kit guru 8970/50 in JUNE ???

[Keysplayr]

It just changed from not a single GPU to volume shipments!

 

[SolMiester]

deleted, already posted

 

[SirPauly]

NV has a right to charge more for an equivalent or lesser product, because many years of sustained delivery and good marketting should be rewarded.

Physx is the bomb. Didn't you all hear?

A few games a year with Physx is all the masses need to pay more for NV.

Imho,

It's more than that though; it is bringing features like adaptive V-sync, frame limiters and FXAA to older GeForce generations that don't help sell new product but improves the gaming experiences to GeForce over-all -- the brand.

It's also trying to raise the gaming experience bar in popular titles by offering control panel multi-sampling and ambient Occlusion like Diablo 3, which many generations of GeForce sku's may enjoy.

With virtually every new driver release from nVidia since April, there were just not performance updates but goodie updates for gamers to enjoy on many generations of GeForce. Even Timothy Lottes would like to see TXAA be offered for Fermi hardware.

Something like allowing 3d stereo and multi-GPU content in windowed mode -- why the hell would nVidia do this? Because many gamers are MMO players and play in windowed mode.

It's much more than marketing. The premise is actually simple: IF one works hard, pro-active, try to do more, the market may reward with a slight premium for the name brand.

So, a culmination of pro-active work and the market may reward with a modest premium.

 

[RussianSensation]

Oh look, a familiar face who jumps on the bandwagon and can't connect my comment regarding how volume shipments relate to launching a consumer GK110 in Jan-Feb 2013. What a surprise. I guess you don't need volume to launch a consumer product that needs a lot more than 1000 chips on launch date.

Yes there is.

http://www.hpcwire.com/hpcwire/2012...h_of_kepler_gpus_for_titan_supercomputer.html

They are getting K20's.

Yes, in the article you linked me. In the original article which I was linked earlier, there was no mention at all it was K20 part. I don't follow Oak Ridge national laboratory daily to know they had 14000 order for K20. Thanks for linking the other article. :thumbsup:

My point is, these volumes are too low to imply that NV can launch GK110 by January or even February 2013. I seriously doubt that NV will be able to launch GK110 before 8970 unless AMD flops or NV pulls a surprise. We don't even know if GK110 will be a full 15 SMX part or a cut-down 12-13 SMX part. If it's the latter, NV would need to build up a bunch of failed GK110 chips and that would take them 3-4 months of K20 production I bet. If it's a full 15 SMX part, the issue becomes yields and fulfilling all the pre-orders for K20 parts first.

Then you have a situation where consumers are buying $400-500 GTX670/680 parts still. JHH won't be selling a 520-600mm^2 die for $500 in January 2013 without any major competition from AMD simply because each of those chips sells out for thousands of dollars in the compute space. Also, as of late NV enjoys launching later which allows them to see exactly what AMD has to offer and price their cards accordingly. Based on how long NV kept $450 prices on the 580 parts, you can be sure they'll milk GTX680 at $450-500 as long as possible. I think GTX780 will launch after 8970.

 

[SirPauly]

There is still a lot of doubt that nVidia releases the GK-110 to GeForce. One may hope, one may believe but what if they release a GK-114 to GeForce as their Flag ship single GPU?

Personally feel that they will eventually release the GK-110 to GeForce, but how will it perform, yield, power efficiency compared to a GK-114? What will the competitive landscape look like in the Spring of 2013? So many questions -- so little data.

 

[RussianSensation]

Let's discuss HD8900 series though since this thread is not about GK110 afterall.

I recall certain forum members make comments like these:

1) AMD can't increase die size to > 400mm^2 because the part would consume more power than 250W, and even implying it will consume more power than Fermi GTX480 since supposedly HD7970 GE uses as much power as a GTX690.....;

2) AMD can't add more ROPs than 32 since their GCN architecture "does not allow for it."

HD8970 rumored specs:

2560 SPs
48 ROPs
410-420mm^2 die size

Discuss please!

