Kepler Unbeatable: NVIDIA

[Recus]

Who is that idiot that translated the message?
Aspettando Kepler... pazienza, pazienza, pazienza che il momento giusto arriverà, e allora... non ce ne sarà più per nessuno!
===
Waiting for Kepler ... patience, patience, patience, the right time will come, and then... there will be none!

Meaning they all will be sold! Is like publicity, nothing to do with kepler itself.
Open google translate see for yourself!

http://i41.tinypic.com/2yjzyti.png

I never had problem with ATI and now AMD drivers. I know nobody that had problems. The only time I remember was back then with the good old 9700 Pro, 9800 Pro and 9800 XT where the 1st driver release were a little buggy, but got resolved quick.

Blast from the Past
http://1.bp.blogspot.com/_9vgJ1nwu_xA/S774cTvfPTI/AAAAAAAACv0/aS3Sc2AGETE/s1600/ati_vs_nvidia.jpg
http://fc01.deviantart.net/fs70/i/2010/139/e/f/ATI_vs_Nvidia_by_masquevale.jpg
http://www.notebook4game.com/web/wp-content/uploads/2010/02/ATI-VS-NVIDIA.jpg
http://cdn.overclock.net/b/b4/b4340e19_vbattach172842.jpeg


http://techbuket.net/images/M5YDu.jpg

 

[Horrux]

Who cares about power usage,
People who buy high end gpu's don't care about power usage, most hardware junkies get water blocks or aftermarket coolers

I know i can't wait and at the end of this year a will grap one, i don't care about price, my gtx275 is getting old as shit

Actually, if you had to change your PSU in order to accommodate your new videocard, I bet you would care. A lot.

I pay $0.15/kWh. That means it cost me near $36/month for a 300W GPU, if it ran 24/7. My power bill for January was $437.30. You bet performance/watt matters, because over 12 months, that's $432 to buy Mcdonalds with. I'll take that cheque, please, as it's easy enough for me personally to make it worthwhile. For the average user, it still might buy that McDonalds. I can't be bothered to guess at how long a GPU is at full load, on average...depends on the app and such, but it's be interesting to get a real number.

You are forgetting that for 7-8 months of the year at least, you are heating your house, and the power used by your GPU is power you're not using to heat your house, so remove at least half of what you calculated. Moreover, you likely NEED it to run a lot less than 24/7... Say 4 hours a day to be reasonable, remove 5/6th of that... A more realistic figure for that $432 thus becomes $36 a year. Well, you can buy one game. With the power your whole GPU uses. But say the difference in power consumption between competing GPUs is a whopping 20%.... Then your $36 a year becomes $7.20. Barely enough for a taste of the shit they serve at McDonalds.

You simply forget that 6990 did beat 590. So, Fermi was beaten in the end. Somehow, I suspect the same thing is going to happen again this time...

It's almost guaranteed it will, given that AMD can pack higher performance in the power budget of a single card, due to their higher performance per watt. This is another reason why power consumption DOES matter. A LOT.

If some rumours and slides are to believed Keplar will have 100% increase in performance over the 5** series for the respectable replacements, I seriously hope this is true as I would buy double the performance of an 570 for the same price, nevermind TWIMTBP, TWTUTBM = the way they used to be made

Not gonna happen. The engineers are really making the best they can out of the current tech. Given that, there is no way you can expect anything close to 100% increase in performance. In the early days of a technology, engineers still learn new tricks and can squeeze huge performance gains from their chips. Take for example the 486DX2 VS 486. Pretty much double the performance. But the difference between a 2600k and a 2700k? A gain of 2.9%. Yay. But that's what a new CPU model gives you today. Same with GPUs. 25% to 33% between generations is bound to become more the norm than the exception, and soon progress will get closer to 25%, then 20%... And so on.

Really?

In the summary page (7970 xfire review), the 590 beats the 6990 at every resolution. Here's the 2560 res summary.

http://tpucdn.com/reviews/ASUS/HD_7970_CrossFire/images/perfrel_2560.gif

I'm only putting this in to stop blatant mistruths. Lots of people give the 590 a hard time but it runs cooler and quieter by most accounts and the very own TPU round up for the link above puts 590 as better for every resolution. But as always, it's really game dependant.

I'm pissed NV is holding back info on Kepler as I'm looking to upgrade but it's so close I need to wait to see how Kepler performs as I'm keen to see a 7970 price drop. Unless Kepler is way better (doubt it).

