GM200 closer than we thought [wccf tech]

[ShintaiDK]

The problem is in Canada a 980 sells at more or less flagship prices of $650+ CDN,

You do know 550USD is 660CAD, right?

 

[RussianSensation]

You do know 550USD is 660CAD, right?

Misleading comment since before oil prices tanked, and CDN dollar was stronger, that $650 CDN 980 was more like $650 US after tax for us. Cards like G1 980 are nearly $700 CDN before tax, which after taxes and today's FX is about $650 US:

http://m.newegg.ca/Product/index?itemnumber=14-125-682

I also know that $700-750 CDN used to buy flagship NV cards for years, which is almost what an after-market 980 costs here. You also ignored my comment on how even US pricing for NV moved up. Were 5700Ultra, 6600GT, 7900GT, 8800GT, GTX260 216, GTX460, 560Ti $500-600 cards like 680 2/4GB and 980 are? No they weren't but every single one of those was a Next gen mid-range card that outperformed the previous gen NV flagship. Since GK104 GTX680 launched, you have continued to refuse to accept these facts that NV bifurcated a generation and created 2 "flagships" out of 1 architecture/ generation. Grooveriding's point that NV now sells mid-range next gen x80 cards for near flagship prices is correct. Never in the history of NV has a true next gen flagship only beat the last gen flagship by only 10-35%. At least 680 was 35% faster which in hindsight is not too bad; the 980's performance at $550-600 USD is simply awful. That's why I recommend 970 SLI right now for someone looking to spend $550-600 on NV.

I am not placing 100% of the blame on NV or AMD for moving up mid-range prices but just stating how it is. If consumers keep paying these prices, that's how the market works but it doesn't change that GM204
980 is a mid-range Maxwell chip. Looking forward to GM200, the true Maxwell flagship. With overclocking + lots of functional units, it should scale really well.

 

[ShintaiDK]

S

I also know that $700-750 CDN used to buy flagship NV cards for years, which is almost what an after-market 980 costs here.

Whatever you could in the past is completely irrelevant if the currency exchange rates changes. Not to mention inflation. Try Keep up with the times, tho it can be hard when you are on your "quest".

http://www.xe.com/currencycharts/?from=USD&to=CAD&view=5Y

 

[Ancalagon44]

Yep. While I don't regret getting a 980, I'm done with cyclical upgrading. I want 2x performance at the same TDP or lower for ~$400 for my next upgrade. Guess I'm waiting until 16nmFF Volta or GCN 3.0.

Given the rumours that we have all heard about 20nm, you might need to wait for the generation after GM200 for a decent upgrade over 28nm. There will still be a lot of yield and leakage issues.

 

[RussianSensation]

Whatever you could in the past is completely irrelevant if the currency exchange rates changes. Not to mention inflation. Try Keep up with the times, tho it can be hard when you are on your "quest".

http://www.xe.com/currencycharts/?from=USD&to=CAD&view=5Y

Ok, let's ignore all the points I made above. Maybe try to actually read what I post instead of trying to make smart remarks as per your usual posting style. Maybe you can tell us when those next gen mid-range cards from NV I listed above cost $600-700 CDN? :hmm: I do live up with the times which is why I do look at history and skip overpriced mid-range cards like 680 4GB and 980. If you want to pay near high-end prices for mid-range cards and ignore 20 years of GPU trends and prices just because March 2012 changed things, it's entirely your choice but doesn't change historical data of what the average mid-range pricing for NV was in either US or Canada.

In the past I also outlined some other points why GM200/390/390X are worth waiting for for 680/7970/7970Ghz users that you also ignored. The opportunity cost of waiting is one of them.

January-March 2012 - Buy 7970/680

November 2013 - buy 780Ti for $700

September 2014 - buy 980 for $150 less for 10% more performance over 780Ti

Summer 2015 - buy 390X/GM200 for $700 with 30% more performance than a 980. The benefit of waiting for 390X/GM200 to get a lot more performance over the 980 is greater than the benefit of waiting for a 980 when one could have just gotten a 780Ti in late 2013 and enjoyed it all this time. I also typed out the price/performance which showed 980 costs the most for the extra percentage gained over last gen flagship compared to any other $500+ NV card. I mean one doesn't even need to do math since no Next gen NV card was just 10% faster than a 780Ti. Even the refresh 580 over 480 brought more out of the gate.