Alos, what are people's thoughts on HD8950 only have 32 ROPs vs. 48 ROPs on the 8970?

If true, that would mean consumers won't be able to buy a 2nd best AMD card and just overclock it to reach a $500 flagship (5850/ 6950 / 7950 could do it).

 

[SirPauly]

It may depend on how power efficient, innovative, over-all efficient their stream processors/arch is with the 8XXX generation. It's tough to gauge a current architecture with a future one, even if it is still on a same node.

 

[RussianSensation]

I completely disagree with just about everything you are saying in pretty much your entire post (except the hd5800's launch prices) and there has been plenty of instances to easily refute your view of the market, but I'm not going to take the time to start 10 page long arguments/discussions with you. Three easy instances are GTX580 and GTX570 launched before Cayman and both were faster, and 8800GT launched to no competition and undercut it's own lineup significantly.

Can you explain what you are disagreeing with?

This is my point: why would AMD have parts ready, but supposedly delay them on purpose to June 2013 and let NV launch first? That makes no sense at all unless AMD is having issues with yields, volume production or power consumption. There is no way AMD will allow NV to launch first if they are ready to launch. If NV launches first, it's because AMD couldn't beat them to market. Based on how HD7000 series launch went, AMD will want to launch first since it allows them to dictate prices and a small window of opportunity where competition is not yet fierce.

What I find surprising is that people are surprised that AMD finally changed its strategy, decided to launch first with high prices for early adopters when they had a clearly superior product in nearly every way, and then naturally were forced to drop prices over time when the competitor released a better product. We as consumers didn't like it but knew this was coming sooner or later since the price/performance strategy failed for 3 generations. This is what I said about NV doing it for years - launching flagship cards at $500-600. When NV priced their high-end chips at $500-600, you didn't say a word but $550 for HD7970 was a total failure supposedly even though it smacked 580 around. So what do you have to say about GTX260/280? Or about NV not lowering prices on GTX580 for more than a year? My point is NV always charged high prices for flagships and AMD could not since they didn't have a competitive enough flagship part (this is why I said HD5870 is an exception since AMD could have easily launched 5870 at $550 but they didn't because their management was incompetent to take advantage of the 6 months head start).

AMD's line-up lower price points even after GTX670/680 launch were still much higher than the previous strategy and market share was not lost = success. If anything that shows this strategy was way more successful = made more $, didn't lose desktop discrete market share. All AMD did was raise prices to historical ATI/NV levels and everyone got sticker shock. However, we knew AMD's GPU division was losing $ or making little $ and it was unsustainable. And actually, the expectation that AMD needs to go back to price/performance to sell HD8950 for $299 and HD8970 at $369 even highlights how the value of their graphics brand has eroded. AMD should try the old strategy of going back to ATI days of making a fast flagship card like 9800XT, X800XT PE and X1950XTX and sell them for $500-550. If you looked at last quarter's market share, AMD gained 2.5% and continues to sell HD7950/7970 for higher prices than 5850/5870 and 6950/6970 debuted. That means this strategy was successful. They made more $ and hardly lost any market share. That's a win-win.

Also, I never said anything about neither company not launching first ever or mid-range or low-end parts coming in at aggressive prices. Sure it happened before. I am talking about flagship products which is what the thread is about. The old ATI or NV didn't launch flagship products for $299-369. It's amusing that people think AMD should go back to their price performance strategy since their new pricing hurt *mind-share* when the majority of people saying this are the same people haven't bought a single AMD product since the small die strategy went into effect.

If GTX700 series is far superior, AMD can always go back to price/performance. However, I am liking strong competition at the top which can only happen if AMD releases high-end flagships and ditches the small die/"budget brand" strategy entirely. AMD going back to head-to-head vs. NVidia is the best thing to happen to the GPU industry. Before, NV almost took it for granted that they would have the $400-500 market all to themselves. Now, NV can't just sit there and expect AMD to flop and price their flagships at $299-369. That means a much more competitive NV as well. Because NV knows AMD is attacking the GPU market like the good old days, they'll try that much harder to make GTX780 even faster.

