What will be AMD'S next Move?

[krumme]

What do you think it would happen to AMD if they lose some 100-120MM in gross profits per quarter, knowing that they only have 450MM in gross profits to live with it to pay its R&D bills? Do you really think they could withstand a blow of this size?

Can we get things in perspective. This 970/980 hysteria needs to come down to reality.

Gtx970/980 is aprox 400mm2 or more or less the same as a 290x -10%. This is not a 680 vs. 7970 situation, where nv clearly did hold a significant lead of cost or a 550mm2 780 vs. the 290 where the 780 was way behind.
And as for efficiency in a desktop. People was buying 480 in huge numbers back then even if it was made with expensive tantalum capacitors and was a toaster by its time and they will continue to do so to get best gaming performance.

 

[krumme]

AMD did manage to fix this but the damage was done. Few will buy a notebook with high end AMD GPUs and not many sellers want to take the risk of selling one. Enduro is still worse than optimus but its significantly better than before.

Optimus is a good example of a technology that really helps users, and gives a tangible benefit every single day all over. AMD have been way to late to implement that kind of user benefit, and is instead focusing of expensive engineering heavy technology like HSA that - perhaps- in years will give a benefit.

 

[sontin]

Can we get things in perspective. This 970/980 hysteria needs to come down to reality.

GM204 is 10% bigger than Tonga and will provide around 50% more performance while using less power.

Why dont we look this way? :hmm:

 

[SirPauly]

This is not a 680 vs. 7970 situation, where nv clearly did hold a significant lead of cost or a 550mm2 780 vs. the 290 where the 780 was way behind.

I don't know:

http://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/27.html

And as for efficiency in a desktop. People was buying 480 in huge numbers back then even if it was made with expensive tantalum capacitors and was a toaster by its time and they will continue to do so to get best gaming performance.

Q2 and Q3 of 2010 AMD did retake discrete leadership --- It wasn't 'till the refresh of Fermi where there was separation once again.

 

[guskline]

3DVagabond, legit question about games. All three of my rigs had Nvidia cards in them before I bought the first R9 290. I had a EVGA GTX 780 Classified in the 3930k; I have 2 GTx 670FTWs in SLI in my 3770k and I had a GTX 680 in my FX8350.

I got the bug to buy a R9 290 and unloaded the GTX 680. When the chance came to Crossfire the R9 290 I moved both of them to my 3930k rig and put both of them under water. I had tons of rad room and keeping them watercooled addressed one of the R9 290/290x issues, heat.

Candidly, I have a number of games but keep coming back to COD Ghost. It has trouble running in CF but runs great with Nvidia drivers.

BF4 runs well in DX11 but faster in Mantle.

I ran some MetroLL benchies with my rig below and at 2560x1440 and all settings maxed and SSSA? enabled I got 51fps average. Without SSSA it was 83fps. A poster said SSSA is equivalent to running in 4k.

It runs IL2 BOS very well.

Bioshock Infinite runs well also.

AMD needs to focus more on the drivers for the R9 series; however, I think they are making an honest effort with a very limited staff.

Need to see more games actually taking advantage of Mantle.

I cannot argue against the more power for better efficiency concept.

Hope this answered some of your questions.

If the 970 was out when I wanted to upgrade from the 680 to the R9 290, I would have bought the GTX 970.

 

[krumme]

GM204 is 10% bigger than Tonga and will provide around 50% more performance while using less power.

Why dont we look this way? :hmm:

Why do you ask, when you perfectly know the answer to that yourself? :hmm:

In q1 we will see what gcn3 is all about. Until then 290 have been depreciated for the most of a year and have about the same size as 970. Amd will just lower prices, and have plenty room to do so. Thats how it works and always have.

 

[krumme]

Q2 and Q3 of 2010 AMD did retake discrete leadership --- It wasn't 'till the refresh of Fermi where there was separation once again.

And the difference was that 5870 was small and 480 was big so AMD could sell 5870 way below fermi. And they hit that low price from day 1.