If Kepler bombs, I'm buying 2 7970's just as a capitalist reaction!!

Again using "All resolutions"... How about comparing them at 1080p or 1920x1200 or 2560xwhatever? You think people are getting $750 video cards to play at 1024x768?

I don't think anyone can really refute that Nvidia will offer more powerful cards, they have pretty reliably, but the real question is both cost and relative performance. If you could pay $1000 for a 680 that has 30-40% higher performance than an HD7970, you're not exactly compelled to buy it. If Nvidia can match AMD at price points and offer much higher performance they will crush the competition.

I have high hopes for Kepler and plan on getting a 6xx series GPU to give Nvidia a shot since the last time I had an Nvidia card was a 7900GS a few years back. Nvidia is just trying to keep people in anticipation of their new line, but given the problems with the 7xxx series and drivers I've seen around, I don't think they really have much to worry about.

Of course nV will release cards at prices that put AMD's products to shame. That's because AMD is milking its business the way they should. Once Kepler is actually available, AMD will adjust its pricing to compete. Yes, that should mean price reductions...

 

[Benetanegia]

Again using "All resolutions"... How about comparing them at 1080p or 1920x1200 or 2560xwhatever? You think people are getting $750 video cards to play at 1024x768?

Lol, blind much? He posted 2560x1600, where it's 7% faster, just saying.

 

[Horrux]

Lol, blind much? He posted 2560x1600, where it's 7% faster, just saying.

I stand corrected. Damn, and I just got told by my optometrist that my glasses are perfect, yesterday. Can't trust anyone these days.

 

[Super XP]

Kepler Unbeatable: NVIDIA?
Are they kidding

OK I like to buy one now please. Oh, but wait they are not available.

 

[Horrux]

Kepler Unbeatable: NVIDIA?
Are they kidding

OK I like to buy one now please. Oh, but wait they are not available.

Unbeatable at being impossible to find?

 

[Benetanegia]

Kepler Unbeatable: NVIDIA?
Are they kidding

OK I like to buy one now please. Oh, but wait they are not available.

"Waiting for Kepler ... patience, patience, patience, the right time will come, and then ... it will be unbeatable"

It's not exactly like he is saying that Kepler is out now, considering that in 1 sentence he used 4 different expressions that point out that it's not here yet.

 

[Aquinus]

"Waiting for Kepler ... patience, patience, patience, the right time will come, and then ... it will be unbeatable"

It's not exactly like he is saying that Kepler is out now, considering that in 1 sentence he used 4 different expressions that point out that it's not here yet.

Last I checked the 7970 was still on the market and Kepler is waiting to be released. What is that again? AMD actually has a working product? Kepler might end up being faster but it's taking them enough time to release it and in that time a lot more people are willing to buy a 7970 instead of waiting for Kepler.

Plus, hype is just hype. It's not making the GTX 6xx series come any faster. When push comes to shove, at least the 7970 has verified numbers on how it performs. nVidia has released no numbers and has just said it is fast. If anyone else said something like that, the community would laugh, and honestly, nVidia is no different.

I laugh at Kepler until it can prove to me that it will deliver.

 

[Benetanegia]

Last I checked the 7970 was still on the market and Kepler is waiting to be released. What is that again? AMD actually has a working product? Kepler might end up being faster but it's taking them enough time to release it and in that time a lot more people are willing to buy a 7970 instead of waiting for Kepler.

Plus, hype is just hype. It's not making the GTX 6xx series come any faster. When push comes to shove, at least the 7970 has verified numbers on how it performs. nVidia has released no numbers and has just said it is fast. If anyone else said something like that, the community would laugh, and honestly, nVidia is no different.

I laugh at Kepler until it can prove to me that it will deliver.

All I read is bla bla bla, already said a million times bla bla. We know that Kepler has not been released yet, I don't think we need just one more guy saying it and it's not what is being discused, it's not news, but it's performance. Kepler might come later, it might even be late (delayed) but we don't really know, since no schedule has ever been released (besides 2012). But the fact of the matter is that it will posibly be much faster, or that's what the Nvidia guy said. He also said we will have to wait, which is what my post was pointing out. "Nvidia" is not saying that they have a faster product, it's saying that they will have a faster product. If true no amount of "but it's been released yet? AMD has. bla bla" is going to change that fact.

