long ago to the public.

willing to bet Kepler and Hawaii were available ages ago too. willing to bet Maxwell is just collecting dust waiting to launch only after until Kepler is totally milked dry. willing to bet NVidia is currently working on Volta as we speak.

I really hope this is intended to be sarcasm.

 

[UaVaj]

I really hope this is intended to be sarcasm.

if you were nVidia/AMD/Intel. would you milk the customer for every cents you can legally get.
or
would you be so nice and only ask for absolute minimum profit.

decide as you see fit.

 

[2is]

Don't confuse need with demand. You don't need more than two cores on a smart phone, but there is plenty of demand for quad core chips nonetheless. If Intel put out six and eight core chips at mainstream prices there would certainly be demand for them.

Don't confuses enthusiasts demand (which is minuscule) with the demand of the general population.

 

[lehtv]

if you were nVidia/AMD/Intel. would you milk the customer for every cents you can legally get.
or
would you be so nice and only ask for absolute minimum profit.

decide as you see fit.

Even if you were right about that, it would still be a huge leap from that to "Maxwell is just collecting dust waiting to launch only after until Kepler is totally milked dry"

 

[HisDivineOrder]

Even if you were right about that, it would still be a huge leap from that to "Maxwell is just collecting dust waiting to launch only after until Kepler is totally milked dry"

I think what he's saying is that Big Kepler could afford to be delayed for the consumer market a year because AMD didn't have anything that required it be released in 2012. That is, nVidia could let the consumer market have a mid-range card overclocked to match AMD's fastest for most of 2012 and let Big Kepler focus on just the Compute market.

Then when 2013 came and AMD delayed the GCN update that should have been, nVidia who had prepared all of 2012 to release Big Kepler saw they were again going to have nothing to compete against.

They looked at the fact that back in December of 2011 when the 7970 was released they saw that AMD had released a card with 10-20% improvement over the ancient-by-then 580 (due mostly to drivers at the time) for $550+. So they figured Big Kepler, a year later worth of anticipation, would be worth at least $100 more, right?

But hey. AMD didn't have anything. So, nVidia could do whatever they wanted because they were so far ahead of the game and Big Kepler had been popping out of fabs for a good long while by then. Lots of defective dies were coming out and many of them could mostly hit the high end spec. So why not throw 6 GB on it and ramp up demand by making it a card unto itself since there wasn't anything Radeon that could even remotely come near it with the Big Tahiti's delay to end of 2013?

Bam, Titan. Cleverly using a halo product to ramp up demand for Big Kepler beyond the already ludicrous hype, they wait a few months and then release a Kepler at that $650 price point. They clever add to the pricing AMD had already raised themselves a year earlier and the beauty of it is because of Titan, there are people who scream like Twilight girls, "Omgz, it's almost half the price of Titan, it's a DEAL!" On a $650+ GPU.

It's not that nVidia is sitting around on Maxwell. But we know it was sitting on Big Kepler. If AMD had anything that came close to Big Kepler (GK110), they'd have had GK110 out there just like they have 780Ti coming out the very second AMD shows up with something comparable.

Hell, if AMD had showed up earlier this year when they were supposed to with Tahiti XL, it's hard to imagine nVidia even showing up with Titan the way they did at all.

Competition is healthy and AMD hasn't been competing on the GPU or CPU level except serving reheated leftovers. That's really, really bad because there's no reason for Intel and nVidia to do anything but tread water.

That's what he's saying. Fact shows it's happening, too. Look at Haswell, a mediocre improvement over IVB, which was itself a mediocre improvement over SB. If AMD had a majority of CPU's that could even beat the two year old SB right now, do you think Intel would be so lackadaisical about performance gains? Intel can afford to because AMD is not showing up with anything remotely compelling to anyone outside the fanboys and the cheapest of the cheap with an obsessive need for multicore beyond reasonable sense.

Intel and nVidia are just going through the motions because AMD isn't giving them much reason to look up from their mobility plans to do more than pop out a refresh or a rebrand.

 

[Erenhardt]

^ I don't get it. I read your post twice, and I can't understand. What was nvidia waiting for the whole time? Wouldn't they want to have the fastest card on the market?

 

[HurleyBird]

The demand is so low, that AMD will only sell 2M/4T streamrollers.

The die size of a 4M/8C APU would be unfeasible, and nobody is going to buy a non-APU 4M/8C CPU from AMD while they are so far behind in single threaded performance and power consumption.

Don't confuses enthusiasts demand (which is minuscule) with the demand of the general population.

The Galaxy S4 sold pretty damn well to the general population, despite the fact that they had to shell out extra money for the "unnecessary" quad core.

If all that the mass market cared about was "good enough" performance, AMD and Intel would have close to equal market share by now.

 

[sontin]

^ I don't get it. I read your post twice, and I can't understand. What was nvidia waiting for the whole time? Wouldn't they want to have the fastest card on the market?

Woot? They have the GTX690 since may 2012.
There was no reason for GK110 based cards at the end of 2012 because they sold enough GK104 cards for >$250.

They have Titan since 8 months. They have the GTX780 for nearly 5 months.