The numbers here is 400 vs. 440 mm2.

As for the performance vs. watt, i guess most is interested in performance and secondly performance vs. price for the product when we talk a product of this highend caliber in a desktop. And its not like you are going to run that 970 of a single 6pin - lol.

 

[Keysplayr]

Can we get things in perspective. This 970/980 hysteria needs to come down to reality.

Gtx970/980 is aprox 400mm2 or more or less the same as a 290x -10%. This is not a 680 vs. 7970 situation, where nv clearly did hold a significant lead of cost or a 550mm2 780 vs. the 290 where the 780 was way behind.
And as for efficiency in a desktop. People was buying 480 in huge numbers back then even if it was made with expensive tantalum capacitors and was a toaster by its time and they will continue to do so to get best gaming performance.

The perspective you ordered.
http://www.pcper.com/reviews/Graphi...and-GTX-970-GM204-Review-Power-and-Efficiency

I could be wrong but it might just be your perspective that needs some alteration.

 

[krumme]

The perspective you ordered.
http://www.pcper.com/reviews/Graphi...and-GTX-970-GM204-Review-Power-and-Efficiency

I could be wrong but it might just be your perspective that needs some alteration.

The perspective that 680 is 294mm2 and 980 is 398mm2?

 

[krumme]

As for the question "What will be AMD'S next Move?"

My strategy would be to implement a "super super balancing" tech on 600mm2 cards with a 195Watt tdp, albeit with 2 eight pins for power, and ask people to mod their ps with capacitors the size of dino balls.

 

[SirPauly]

And the difference was that 5870 was small and 480 was big so AMD could sell 5870 way below fermi. And they hit that low price from day 1.

The constructive nit-pick was performance/watt and not its size, imho.

As for the performance vs. watt, i guess most is interested in performance and secondly performance vs. price for the product when we talk a product of this highend caliber in a desktop. And its not like you are going to run that 970 of a single 6pin - lol.

I don't know what is funny but efficiency is key for architectures over-all, imho! I'll allow the market to decide what most are interested in.

 

[Keysplayr]

The perspective that 680 is 294mm2 and 980 is 398mm2?

So you are calling appreciation for the 970/980 price, performance and power consumption "hysteria" because of the die size difference between GK104 and GM204?
Is that reasonable?

 

[alexruiz]

I am just telling facts and if u dont like it than than plz u dont need to post.

Facts is a reality not flaming.

No you are not, you are posting opinions.
If you really wanted to have good answers, you should have started with solid neutral comments. You are not, all of what you type is completely biased despite asking for "not flaming" and you have been called by more one poster.
You started the thread as a flamebait, and in fact, I am sure the mods already should have an eye on this thread and close it, as you obviously started it to troll.

The GTX980 is fast, but expensive; doesn't change anything other that obsoleting the very card you have, and hence why you are so passionate trying to justify your choice. The GTX970 is the one that shakes the market, but the R9 290 all it needs is a minor price adjustment, as it had been hovering around 340-$350 NEW often, I know it, I got a Diamond card (HIS IceQ x2 rebadge) at that price. The R9 290X situation doesn't change, it was already obsoleted by its own sibling. A R9 290 at $290-$300 is a perfect alternative for multi-GPU users, as crossfire scales better than SLI.

 

[desprado]

Well we all hoped that something was coming from AMD on 25th September well sadly that is not the case so the there some current product which is launching in India so move alone folks and discuss some other possibilities.

 

[SlowSpyder]

I don't think AMD is too worried. Nvidia launched a good part, no doubt about it. AMD will respond. I remember when no one thought the relatively inexpensive R9 290 would keep up with Titan or the GTX780. Right now AMD should lower prices to keep parts moving, keep working on the next GPU. Not too big of a deal to me, this tug of war tends to go back and forth.

 

[mrmt]

Can we get things in perspective. This 970/980 hysteria needs to come down to reality.