As to sales, if you want to be any realistic, AMD is not selling many HD7000 cards right now, because the ramp up has just begun. In Q2 AMD and Nvidia will sell much much more cards than in Q1, and in Q3 a lot more than that and in Q4 a hell of a lot more, so much more, in fact, that the number of HD7000 cards sold in these couple of months will be negligible. And that's the reality, that's why Nvidia is not in a rush to release Kepler. This is not HD5000 with new DX11, 30% faster than previous gen, et al. HD7000 and Kepler bring very little "new", and HD7970 is 10-15% faster than previous gen, so there's no rush. Even HD5000 didn't sell so much more than GTX400 when factoring the fact that it was more than double the time for sale. And HD6000 vs GTX500, GTX500 won in sales, by a good margin too, considering its higher ASP. So they are not in a rush at all.

 

[faramir]

Unbeatable power hungry?

My thoughts exactly.

 

[Horrux]

If it's unbeatable, it means it's the last architecture nVidia will EVER release.

 

[N3M3515]

Of course it will be unbeatable. It isn't hard to get out a product that is better than your competition when you wait for your competition to release theirs so you have a line in the sand set for you to cross...

My exact thoughts.

 

[HD64G]

Really?

In the summary page (7970 xfire review), the 590 beats the 6990 at every resolution. Here's the 2560 res summary.

http://tpucdn.com/reviews/ASUS/HD_7970_CrossFire/images/perfrel_2560.gif

I'm only putting this in to stop blatant mistruths. Lots of people give the 590 a hard time but it runs cooler and quieter by most accounts and the very own TPU round up for the link above puts 590 as better for every resolution. But as always, it's really game dependant.

I'm pissed NV is holding back info on Kepler as I'm looking to upgrade but it's so close I need to wait to see how Kepler performs as I'm keen to see a 7970 price drop. Unless Kepler is way better (doubt it).



If Kepler bombs, I'm buying 2 7970's just as a capitalist reaction!!

http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_590/23.html

This is the first review of 590 and it loses slightly from 6990 on top resolutions. So, it's a matter of driver maturing. Besides that, in the graph you posted difference in tiny (3-4%) and efficiency is much better for AMD...

 

[N3M3515]

All I read is bla bla bla, already said a million times bla bla. We know that Kepler has not been released yet, I don't think we need just one more guy saying it and it's not what is being discused, it's not news, but it's performance. Kepler might come later, it might even be late (delayed) but we don't really know, since no schedule has ever been released (besides 2012). But the fact of the matter is that it will posibly be much faster, or that's what the Nvidia guy said. He also said we will have to wait, which is what my post was pointing out. "Nvidia" is not saying that they have a faster product, it's saying that they will have a faster product. If true no amount of "but it's been released yet? AMD has. bla bla" is going to change that fact.

As to sales, if you want to be any realistic, AMD is not selling many HD7000 cards right now, because the ramp up has just begun. In Q2 AMD and Nvidia will sell much much more cards than in Q1, and in Q3 a lot more than that and in Q4 a hell of a lot more, so much more, in fact, that the number of HD7000 cards sold in these couple of months will be negligible. And that's the reality, that's why Nvidia is not in a rush to release Kepler. This is not HD5000 with new DX11, 30% faster than previous gen, et al. HD7000 and Kepler bring very little "new", and HD7970 is 10-15% faster than previous gen, so there's no rush. Even HD5000 didn't sell so much more than GTX400 when factoring the fact that it was more than double the time for sale. And HD6000 vs GTX500, GTX500 won in sales, by a good margin too, considering its higher ASP. So they are not in a rush at all.

Well.......arriving almost half a year later it must by all means beat amd by a decent margin don't you think? It's so obvious.
What would be nice is if nvidia prices gk110 the same as hd7970.

 

[GSquadron]

That guy is wrong too and looks like a lot of people don't know italian....
How can these people be so foolish to accept that translation???
and he is surely an italian, but "EXTREME ENTHUSIAST" of Nvidia

EDIT: lol, now i am seeing even in english is translated: there will be no cut for everyone

 

[Benetanegia]

Well.......arriving almost half a year later it must by all means beat amd by a decent margin don't you think? It's so obvious.
What would be nice is if nvidia prices gk114 the same as hd7970.

I don't know in your universe, but in mine 3 months is not half a year.