 

[SirPauly]

It wasn't the sticker shock for me but more-so the lack of performance with a substantial and significant node and arch - for both companies.

Consumer and company still had a win-win with AMD execution advantage -- AMD, could maximize their revenue and margins and the consumer had a 28nm choice that offered the fastest GPU on the planet. No one was forcing anyone to pay.

AMD gained 2.5 percent discrete desktop but also lost revenue sequentially, while nVidia may of lost 2.5 percent discrete desktop garnered 15 percent more GeForce revenue sequentially, and GTX performance/enthusiast GeForce sectors were exclaimed for part of the success.

 

[maniacalpha1-1]

I'm just curious if anything can be done about certain games, such as BF3, favoring Nvidia beyond the actual average performance differences? Or is that going to continue?

 

[RussianSensation]

You have to remember HD5850/5870/HD6950/6970 were all underpriced. This helped to make $550 price of 7970 seem especially high - sticker shock. At 1150mhz a 7970 is 70-75% faster in games than a 6970. After bitcoin mining, it's free. For people who took advantage of this perk, this was a great generation. On the NV side, a prospective gamer was stuck paying $500 for 35% more performance from the 580 and no way to use your GPU to make $ when not gaming. That's a far worse deal in my eyes. In fact after bitcoin mining, you could have easily walked out with 2 7970s for free. :thumbsup:

lost revenue sequentially[/B], while nVidia may of lost 2.5 percent discrete desktop garnered 15 percent more GeForce revenue sequentially, and GTX performance/enthusiast GeForce sectors were exclaimed for part of the success.

I think this is because the entire AIB GPU market has been shrinking lately. AMD gained market share in a shrinking discrete GPU market segment due to decline of the PC DIY and desktop PC market segments worldwide. Based on what I read on MarketWatch, most of the success for NV's earnings have been attributed to professional markets, notebook design wins (double of Fermis) and Tegra wins in Google Nexus 7, smartphone design wins, etc. I don't recall JHH mentioning that it was desktop Kepler their GeForce division was doing much better. Do you have a link for that? GeForce brand as a whole includes both desktop and notebook parts.

 

[exar333]

I don't agree with that. AMD tried this strategy for 3 generations in a row and it didn't work:

1) Didn't result in high consumer switching from NV;
2) Didn't result in significantly higher market share (actually NV ended with 62-63% desktop discrete after HD4800/5800/6900 series were all counted);
3) AMD's GPU division barely made $ on an annual basis.

It's pretty clear that strategy was great for consumers but a failure overall for the company.

Actually in some quarters, AMD's GPU division lost $ using this strategy. ATi's pricing method made was Premium, never price/performance, and it resulted in them producing very fast high-end GPUs. The end result was that company was hugely profitable and ATi's brand name was considered premium because they wouldn't dare sell an X850Pro / X850XT for $269/ $369 on launch day. If AMD goes back to price/performance ($299 8950 and $369 for 8970) voluntarily without GK110 clobbering them, it's an admit of failure for the new strategy the firm exercised (First Mover Advantage). You can't market yourself as a premium brand and sell a $500 GPU as a $350 GPU. Of course as customers, we would have loved a $350 HD7970 at launch but AMD made the right move since selling $300-350 flagship cards is not sustainable with rising wafer prices as continuous node shrinks get more expensive. $500 GTX680 could not even beat the 7970 series this round. So why should AMD have priced 7970 at $369? People have a serious double standard with paying $400-500 for AMD cards that perform the same or faster than a GTX680 this generation it seems. Even when unlocked 6950s were going for $250 and you could get 2 of those for the price of a 580, it still didn't really matter. Why continue on the same failed strategy that hasn't worked for 3 generations in a row? This was exactly why AMD rushed HD7900 series with 925mhz clocks since they wanted those high margins.

