Radeon R9 290X Priced at $730 at Newegg

[Silverforce11]

It's faster than the 780, it has to be priced higher than the 780.

If thats the criteria, every new generation should cost much more than the last, right?

Except no, thats not how it works. There's a price range many people are willing to folk out, and there's a range where few would. Titan and 780 are harvested GK110 dies that didn't make it into Teslas and servers. R 290 dies first and foremost are consumer products. There's a difference.

AMD is making a big mistake if they think its more beneficial to them to price it around $700. Price it around $500-550 and they would sell much much more.

 

[King Mustard]

Excellent. Looks like AMD is about to cornhole nvidia with a Titan killer.

I should hope so given the time, the Titan was released in February - that was 8 months ago...

 

[3DVagabond]

If thats the criteria, every new generation should cost much more than the last, right?

Except no, thats not how it works. There's a price range many people are willing to folk out, and there's a range where few would. Titan and 780 are harvested GK110 dies that didn't make it into Teslas and servers. R 290 dies first and foremost are consumer products. There's a difference.

AMD is making a big mistake if they think its more beneficial to them to price it around $700. Price it around $500-550 and they would sell much much more.

I haven't read anymore about it, but suppose it's 1/2 DP as it's been rumored to be? What's it worth then? That's what supposedly makes Titan worth more because it does 1/3 DP. What if Hawaii beats Titan in DP?

For me it doesn't mean Jack. Just like Titan's DP being 1/3 means absolutely zero to me. There are those who have used it to justify Titan's price still til today. I only ask you because you referenced Titan being a Tesla chip.

 

[Silverforce11]

I haven't read anymore about it, but suppose it's 1/2 DP as it's been rumored to be? What's it worth then? That's what supposedly makes Titan worth more because it does 1/3 DP. What if Hawaii beats Titan in DP?

For me it doesn't mean Jack. Just like Titan's DP being 1/3 means absolutely zero to me. There are those who have used it to justify Titan's price still til today. I only ask you because you referenced Titan being a Tesla chip.

Its worth something because CUDA is widely used where it matters in the sectors that think $1000 for a GPU is cheap. AMD doesn't cut the mustard for those sectors yet, so it doesn't matter until they do penetrate better.

Also, Hawaii being a consumer product first and foremost, means most of the dies are going towards R 290X, with a relatively small die size on a mature node, yields should be high, as such, its not a limited product. They can easily match massive demands on it if they wanted to. By pricing it reasonable, they can ensure there WILL be massive demand. Pricing it at $700, its neutering demand by many folds, as any market analysis will tell you, $500 segment is already small, $700 or beyond would be minuscule.

Why artificially reduce demand on a product that has the potential to sell like crazy??

Edit: The ONLY justifiable reason for them to mark up massively, ie. selling much fewer units but with more margins and end up making LESS overall profit, is if the dies are limited in quantity, they are banking on the fact any that is made, will be sold regardless of whether its $500 or $700. If there's another non-valid, non-economical reason why its priced so high (ie. "HALO" brand building etc), then heads will roll after this debacle is through.

 

[3DVagabond]

I would love to see the % of Titans that were purchased to run CUDA in applications that leverage DP. I'd wager it's next to 0%. Titan doesn't have ECC RAM and has zero support from nVidia for professional use, being a Geforce card. Nobody who has Titans on these boards has ever said they use them for any type of DP application. Everyone who purchased them here, and any other forum I've gone to, uses them for gaming.

 

[RussianSensation]

So Intel's 4670K is 30% faster than the 3570K and also $30+ dollars cheaper?

This comparison is not as relevant for these reasons:

1) Intel CPUs keep their resale value which means if you buy a $225-325 Intel CPU, you can recoup 60-70% of its value after selling in 2 years. Now look at GTX580 that came out at $499 and 2.5 years later HD7870 is $170 with the same performance, while HD7970 fell from $549 to $280 for 1Ghz versions. If you happen to have a MC near you, then you can buy Intel CPUs, resell them, and lose almost nothing, making Intel CPU ownership extremely cheap.

2) If you do not sell the CPU, i5/i7 overclocked now easily lasts 4-5 years. i7 920 @ 4.0-4.2ghz is still a good gaming CPU. The total cost of ownership per year assuming a resale value of $0 is very low since modern Intel CPUs have a much longer useful life and much lower cost than modern GPUs ($225-325 vs. $650-700 for the GPU!)