And no just because you come later, that does not mean that it has to or that you can be faster*. GPUs are designed and manufactured in a process that lasts 3-5 years, and this process ends whenever it ends. And you can't change much either later on on the cycle, in the last 3 months you can change nothing at all, except clocks and fully knowing that changing clocks will affect yields. So GK104 is what Nvidia expected to be at least 2+ years ago, accomodated to the real/final state of 28nm process and maybe slightly adjusting clocks to fine tune where in the performance scale they want to end up. And that's it. According to them such cycle ended up making GK104 "unbeatable" by AMD. I'm sure that means that AMD does not have anything coming soon that will be able to beat it, including a higher clocked HD7980 or something like that.

* GeForce FX? HD2900? Bulldozer? etc etc Also every AMD card has launched later than Nvidia's in the past years (usually 1-2 months later) except HD5000 and they were never faster. When a card is launched, has nothing to do with how it performs, the physical limits that the manufacturing process imposes is pretty much the only thing that matters (and because of this which die size you choose to go with) and 28 nm will be used now and until 20 nm launches in 2 years or so. neither AMD or Nvidia will be able to make a much better chip than they already did. They can make a bigger one and a slightly better one, but no magic will be made. For example, the best that AMD did 12+ months after Cypress was Cayman, a chip that despite using the VLIW4 advantage over VLIW5 and being ~15% bigger than Cypress, it's only 15% faster. So they achieved what the process allowed them to do.

 

[radrok]

I can confirm that the translation is a bit forced on the unbeatable part, and yes I am italian born.

 

[Horrux]

I don't know in your universe, but in mine 3 months is not half a year.

And no just because you come later, that does not mean that it has to or that you can be faster*. GPUs are designed and manufactured in a process that lasts 3-5 years, and this process ends whenever it ends. And you can't change much either later on on the cycle, in the last 3 months you can change nothing at all, except clocks and fully knowing that changing clocks will affect yields. So GK104 is what Nvidia expected to be at least 2+ years ago, accomodated to the real/final state of 28nm process and maybe slightly adjusting clocks to fine tune where in the performance scale they want to end up. And that's it. According to them such cycle ended up making GK104 "unbeatable" by AMD. I'm sure that means that AMD does not have anything coming soon that will be able to beat it, including a higher clocked HD7980 or something like that.

* GeForce FX? HD2900? Bulldozer? etc etc Also every AMD card has launched later than Nvidia's in the past years (usually 1-2 months later) except HD5000 and they were never faster. When a card is launched, has nothing to do with how it performs, the physical limits that the manufacturing process imposes is pretty much the only thing that matters (and because of this which die size you choose to go with) and 28 nm will be used now and until 20 nm launches in 2 years or so. neither AMD or Nvidia will be able to make a much better chip than they already did. They can make a bigger one and a slightly better one, but no magic will be made. For example, the best that AMD did 12+ months after Cypress was Cayman, a chip that despite using the VLIW4 advantage over VLIW5 and being ~15% bigger than Cypress, it's only 15% faster. So they achieved what the process allowed them to do.

The 7970 was launched 3 months ago...

 

[erocker]

The 7970 was launched 3 months ago...

January 9th was three months ago?

 

[cadaveca]

January 9th was three months ago?

technically, it launched in December. not quite 3 months, but close.

 

[Aquinus]

It will at least be half a year until Kepler actually gets released? That's a lot of extra sales for AMD.

 

[Benetanegia]

Even taking late December (21st I think?) as launch, which we know it is not really true.

21st january == 1 month
21st feb == 2 months
23rd march == 3 months

Far really far from half a year. And if it's hard launch in the end, it's more like 2 months.

 

[claylomax]

January 9th was three months ago?

Yep; time flies, doesn't it?

 

[Benetanegia]

Yep; time flies, doesn't it?

No it's not, January 9 was not 3 months ago, it's not even been 2 full months (~60 days) since then.

So while the "time flies" expression is usually true, it's pretty obvious that it does not apply in this case. Apparently time passes reaaaaaaaaaaally sloooooow for some people when they are waiting for a release.

 

[claylomax]

No it's not, January 9 was not 3 months ago, it's not even been 2 full months (~60 days) since then.

So while the "time flies" expression is usually true, it's pretty obvious that it does not apply in this case. Apparently time passes reaaaaaaaaaaally sloooooow for some people when they are waiting for a release.

Joking? Irony? Sarcasm?
Where the heck are you from?