When GTX280 launched at $649 and even later fell to $499, it was still $200 more expensive than 4870 but 20% faster. HD7970 was easily 20% faster than a 580 without overclocking and cost $110 more than $440 market rate for 580s then. The main reason AMD got away with this pricing strategy is because NV went MIA on GK110 and delivered 680 that was barely faster. You can only put 50% of the blame on AMD. The other 50% has to fall on NV for way under-delivering with the GTX680.

This is the fundamental difference: It's impossible to claim that only AMD is ripping us off now this generation when NV is doing the same (and frankly has been doing so for 6 years with $500-600 flagship cards). If you want to say AMD is ripping us off, then so is NV. $500 GTX680 294mm^2 is not exactly the price/performance part. Ironically, AMD gets all the heat for raising prices because "well NV kept the price at $500". This is the same story Intel is pulling on us. The price stays the same but the die keeps shrinking. Damn right NV raised prices big time. They went from selling us a 520mm^2 384-bit bus chip to a 294mm^2 256-bit bus chip. NV definitely raising prices but did so in a sneaky way that only people who follow tech closely would see. Also, GTX660Ti is a real GTX460 replacement and price went up from $229 to $299.

Most ironic NV only had superior price/performance for 1 quarter (March to June), while 7970 went totally uncontested from January to March and then after June price cuts and Cats 12.7, again 7970/7970GE delivered better value and performance than 670/680s. NV continues to charge $580 for their premium 680, but a premium 7970 GE is $450. NV continues to overcharge by $75-100 on their 680 line overall and it's not getting any backlash. It looks like AMD has actually been delivering both the price/performance and performance since June. If anything, it's NV that continues to rip off consumers with $400-420 670 and $500-550 680s.

Kind of rediculous to say NV is 'ripping off their consumers'. If they are will to pay, so be it. NV has been able to charge a premium for their products vs. AMD. If they can get away with it, and make money, it's their decision I guess.

 

[SirPauly]

RS:

Actually the revenue dropped in the professional sector for nVidia last quarter.

Our Professional business was down 8% sequentially due to continued weak economic conditions in Europe and the slower ramp of Romley platforms.

On GeForce:

Revenue results by business segment were as follows. Our GPU business was up 15% sequentially due to the strong demand for our desktop performance and enthusiast products, and growing notebook share as the industry continues to move to the Ivy Bridge platform

http://seekingalpha.com/article/800...f2q13-results-earnings-call-transcript?page=2

 

[jackstar7]

I'm hoping that higher adoption rate of 1440p+ will force some pressure to put out the cards that can keep up.

 

[RussianSensation]

I'm just curious if anything can be done about certain games, such as BF3, favoring Nvidia beyond the actual average performance differences? Or is that going to continue?

Deferred MSAA hurts AMD cards more in BF3 in % terms but driver updates have allowed them to crawl back the performance delta to just 5-10%.

The performance difference in BF3 / expansions with current drivers is within a spitting distance.

With Cats 12.8

With Cats 12.7

With Cats 12.7

With Cats 12.7 Beta

Cats 12.7 Beta @ 1600P

I think the key will be how much faster GK110 could be over 8970. The performance difference in BF3 is actually very close now that it comes down to the overall speed of the GPU. Who knows what happens Medal of Honor: W now that it's an AMD Gaming Evolved title, while BF4 is far away.

 

[blackened23]

I'm hoping that higher adoption rate of 1440p+ will force some pressure to put out the cards that can keep up.

Intel has published several documents and they are actually pushing 4k resolution in 2013. Going forward, I imagine that GPUs will have better compatibility with 4k (currently, it requires 2x DVI-D connectors or Displayport) and this will be the focus. Super high resolution displays has become a brand differentiator in mobile, and the desktop will follow suit - and like I mentioned intel will be pushing 4k in a big way in 2013.

The other thing is, windows 8 is supposed to have scaling kinda like Mac OSX lion does, hopefully this will ease the transition a bit since 4k will obviously require a great deal of graphics power.

