kit guru 8970/50 in JUNE ???

[Lonbjerg]

It's painfully obvious that they're going to use the K-10 to enjoy single precision in a large super computer that demands double precision.

Sometimes you have to wonder if people posting actually evenknow what they are talking about...this time it is more obvious though...good catch

And what is up with the "1 GHZ" red herring?

~780 Mhz should be Big K's sweetspot

 

[f1sherman]

GK110 produces so much anxiety that the only way to deal with it - is denying it's existence

It would be great if you looked at your sources

http://blogs.knoxnews.com/munger/2012/09/the-big-computing-change-is-ta.html

because the original source only names the chip as Kepler, not K20.

no it wouldn't


everyone interested enough already knows what goes into Titan and future Crays

~780 Mhz should be Big K's sweetspot

You mean this?

Sounds about right

 

[Granseth]

It's painfully obvious that they're going to use the K-10 to enjoy single precision in a large super computer that demands double precision.

I believe that they won't be using a K-10, the same as you. But I'm guessing, and everyone here should be better at checking their facts or point out that they are guessing (even if it's a qualified guess).

 

[Granseth]

no it wouldn't


everyone interested enough already knows what goes into Titan and future Crays

How can you say that when no one can point to a source that says what GPU that goes into Titan?

Even though it seems likely that the K20 will be the GPU part, there might be a reason not to use it. I just want to question the truth until we actually knows it to be true.

 

[SirPauly]

That's not a guess though because the K-10 is specifically being catered to oil and gas reserves and signal and image processing, which use single precision -- the key is this is going to be a scientific super-computer.

 

[biostud]

To some extent yes, but there are other factors involved(R&D is simply the largest the way things are done currently). Say nVidia orders a wafer of K20 parts, 75% of them have one bad cluster. If it is a compute only solution, they have yields of 25%. By sharing the same die, they have a healthy batch of 770s to sell for ~$400 and another 25% of the wafer they can sell for ~$5K.

They could us the defect dies in $3-4K K20 tesla cards.

 

[biostud]

GK110 is only a bigger GK104 with a few more tweaks and DP-Compute-Units.
It makes no sense for them to design another Geforce only part when they have GK110.

according to anandtech the GK110 consist of ~7.1B transistors while the GK104 ~3.5B.

 

[f1sherman]

How can you say that when no one can point to a source that says what GPU that goes into Titan?

Even though it seems likely that the K20 will be the GPU part, there might be a reason not to use it. I just want to question the truth until we actually knows it to be true.

I am not going to Google that for you. If you are interested in subject, you would have already knew what goes into Titan.

"checking facts" does not equal being fastidious about known facts

throwing 0.19 DP TFlops K10 around as possible ORNL candidate is a nonsense and subject dragging

 

[NIGELG]

Another thread 'Wrecked' by talking about Nvidia.

Keep on topic please.Note the thread title.It says kit guru 8970/50 in JUNE ???

Did you see 'Big Kepler' anywhere in that title?D:.

I am looking forward to seeing improvements from the 8900 series over the 7900 series in terms of heat/noise and power consumption as well as a decent increase in speed.

The world needs more efficient chips.

 

[Granseth]

I am not going to Google that for you. If you are interested in subject, you would have already knew what goes into Titan.

"checking facts" does not equal being fastidious about known facts

throwing 0.19 DP TFlops K10 around as possible ORNL candidate is a nonsense and subject dragging

How can I google my claim that there are no facts, just assumptions. I've actually done what I can as I have disproved the facts other have posted. (and I can only repeat myself by saying that I have no problem with people believing that Titan will include K20, It's just that as far as I have seen posted no one can claim that as a fact)

But my main points is that one says something is the truth it would be nice if they have checked their sources.

 

[jackstar7]

Gee whiz, this AMD thread sure has a lot of nV-ious people...

 

[3DVagabond]

Gee whiz, this AMD thread sure has a lot of nV-ious people...

Banter. It makes these forums go 'round. Occasionally though, something good does come out of it. :thumbsup:

 

[boxleitnerb]

Maybe AMD releases Oland first and later Venus? That would certainly explain the late rumored full launch in June.

 

[scooterlibby]

Higher resolutions will become a differentiator next year, trust me. I think the added GPU horsepower will not go to waste! Intel is pushing for 4k, and nvidia/AMD will deliver products designed for that. Apple (even though I hate them) has GPU performance as #1 on their wishlist - so they can differentiate their products further. In fact, this is one of the reasons why intel is so fervent in their pursuit of improving iGPU performance. Mobile has become much more relevant in recent years, so power consumption / graphics performance is far more important to intel now than ever before -- consider that the IB is not much faster IPC wise than the SB. Their focus was iGPU performance because Apple practically begged them for it (if reports are accurate), and I think that will remain the case with haswell.

Things will change soon IMHO. Also consider that new consoles with DX11 features built in will be released next year as well. GK110, 8970, 4k screens, I think 2013 has the makings of a glorious year in terms of gaming.

I really hope you are right!

 

[RussianSensation]

Semiaccurate reports that Sea Islands is delayed to March 2013 and the performance increase will only be 15%, down from the previous estimate of 30%.
http://semiaccurate.com/2012/10/11/what-is-the-latest-on-amds-sea-islands/

If true, that would be really poor showing from HD8000 series. They had more than 1 year and 15% is all they are going to come up with? GK110-based GTX780 will destroy such a chip if these rumors are true. Perhaps the financial woes of AMD are finally affecting the graphics division. Maybe they had no $ to make a major redesign for Tahiti in 2013? Hmm... disappointing if true.