I expect 2nd best SKUs like 390 and GM200 780 successor to deliver a lot more performance at $550-600 US over the 980 than the 980 did in nearly a year over the 780Ti. I'll make sure to point it out to you too. You also seem to forget that many high end enthusiasts on our board game at 1600p and even 4K, where 980 brought very little change in overall gameplay ability over the 290X/780Ti. This is why A LOT of high end gamers are patiently waiting for GM200/390 series.

 

[ShintaiDK]

Ok, let's ignore all the points I made above. Maybe try to actually read what I post instead of trying to make smart remarks as per your usual posting style. Maybe you can tell us when those next gen mid-range cards from NV I listed above cost $600-700 CDN? :hmm: I do live up with the times which is why I do look at history and skip overpriced mid-range cards like 680 4GB and 980. If you want to pay near high-end prices for mid-range cards and ignore 20 years of GPU trends and prices just because March 2012 changed things, it's entirely your choice but doesn't change historical data of what the average mid-range pricing for NV was in either US or Canada.

Its irrelevant to use CDN when the cards are traded in USD.

 

[skipsneeky2]

This really has been the first time i have been more excited about the second to best then the top dog.The whole 680/670 to 780/770 transition really taught me how lousy the"high end" has been lately on the Nvidia camp.

Nvidia is really gonna play the fool most likely again,gimped 256 bit 4gb"high end",big ass $1k Titan,lets release a 6gb "1080" for $499 to attack the 380x and drop the 980 pricing.The 770 took nearly a whole year to release at the same price the 670 debuted at lol.

 

[Silverforce11]

Imagine the margins on the 980. NV doesn't need to release GM200 early unless AMD force them to with a good R380X that stomps all over the 980.

Competition is good people. Don't forget that.

 

[XiandreX]

This is really AMD's fault.

They make high-performing chips, yes, but they require a ton of power to be thrown at it to get that performance, which means a ton of heat.

Nvidia has found that for the past few generations, they didn't need to throw their large chip into the consumer space simply to stay competitive. On the contrary, their typically mid-range parts, with low TDP, are capable of keeping up with parts AMD needs to throw nearly twice the TDP with die's twice as large.
If you were Nvidia, what would you do? It sucks for us, who were used to the beast chips at the current 980 prices, but for them, if they can compete with what AMD is producing, but with fewer transistors on a smaller die, then they can likely have a higher profit margin thanks to a lower production cost.

They even priced lower than AMD's original pricing, and AMD was forced to cut the price. AMD is almost bleeding due to their current prices, but Nvidia forced their hand.
I've been preferring Nvidia for my last few purchases (save for the AMD 7850, I think, that is in my HTPC), but we all need AMD's next generation to really push Nvidia. I want a big chip from Nvidia at fair market pricing, but Titan and 780 Ti type prices are frankly ridiculous.

This is the reason I bought the GTX 970. Its fast, quiet, runs cool and has a low TDP. The last Nvidia card I owned was Geforce 2.
From this point on, power consumption and heat (go hand in hand) will be my deciding factor if performance is close enough.

 

[Mondozei]

It's going to be 28 nm. Nvidia has more or less already ruled out 20 nm. The big question is on the flagship 390X. AMD has not categorically said that the high-end GPUs will be on 20 nm, but it has said that some GPUs will and it makes sense to do it on the cutting edge.

If that's so, then it would explain why Nvidia is rushing out full fat Maxwell.

We all know what Nvidia can play this year, full fat Maxwell at 28 nm. That's their card. And because we all know it, AMD has the luxury of the initiative, of timing. As such, the GM200 will come when the 300 series come. I doubt Nvidia will try to pre-empt them, only to find themselves in a similar Titan Z-situation like last year.

 

[RussianSensation]

Imagine the margins on the 980. NV doesn't need to release GM200 early unless AMD force them to with a good R380X that stomps all over the 980.

Competition is good people. Don't forget that.

Thumbs up! When 980 is selling like hot cakes, NV will either launch a high priced GM200 (aka to follow the footsteps of a $1000 Titan --> $650 780, while keeping the best for later), or they will just milk 980 for another 3-4 months and launch GM200 in anticipation of 390X's spring/summer launch. Either way NV has little incentive to lower the 980's price from $550-600. Thus, if GM200's true 780Ti successor launches now, I can't see it being $650-699 and 30-40% faster than a 980 as that would make the 980 irrelevant. NV is cashing in big time on the perf/watt trends. The longer they can sell GM204 at $550+, the better it is. It doesn't rule out GM200, but to me it rules out a $650 GM200 right now.