See here: Intel: "Higher resolution displays coming in 2013"

http://www.tomshardware.com/news/Intel-Higher-Resolution-Displays-Coming,15329.html

 

[RussianSensation]

Kind of rediculous to say NV is 'ripping off their consumers'. If they are will to pay, so be it. NV has been able to charge a premium for their products vs. AMD. If they can get away with it, and make money, it's their decision I guess.

Just as ridiculous as claiming AMD charging $400-500 for HD7970/7970 GE is ripping off customers. AMD did it by raising prices from 6970. NV is doing it by selling the lowest generational jump in its history by lowering die size. Either you claim both are doing it, or neither is doing it. But who got 99% of the heat for raising prices this generation? AMD, not NV. That's my point.

RS:
Actually the revenue dropped in the professional sector for nVidia last quarter.

On GeForce:

http://seekingalpha.com/article/800...f2q13-results-earnings-call-transcript?page=2

Thanks for the link! :thumbsup:

Keep in mind, despite Revenues falling, AMD made more $ using the new strategy:
Q1 2011 Revenue = 413M
Q1 2012 Revenue = 382M (less than 1% drop)

vs.

Q1 2011 Operating Income = 19M
Q1 2012 Operating Income = 34M (79% higher)
http://www.anandtech.com/show/5764/amd-q112-earnings-report-158b-revenue-590m-net-loss

^ I can look for Q2 and Q3 data if you want.

This is why you cannot just look at revenue to assess how successful a new strategy is. In a mature company, one of the most important components investors look at are cash flows, not revenue. If you have higher cash flows despite shrinking revenues, your business is actually better off long-term. I think AMD's discrete GPU division is doing better despite a decline in revenues. We'll have to see if they continue with the new premium pricing strategy or resort back to the old price/performance strategy. Of course a lot of that depends on how awesome GTX700 series will be.

AIB discrete GPU shipments continue to decline, and AMD's market share grows. Total AIB shipments decreased this quarter, from the previous quarter, by 6.5% to 14.8 million units. (see Figure 1).
"AMD introduced the new Radeon HD7000 series early in the quarter and as a result picked up market share."
http://jonpeddie.com/press-releases...-shipments-seasonally-down-from-last-quarter/

Sounds to me like high prices on HD7000 series made the GPU unit more profitable and didn't negatively affect market share. By launching first, AMD won market share and earned higher profits. From a business perspective, that's a win-win for the new strategy, despite a drop in revenues.

This is why I don't believe that AMD will purposely allow NV to launch GK110 first and somehow hold back HD8970 on purpose. I think if NV launches first, it'll be because they executed better than AMD.

 

[Olikan]

2) AMD can't add more ROPs than 32 since their GCN architecture "does not allow for it."

not true...
...GCN is very flexible at rops count,
both the 7870 and 7970 have 32 ROPs, yet 256-bus and 384-bus

 

[RussianSensation]

not true...
...GCN is very flexible at rops count,
both the 7870 and 7970 have 32 ROPs, yet 256-bus and 384-bus

Don't tell me, ask the people who made that comment earlier. That's why the rumors of 48 ROPs keep contradicting the position that AMD cannot improve its ROP performance much.

 

[scooterlibby]

Kind of underwhelmed at the prospect of next gen cards. Who will need the flagship single GPU other than multi-monitor players? The most taxing games out there I've played have been Metro 2033, BF3, and a heavily modded Skyrim. My two 580's destroy them. Hopefully the upcoming generation of consoles will drive a graphics push with PC games and we'll see something that justifies an 8970 or a 780.

 

[iCyborg]

AMD gained 2.5 percent discrete desktop but also lost revenue sequentially, while nVidia may of lost 2.5 percent discrete desktop garnered 15 percent more GeForce revenue sequentially, and GTX performance/enthusiast GeForce sectors were exclaimed for part of the success.

nVidia Q1->Q2 going from pretty much no competition to a launch of a highly competitive product + IB launch
Is it really a surprise that their discrete GPU revenue picked up in Q2?

AMD Q1->Q2 going from no competition from nV at high end to a competition against gtx 670/680. Given that it was a seasonally down quarter, I'd call going down only 4% a solid result given the circumstances.