It'll be interesting to see if NV decides to launch the 7 billion GK110 monster then if all AMD will bring is a 15% faster HD8970. Maybe AMD should have taken out compute/double precision out for next generation and just built a large Pitcairn XT Version 2.0 and waited until 22nm to reintroduced back compute because clearly they are running into power consumption limits on 28nm. Not sure what to say other than 15% is only good enough for a refresh product. This is starting to smell like another HD5870 --> HD6970 update.

 

[AnandThenMan]

15% is not enough to even warrant a new series designation. If true, so long AMD nice knowing you.

 

[RussianSensation]

True. They'll be forced to compete on price again if this pans out but not sure how they are going to do that with a ~400mm^2 die. I am hoping NV brings something more than 15-20% over 680 or it'll be another GTX500/6900 refresh generation, which was disappointing.

 

[OILFIELDTRASH]

15% is not enough to even warrant a new series designation. If true, so long AMD nice knowing you.

Take off a gen and wipe your tears. I didn't think this gen was worthy of an upgrade so I politely declined. By the time next March rolls around I'll be in line for a solid upgrade.

 

[2is]

Whatever it is AMD/nVidia is going to put out there, I just hope it's around the time Crysis 3 is released. That way I can read all the reviews and decide on a 2nd 680 or one of the new beasts. If 15% is accurate, i'll probably go with a 2nd 680, by then it will a lot cheaper (especially on the used market) and more powerful than a single next gen.

 

[AnandThenMan]

Take off a gen and wipe your tears. I didn't think this gen was worthy of an upgrade so I politely declined. By the time next March rolls around I'll be in line for a solid upgrade.

I have nothing to do with it. AMD is in real trouble, but their graphics division has been a bright spot for them. Even given that desktop graphics are a small part of their revenue, if they falter here the image hit will be substantial. AMD can't afford anymore hits, to image or their bottom line.

 

[Smoblikat]

Pathetic, I remember the HD 4000 - HD5000 and GTX 2XX to GTX 4XX was an amazing transition. Now the companies dont care enough about us to give us any new performance.

 

[RussianSensation]

Take off a gen and wipe your tears. I didn't think this gen was worthy of an upgrade so I politely declined. By the time next March rolls around I'll be in line for a solid upgrade.

So you skipped a generation where HD7970 @ 1150mhz / GTX680 @ 1280mhz were ~50% faster than a GTX580 but 15% more on top of that is somehow worth waiting for 15 months to get (from January 2012 to March 2013)? The other thing is the price. By the time March 2013 rolls around, HD7970 will be $250-275 I bet. Who is going to pay $499 for a 15% faster HD8970?

This Greenpeace focus on performance/watt is killing our generational performance increases. If only NV launched a 250W GTX780!

Pathetic, I remember the HD 4000 - HD5000 and GTX 2XX to GTX 4XX was an amazing transition. Now the companies dont care enough about us to give us any new performance.

With node shrinks slowing down, so have the performance increases. However, consumers threw $ at GTX680/HD7970 despite 30-35% faster performance from 580. NV has no need to give us 50-100% faster increase since they just saw that with just 35% gamers still upgraded.

 

[2is]

Sure they do... Unless you think the number of upgraders would have remained the same even if they gave us 100% performance increase?

 

[RussianSensation]

It's not that simple. If you get more people to upgrade with 100% increase now, what are you going to do when Maxwell launches? That's a reason Intel still won't release a 6-core CPU for $300. To get 50-100% increase would require a much larger die. If HD8970 is just 15% faster, NV has no reason to deliver more than 30-40%. It's enough to make a killing without ruining their margins with a massive die, unless you think people want to buy $800-1000 single-GPUs?

We've already seen Intel execute this strategy perfectly - release very small incremental increases and shrink die sizes to maximize gross margins when there is lack of competition. If AMD graphics follows AMD's CPU path of noncompetitive products, NV will respond accordingly (either raise prices for fastest GPUs as they have done in the past when AMD flopped), or prolong generational increases by giving us enough to warrant upgrades, but not enough to make us skip the next generation. They did exactly that with GTX680. 35% faster was the worst generational increase from NV ever (GTX580 is not a real new generation from 480 so it doesn't count as a new generation imo).

Think about it, if NV release 40% faster GTX780 and 40% faster Maxwell, a lot of people will upgrade to both. If NV releases 100% faster GTX780, they are going to have less people upgrading for Maxwell. Why would they do that? 15% faster HD8970 is disastrous now. If this is true, NV has even less reason to push a 50-100% faster GTX780.

On the Intel side we have seen mind-blowing slowdown already. Intel used to pump out 2x faster CPUs every 2-2.5 years since late 90s. Since November 2008 when Core i7 920 was $284 and overclocked to 4.0ghz, we hardly moved by 35% at this price in 4 years (4.5 ghz IVB x 15% IPC of SB x 5% IPC of IVB = 5.43ghz Nehalem).

Every time AMD's graphics division flopped, we got rapped by NV. 8800GTX was $599 and 8800GTS 640mb was $449 at launch. NV happily waited 1 year to launch 8800GT. Why? Because they could.

If AMD goes bust, NV and Intel will still launch faster CPUs/GPUs but the pace of innovation will slow down dramatically (well it has already happened for Intel on the CPU side without question), but AMD GPU division wasn't that far behind NV and caught up with HD7000 series.

Without competition in the marketplace, the pace of innovation will slow down and/or prices rise long-term, and ultimately limited choice will hurt consumers. This is why there are rules against monopolies.

 

[tviceman]

Everyone's favorite Oracle Charlie is at it again: http://semiaccurate.com/2012/10/11/what-is-the-latest-on-amds-sea-islands/ TLDR - he says sea islands is coming in March and performance improvements aren't currently reaching what AMD had previously told AIB's.

Whether you believe it or not, it's Charlie D so make sure to sprinkle a heavy dose of salt on it regardless.