Gtx970/980 is aprox 400mm2 or more or less the same as a 290x -10%. This is not a 680 vs. 7970 situation, where nv clearly did hold a significant lead of cost or a 550mm2 780 vs. the 290 where the 780 was way behind.
And as for efficiency in a desktop. People was buying 480 in huge numbers back then even if it was made with expensive tantalum capacitors and was a toaster by its time and they will continue to do so to get best gaming performance.

I agree with you. NVidia doesn't have a good area advantage so the costs should be more or less the same for both parties. What Nvidia is trying to do is to get a premium for its higher efficiency. I expect Nvidia revenue share to grow in the next two quarters.

That said, I don't expect AMD market share to fall 50%, I was just pointing out to RussianSensation that AMD has to manage its cash inflows very carefuly, because they are in a very tight situation on the matter. AMD must carefuly calibrate how much revenue volume they can afford to lose to Nvidia otherwise they will burn too deep into their thin cash reserves.

 

[StereoPixel]

AMD's R9 390X will not become available until the first half of 2015, but the GPU vendor has already started cutting some of its GPU prices to boost sales

http://digitimes.com/news/a20140924PD203.html

 

[krumme]

I agree with you. NVidia doesn't have a good area advantage so the costs should be more or less the same for both parties. What Nvidia is trying to do is to get a premium for its higher efficiency. I expect Nvidia revenue share to grow in the next two quarters.

That said, I don't expect AMD market share to fall 50%, I was just pointing out to RussianSensation that AMD has to manage its cash inflows very carefuly, because they are in a very tight situation on the matter. AMD must carefuly calibrate how much revenue volume they can afford to lose to Nvidia otherwise they will burn too deep into their thin cash reserves.

Nvidia can get a higher price because of superior brand value. Brand value is also having a huge impact on what matters most be it performance, what features, efficiency. Nvidia is better in making a business and always have been. Thats positive in my book, and its also translated into a more user oriented tech development as i mentioned.

As for AMD cash inflow. Yes you are probably right, but havnt that been a problem for years now? not to mention the lack of profit and still keep it going as the basic problem - we just tend to forget how much the big cores have eaten of the R&D budget and entire capacity for the prior years. I dont think its any worse - but ofcource that depends on expectations. At least AMD is not in a position to apply so much more lipstick to the results, so we have a better and more realistic picture today than 3 years ago.

Besides from the consoles, and to a smaller degree beema, if there is a place where AMD is actually somewhat a little competitive its on the desktop gpu (and that sad situation does explain the lack of profit - lol ). So long doomsday threads because of the 970/980 is a farce.

 

[krumme]

So you are calling appreciation for the 970/980 price, performance and power consumption "hysteria" because of the die size difference between GK104 and GM204?
Is that reasonable?

Well for your information this is not the thread:

"OMG the 400mm2 970 is the best thing since sliced bread its 20% faster than the year old 290 of similar size - just remember to have two 6 pins and a 550w powersupply"

We actually have 3 more or less similar threads to that - and fine for me, thats tech enthusiasm. Thats appriciation.

This is the thread:
[OMG ...insert yourself]
"What will be AMD´S next move?"

This is something entirely different. And its not "appriciation".

 

[Dribble]

I agree with you. NVidia doesn't have a good area advantage so the costs should be more or less the same for both parties. What Nvidia is trying to do is to get a premium for its higher efficiency. I expect Nvidia revenue share to grow in the next two quarters.
.

Costs won't be the same - the lower power usage and simpler memory requirements of maxwell lead to cheaper simpler boards. Really AMD can't complete with the 9 series - it almost certainly costs less to make, it goes faster, it's cooler/quieter, it's shiny new, and it has superior nvidia marketing backing it up. They just have to hope nvidia takes their time releasing mid range and lower end cards.

 

[krumme]

Costs won't be the same - the lower power usage and simpler memory requirements of maxwell lead to cheaper simpler boards. Really AMD can't complete with the 9 series - it almost certainly costs less to make, it goes faster, it's cooler/quieter, it's shiny new, and it has superior nvidia marketing backing it up. They just have to hope nvidia takes their time releasing mid range and lower end cards.