 

[Anarchist420]

too bad nvidia's drivers suck, that their advertising isn't entertaining, that they don't support adaptive sync, and that even with Gsync you have vsync lag or tearing at < 30 fps. and the 980 has about the same DP performance per clock as the 780 which is pretty low.

but then it doesn't matter anyway, as there aren't any good monitors.

 

[n0x1ous]

I don't think they can get away with multiple cut down GM200's like they did with GK110 (Titan and vanilla 780).

Reason. GK104 to GM204 is +100mm2 die size. they don't really have that much die size room to grow to 650mm2 on GM200? I think they will need the fully functional GM200 to make the expected increase over 780ti/980 etc.

 

[RussianSensation]

I don't think they can get away with multiple cut down GM200's like they did with GK110 (Titan and vanilla 780).

Reason. GK104 to GM204 is +100mm2 die size. they don't really have that much die size room to grow to 650mm2 on GM200? I think they will need the fully functional GM200 to make the expected increase over 780ti/980 etc.

What to do with non-fully functional GM200 chips? It's too expensive to throw them out. Chances are have GM200 2nd tier at $599, GM200 flagship at $749. Drop 980 to $399-449. I expect this if 300 series brings solid competition.

Think about it, NV goes from $330-350 on a 970 to a $550-600 980 today for 16-19% more performance. You don't think it's possible to have a cut-down GM200 with 16-19% more performance over the 980 between 2048 and GM200's rumoured 3072 CUDA cores? It could actually make a 384-bit, CUDA core/TMU cut down GM200 a great value when overclocked against. 780 scales well when overclocked but stock performance was not great.

 

[n0x1ous]

What to do with non-fully functional GM200 chips? It's too expensive to throw them out. Chances are have GM200 2nd tier at $599, GM200 flagship at $749. Drop 980 to $399-449. I expect this if 300 series brings solid competition.

Keep them for Quadro and Tesla where it wont make nearly as much difference because they don't have much competition there?

I could see one cut down chip but hopefully not 2 like GK110

You made good points though as usual.

 

[BlockheadBrown]

...NV is basically selling mid-range next gen at near flagship prices...

Completely agree. The 980 should be at 970 prices. The 970 should be the $200-250 part and the 960, the $150-180.

Maybe Nvidia should have been more careful with their "4x performance" slides when first discussing Maxwell years ago.

 

[5150Joker]

Misleading comment since before oil prices tanked, and CDN dollar was stronger, that $650 CDN 980 was more like $650 US after tax for us. Cards like G1 980 are nearly $700 CDN before tax, which after taxes and today's FX is about $650 US:

http://m.newegg.ca/Product/index?itemnumber=14-125-682

I also know that $700-750 CDN used to buy flagship NV cards for years, which is almost what an after-market 980 costs here. You also ignored my comment on how even US pricing for NV moved up. Were 5700Ultra, 6600GT, 7900GT, 8800GT, GTX260 216, GTX460, 560Ti $500-600 cards like 680 2/4GB and 980 are? No they weren't but every single one of those was a Next gen mid-range card that outperformed the previous gen NV flagship. Since GK104 GTX680 launched, you have continued to refuse to accept these facts that NV bifurcated a generation and created 2 "flagships" out of 1 architecture/ generation. Grooveriding's point that NV now sells mid-range next gen x80 cards for near flagship prices is correct. Never in the history of NV has a true next gen flagship only beat the last gen flagship by only 10-35%. At least 680 was 35% faster which in hindsight is not too bad; the 980's performance at $550-600 USD is simply awful. That's why I recommend 970 SLI right now for someone looking to spend $550-600 on NV.

I am not placing 100% of the blame on NV or AMD for moving up mid-range prices but just stating how it is. If consumers keep paying these prices, that's how the market works but it doesn't change that GM204
980 is a mid-range Maxwell chip. Looking forward to GM200, the true Maxwell flagship. With overclocking + lots of functional units, it should scale really well.

^^This guy's never heard of inflation, R&D expenditure increases or exchange rate fluctuations.

Imagine the margins on the 980. NV doesn't need to release GM200 early unless AMD force them to with a good R380X that stomps all over the 980.

Competition is good people. Don't forget that.

So true, I'm glad AMD is around. Actually hope they can get some more market share and start making profit on a more consistent basis. One of the biggest blunders in history was the sale of ATi to AMD - especially considering the GPU division is what keeps AMD afloat these days. Too bad the geniuses at AMD sold their mobile phone GPU IP to qualcom, AMD could have been a much stronger company today if it had good leadership.