3) The nature of the GPU vs. CPU industry. GPUs increase in speed anywhere from 50-100% every 2-3 years, while CPUs do not. Since the GPU cycles are much shorter for major performance increases, 6 months late is a long time in the GPU industry but it is not in the CPU industry where performance increases are very small.

4) Opportunity cost of waiting. If you buy i5/i7, you have an idea of Intel's road-map and it's not as if in 6-9 months Haswell refresh will be 30% faster at $225-325. Now look at the opportunity cost of waiting 6 months and not buying say EVGA GTX780 ACX? It makes no sense to have waited for R9 290X. R9 290X barely beating 780 6 months later for $649 is not an accomplishment, and even worse when it's $650 for a reference cooler. Worse, it's not really putting much downward pricing pressure on NV's 780 since after-market 780s are $650-700 and are 15-20% faster than a reference 780 (Galaxy HOF 780, EVGA Classy).

5) GPU technology curve. As mentioned, the closer we get to 20nm GPUs, the less valuable 28nm GPUs become because we are in the 2nd half of 28nm node generation. Would you have said buying a GTX580 for $450 near the 2 year mark of 40nm generation (480->580) before 28nm 680 launched was worth it? Of course not since you know on the technology curve the 580 was old tech, but the market price has not adjusted. That's what you call market pricing inefficiency. If consumers are paying $450 for a near 2 year old 40nm 580 around December 2011, then they are not knowledgeable of tech cycles. Now imo, $650 for a 28nm card that's only 30-35% faster than HD7970Ghz is way too expensive since we are 2 years into "old" 28nm tech. 28nm tech should be more mature which means a 424mm2 Hawaii should not be 2.3x the cost of a 1Ghz 7970 if it only brings 30-35% more performance.

^ In comparison, HD7970 was $549 and overclocked, it beat HD6970 overclocked by 70-75% when 6970 cost $299. R9 290X offers half of the performance increase 7970 oc offered over 6970 and costs $100 more.

 

[lopri]

$730.. ?!

Edit: forgot to say "LOL." ^^

 

[SiliconWars]

If thats the criteria, every new generation should cost much more than the last, right?

Except no, thats not how it works. There's a price range many people are willing to folk out, and there's a range where few would. Titan and 780 are harvested GK110 dies that didn't make it into Teslas and servers. R 290 dies first and foremost are consumer products. There's a difference.

AMD is making a big mistake if they think its more beneficial to them to price it around $700. Price it around $500-550 and they would sell much much more.

There are a bunch of reasons why AMD would want to price it ~$699.

The number of people buying $500 cards isn't that much higher than the number of people buying $700 cards. We're talking tens of thousands, not millions.

They need to price it where the competition is now. Pricing cheaper than a slower card sends out the wrong message. This thing is a couple percent short of Titan making it a good 10%-ish faster than the 780.

Every single spare wafer you can imagine is going on consoles right now. The demand for them is unprecedented and sales are guaranteed. Don't forget that Hawaii is basically taking away wafers from that as it's not a replacement it's a brand new segment of cards.

It all adds up to a limited wafer run and high price cards. I'm sure they'll be under $500 in the new year but right now the sensible option is for AMD to create real desire for the product and price it where it performs.

 

[SirPauly]

Pricing cheaper than a slower card sends out the wrong message.

Didn't effect nVidia's message when they offered the faster GTX 680 cheaper than the slower HD 7970!

 

[ICDP]

Didn't effect nVidia's message when they offered the faster GTX 680 cheaper than the slower HD 7970!

Using that logic, if R9 290X matches Titan and is ~$350 cheaper, then surely it should be judged as an amazing deal? Or if R9 290X comes in $50 cheaper than GTX 780 and is less than 10% faster, will it be hailed as a great product in the same way GTX680 was?

 

[SiliconWars]

There are many variables at work. Yes the 680 was faster and cheaper back then but it was also much cheaper to produce having a smaller die and less RAM. The 290X is smaller than the 780 but if you factor in that the 780 is twice salvaged the actual yield rate might not be all that much different between them. The 290X also has more RAM again.

 

[HisDivineOrder]

Hypothetically, they could also be pricing (extremely) high to make room for price cuts in due course to the wait for 20nm next year.

That said, I think it's still too high for a late 28nm fab'ed part that either only matches or barely beats what we've had for months at roughly the same price point. If you're going to wait this long, you have to bring something new to the table. New performance, new price, some new feature that's actually worthwhile.