 

[blackened23]

Kind of underwhelmed at the prospect of next gen cards. Who will need the flagship single GPU other than multi-monitor players? The most taxing games out there I've played have been Metro 2033, BF3, and a heavily modded Skyrim. My two 580's destroy them. Hopefully the upcoming generation of consoles will drive a graphics push with PC games and we'll see something that justifies an 8970 or a 780.

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.

 

[RussianSensation]

I think Wii U is only DX10.1. Iirc, PS4/Next Xbox are not due until Q4 2013/Q1 2014. 4K displays are far too expensive even for enthusiasts. $20-30k?

I am hoping Metro Last Light, Crysis 3 and some other new game has next gen graphics. Right now it seems the PC gaming industry has stagnated in terms of graphical development. Whatever changes we have seen since Crysis and Metro 2033 have been incremental at best.

Enthusiasts will always find a way to use up more GPU performance with Super-Sampling, 3D vision, multi-monitors but that's probably less than 20% of the market. For the rest of the market though, GPUs are getting so fast and games are falling behind more and more, making upgrading less and less worthwhile each new generation. Even the list of upcoming PC games has an overwhelming number of console ports.

 

[SirPauly]

nVidia Q1->Q2 going from pretty much no competition to a launch of a highly competitive product + IB launch
Is it really a surprise that their discrete GPU revenue picked up in Q2?

AMD Q1->Q2 going from no competition from nV at high end to a competition against gtx 670/680. Given that it was a seasonally down quarter, I'd call going down only 4% a solid result given the circumstances.

Considering AMD had much more of a complete 28nm family, no vocal problems about yields or constrains, wouldn't consider going down only 4 percent a solid result but may of felt disappointment and pressure to improve revenue and sales by price adjustments, imho.

 

[RussianSensation]

Revenue declined less than 1% in Q1, operating income increased 78%. Cash flow is what allows the company to pay its debt obligations and continue operating, not revenue. In fact, the value of any company is the greater of the value of its underlying assets (say a real estate business) or the discounted present value of all its future cash flows (a technology company). Revenue is just a top line of an income statement and by itself is almost meaningless unless you are talking about an early stage start-up company, which has not yet generated sufficient cash flows on which to base a proper valuation. If a company increased its revenue 10x but its free cash flows have not improved, it accomplished little. For example, if you sold 1 GPU for $100 and made $50 on each, that's better than selling 10 GPUs for $1000 and making $40 on all 10 of them. Your revenue increased 10x but you made less $. I don't want to derail the thread into a finance discussion but a company could still be better off despite declining revenues.

Moreso, the entire argument is illogical in nature. High prices hurt AMD's mind-share. What does that even mean for people who cross-shop at current prices. If you were buying a card between March and early June, GTX670/680 were great. Before that NV had nothing, and after they are offering worse price/performance. So NV only had "good mindshare" for just 4 months of the year. AMD priced high for 2.5 months when there was no competition from NV. After AMD lowered prices, its products offered similar or better price/performance to NV. For anyone buying after the price drops, it doesn't matter at all what HD7900 cost in January. Since June 2012, AMD delivered either similar or better price/performance in nearly every price category, especially so in foreign countries where NV tends to have a tangible price premium even now. Until August 2012, AMD delivered better price/performance in every desktop category below $300 since NV had no 28nm parts under $300 until August 16th. You can say HD7850 and 7870 were too expensive vs. 6950/6970, but against GTX570/580 they still offered superior price/performance and performance/watt for 6+ months. AMD got away with it because what NV was selling on 40nm was even worse. Competing against 7850/7870 with GTX560Ti 448 / 570 and 580 was a joke and a half. Of course people still bought GTX285 when AMD had 5850. I am not really surprised that NV was able to sell the inferior 40nm Fermi cards for 6 months against the newer 28nm 7800 series without many consequences for being late......I think AMD needs to deliver cards 2x faster for 50% less before NV loyalists start switching.