The board cost is of minor significance in the broader picture. Remember the lower freq ddr5 and 290 beeing on market for a year giving better yields and binning. And then the depreciation as the most important factor bar none. 10 months on the market is a very long time. Looking at BOM, is not going to cut it half way. If anything AMD have a cost advantage.

There is tons of reasons AMD is fucked, but cost disadvantage vs. 980/970 on the desktop, is absolutely the last of it.

 

[PPB]

Well for your information this is not the thread:

"OMG the 400mm2 970 is the best thing since sliced bread its 20% faster than the year old 290 of similar size - just remember to have two 6 pins and a 550w powersupply"

We actually have 3 more or less similar threads to that - and fine for me, thats tech enthusiasm. Thats appriciation.

This is the thread:
[OMG ...insert yourself]
"What will be AMD´S next move?"

This is something entirely different. And its not "appriciation".

Yep, seems this thread was a not-so-camouflaged attempt at claiming AMD, for a billionth time in their history, is doomed. Except not only they are not doomed, they are actually in a better possition than 7970 vs 680 2 years ago.

7970 vs 680
20% larger die (50% larger bus bandwith)
10-15% perf advantage for the 680
10% higher price for the 7970 (12.5% higher price for the 7950vs 670)
50% more VRAM for the 7970
15-20% delta perf watt in gaming (0% perf/watt difference in GPGPU)
new uarch for both companies (3 months difference between each product)

290x vs 980

10% larger die (100% larger bus bandwith)

15-20% perf advantage for the 980
0% pricing difference (21.21% higher price for the 290 vs 970)
Same ammounts of VRAM for both cards
30-40% perf/watt in gaming (0% perf/watt difference in GPGPU)
NVs new uarch vs amd's n-1 uarch (1 year difference between each product)


This shows that from a cost standpoint, AMD is better suited this time to lower 290/x prices than they were on 7970 launch (less die size difference even sporting a bigger bus bandwith difference, same ammounts of VRAM for both companies now vs 50% more VRAM for the 7970 vs 680, etc).

From the gaming efficiency standpoint there isnt much to be done as now AMD is competing to a quite newer product compared to tahiti launch, which can be said it was pretty even in this regard. That said, I have seen Tom's power analysis on Hawaii and their power consumption spikes (240-50 average with spikes to 420) have a lesser average to spike ratio than what has been seen on GM204 (170 avg and 290 spikes), and both will require you to have a pretty well built power supply to endure that kind of spiking delta. It is obviously easier to build a SLI of GM204 than it is for Hawaii, no doubt about it.

The answer in the end is pretty simple: AMD is not going anywhere, the hysteria indeed is disproportionate (and is coming from the usual suspects as always), but what is important is that this competition is good for us as costumers as this launch will obviously lead AMD to adjust their Hawaii pricing (which is easier for them this time compared to the Tahiti vs GK104 situation a long while ago, because of the points shown above). They will still be behind in perf watt, something expected from a newer uarch vs a 1yr old one. But hey, if GCN 1.2 changes regarding tesselation and bus bandwith efficiency are any indicator, it shows that they are trying to catch up in every indicator they were actually behind, while still focusing in the thing that will actually make them make more money, which is GPGPU and the professional market.

EDIT: Something I forgot: Bus bandwith is a bigger die area cost incurred by whoever implements it. However, it allows you to use lower binned VRAM modules, which also impact in the price. In their whole SKU range since Kepler, NV was forced to use 1500mhz rated modules, while AMD could use 1250mhz rated ones because of the 50% bigger bus width. Now with Maxwell, they are forced to use 1750mhz rated vram.

 

[Keysplayr]

Yep, seems this thread was a not-so-camouflaged attempt at claiming AMD, for a billionth time in their history, is doomed. Except not only they are not doomed, they are actually in a better possition than 7970 vs 680 2 years ago.