I think that's what the reveal was all about. Trying to hit that last one, but I don't think they actually found one that really means much right now, which is why everyone's pointing at the price and performance to look for it.

 

[Z15CAM]

None of us would be here if we where not Geeks, Nerds - Whatever - Other then Z15CAM, I've more then one OnLine Identity.

Apparently due to the low % of "Whoever" our words mean nothing in this INSANE $ Monopoly - Yet Google and various Search Engine Marketing are dependent as to what we have to say.

Consider Price/Performance is ultimately our domain we should control it.

So I say SHOOT the product down and buy with Discretion; in other words - Use our Back Door and fight back saying the GTX and R9 are nothing but Marketing Bullshit that will not work.

 

[SirPauly]

Using that logic, if R9 290X matches Titan and is ~$350 cheaper, then surely it should be judged as an amazing deal? Or if R9 290X comes in $50 cheaper than GTX 780 and is less than 10% faster, will it be hailed as a great product in the same way GTX680 was?

I'll allow third party sites to review and the market to decide judgement over-all!

 

[ICDP]

There are many variables at work. Yes the 680 was faster and cheaper back then but it was also much cheaper to produce having a smaller die and less RAM. The 290X is smaller than the 780 but if you factor in that the 780 is twice salvaged the actual yield rate might not be all that much different between them. The 290X also has more RAM again.

HD 7970 had an extra 1GB of VRAM over GTX680 and was only marginally slower. At GTX 680s release HD 7970 OC cards were available at the same price with custom cooling and 1GHz clock speeds. Yet somehow none of that mattered.

R9 290X is going to have 1GB extra VRAM and is allegedly faster stock for stock vs GTX 780. Somewhow I have a feeling that the usual suspects will be along after release to declare it a failure because it is too little too late. The same people who procclaimed GTX770 and GTX760 as amazing, despite being almost one year late and only matching 7970 GE and 7950 OC cards.

 

[Rvenger]

Didn't effect nVidia's message when they offered the faster GTX 680 cheaper than the slower HD 7970!

Oh jeez... Enough of this already. Just when I thought things are civil then this stuff starts to get spewed to stir the pot.

 

[ICDP]

Oh jeez... Enough of this already. Just when I thought things are civil then this stuff starts to get spewed to stir the pot.

Exactly. Closely followed by the usual passive aggressive, "let the market decide" when called out.

 

[SirPauly]

It was more than civil and engaged this point:

Pricing cheaper than a slower card sends out the wrong message.

Pricing cheaper than a slower card doesn't send out the wrong message and used an example!

 

[SirPauly]

Exactly. Closely followed by the usual passive aggressive, "let the market decide" when called out.

No, simply allowing third party sites to review and the market to decide judgement over-all seems fair and actually reasonable!

 

[ShintaiDK]

Somebody has to pay for the "free" game deals and sponsoring in the multi million dollar amount for games. Hence graphics cards goes against the overall trend in price for electronics.

 

[Vesku]

This comparison is not as relevant for these reasons:

1) Intel CPUs keep their resale value which means if you buy a $225-325 Intel CPU, you can recoup 60-70% of its value after selling in 2 years. Now look at GTX580 that came out at $499 and 2.5 years later HD7870 is $170 with the same performance, while HD7970 fell from $549 to $280 for 1Ghz versions. If you happen to have a MC near you, then you can buy Intel CPUs, resell them, and lose almost nothing, making Intel CPU ownership extremely cheap.

AMD may have problems with brand value (steeper decline between launch price and pricing 6 months later , resale value) but Nvidia is much closer to Intel in that regard.

2) If you do not sell the CPU, i5/i7 overclocked now easily lasts 4-5 years. i7 920 @ 4.0-4.2ghz is still a good gaming CPU. The total cost of ownership per year assuming a resale value of $0 is very low since modern Intel CPUs have a much longer useful life and much lower cost than modern GPUs ($225-325 vs. $650-700 for the GPU!)

GPUs are lasting longer too. 8800 had a long life. Same with 4870 and 5870. 580 is still hanging in there at 1080P. Nvidia and AMD can probably thank Sony and Microsoft: Xbox 360 and PS3 being so old made the GPU busting AAA game titles rarer.