7970 vs 680
20% larger die (50% larger bus bandwith)
10-15% perf advantage for the 680
10% higher price for the 7970 (12.5% higher price for the 7950vs 670)
50% more VRAM for the 7970
15-20% delta perf watt in gaming (0% perf/watt difference in GPGPU)
new uarch for both companies (3 months difference between each product)

290x vs 980

10% larger die (100% larger bus bandwith)

15-20% perf advantage for the 980
0% pricing difference (21.21% higher price for the 290 vs 970)
Same ammounts of VRAM for both cards
30-40% perf/watt in gaming (0% perf/watt difference in GPGPU)
NVs new uarch vs amd's n-1 uarch (1 year difference between each product)


This shows that from a cost standpoint, AMD is better suited this time to lower 290/x prices than they were on 7970 launch (less die size difference even sporting a bigger bus bandwith difference, same ammounts of VRAM for both companies now vs 50% more VRAM for the 7970 vs 680, etc).

From the gaming efficiency standpoint there isnt much to be done as now AMD is competing to a quite newer product compared to tahiti launch, which can be said it was pretty even in this regard. That said, I have seen Tom's power analysis on Hawaii and their power consumption spikes (240-50 average with spikes to 420) have a lesser average to spike ratio than what has been seen on GM204 (170 avg and 290 spikes), and both will require you to have a pretty well built power supply to endure that kind of spiking delta. It is obviously easier to build a SLI of GM204 than it is for Hawaii, no doubt about it.

The answer in the end is pretty simple: AMD is not going anywhere, the hysteria indeed is disproportionate (and is coming from the usual suspects as always), but what is important is that this competition is good for us as costumers as this launch will obviously lead AMD to adjust their Hawaii pricing (which is easier for them this time compared to the Tahiti vs GK104 situation a long while ago, because of the points shown above). They will still be behind in perf watt, something expected from a newer uarch vs a 1yr old one. But hey, if GCN 1.2 changes regarding tesselation and bus bandwith efficiency are any indicator, it shows that they are trying to catch up in every indicator they were actually behind, while still focusing in the thing that will actually make them make more money, which is GPGPU and the professional market.

980 isnt really what puts 290X in the hot seat though. That title belongs to the 970. 970 kind of negates your entire wall of text. And if you say 970 doesn't compete with 290X, I'll show you where you might be mistaken.
AMD doesnt appear to have any answers til 2015. 1H anyway.

 

[rusina]

I agree with you. NVidia doesn't have a good area advantage so the costs should be more or less the same for both parties. What Nvidia is trying to do is to get a premium for its higher efficiency. I expect Nvidia revenue share to grow in the next two quarters.

That said, I don't expect AMD market share to fall 50%, I was just pointing out to RussianSensation that AMD has to manage its cash inflows very carefuly, because they are in a very tight situation on the matter. AMD must carefuly calibrate how much revenue volume they can afford to lose to Nvidia otherwise they will burn too deep into their thin cash reserves.

If I remember correctly AMD pays only for working GPUs while Nvidia pays for all GPUs. Nvidia pays less per GPU. Which system is better depends of how many of those units are working. This explains why we so often see these Nvidia-cards with crippled GPUs..specially in OEM cards.

Smaller power consumption means cheaper power components. Smaller memory bus means simpler pcb.

I think Nvidua is good with profit margins and they could have sold these cards as 960-series if competition would have demanded it. They made profit with 560 ti-cards and these GM204 cards shouldn't be that much more expensive to make

 

[mrmt]

If I remember correctly AMD pays only for working GPUs while Nvidia pays for all GPUs. Nvidia pays less per GPU. Which system is better depends of how many of those units are working. This explains why we so often see these Nvidia-cards with crippled GPUs..specially in OEM cards.

Nope, it was quite the opposite. Nvidia had a waiver to pay only for good dies for a brief monent (can't recall whether it was on the 40nm or in the 28nm node introduction). I couldn't find any data about Nvidia using more fused-off parts than AMD, would you be able to share some?