3) The nature of the GPU vs. CPU industry. GPUs increase in speed anywhere from 50-100% every 2-3 years, while CPUs do not. Since the GPU cycles are much shorter for major performance increases, 6 months late is a long time in the GPU industry but it is not in the CPU industry where performance increases are very small.

Both are dependent on moving to a smaller node to see appreciable performance increases. Unless they mess up a design: Pentium 4, GeForce FX, Radeon 2900.

4) Opportunity cost of waiting. If you buy i5/i7, you have an idea of Intel's road-map and it's not as if in 6-9 months Haswell refresh will be 30% faster at $225-325. Now look at the opportunity cost of waiting 6 months and not buying say EVGA GTX780 ACX? It makes no sense to have waited for R9 290X. R9 290X barely beating 780 6 months later for $649 is not an accomplishment, and even worse when it's $650 for a reference cooler. Worse, it's not really putting much downward pricing pressure on NV's 780 since after-market 780s are $650-700 and are 15-20% faster than a reference 780 (Galaxy HOF 780, EVGA Classy).

Again this is just another aspect of node lifespans increasing. Nvidia had the big die 28nm already being produced for their HPC products. The reason to wait is if you think a node change is coming. Some people were hoping for an early 2014 20nm where as it's now looking like closer to end of year 2014.

5) GPU technology curve. As mentioned, the closer we get to 20nm GPUs, the less valuable 28nm GPUs become because we are in the 2nd half of 28nm node generation. Would you have said buying a GTX580 for $450 near the 2 year mark of 40nm generation (480->580) before 28nm 680 launched was worth it? Of course not since you know on the technology curve the 580 was old tech, but the market price has not adjusted. That's what you call market pricing inefficiency. If consumers are paying $450 for a near 2 year old 40nm 580 around December 2011, then they are not knowledgeable of tech cycles. Now imo, $650 for a 28nm card that's only 30-35% faster than HD7970Ghz is way too expensive since we are 2 years into "old" 28nm tech. 28nm tech should be more mature which means a 424mm2 Hawaii should not be 2.3x the cost of a 1Ghz 7970 if it only brings 30-35% more performance.

^ In comparison, HD7970 was $549 and overclocked, it beat HD6970 overclocked by 70-75% when 6970 cost $299. R9 290X offers half of the performance increase 7970 oc offered over 6970 and costs $100 more.

TSMC 20nm is still probably 9-12 months away. Haswell is on the same 22nm node as Ivybridge with no desktop node change scheduled for next year (14nm Broadwell roadmapped as mobile only). And that's Intel who is the industry leader in terms of Fabrication.

Generally consumers get more transistors/$ with each node change but companies don't necessarily slash the pricing structure. Each product tier just has more transistors than the products made on the previous node. Main factor in all of this is that node changes are slowing down.

 

[ICDP]

No, simply allowing third party sites to review and the market to decide judgement over-all seems fair and actually reasonable!

You took one single line from his post and took it out of context. As Rvenger correctly stated, you are trying to stir the pot.

GTX 680 was hailed as a Tahiti killer right from release. Very few of the reviews or Nvidia fans waited for the "market to decide", they simply hyped GTX680 then and there. I asked if AMD did the same thing and release R9 290X $50 cheaper and ~10% faster than 780, should we use the same criteria and hail it as a great card?

No need to wait for the market to decide, it's a simple question than can be answered easily.

For the record my response would be, no we shouldn't. GTX 780 has been available with that level of price and performance since May. IMHO R9 290 should be ~$450, R9 290X should be released at ~$550, the same price the 7970 was released at.

Somehow many think it is OK for Nvidia to be late and deliver similar performance as an already available AMD card. In fact is more more than likely the Nvidia card will be hyped as amazing (See the truly ordinary GTX 660Ti for proof). Turn the tables and AMD are "too little too late".

 

[SirPauly]

No need to wait for the market to decide, it's a simple question than can be answered easily

It's not a simple question that can be answered quite easily for over-all!

 

[ICDP]

It's not a simple question that can be answered quite easily for over-all!

So you are unable form an opinion without waiting for the masses to tell you how to think?

Ohhhh Kaaay

 

[RussianSensation]

It's not a simple question that can be answered quite easily for over-all!

ICDP has a point there.

660Ti (and even 760 can be clumped here) - brought almost nothing worth talking about vs. 7950 V2. In the end an overclocked 7950 either matched or beat those cards and at similar or lower prices. 760 is now destroyed by barely more expensive 1Ghz 7970s making it pointless.

770 - 1 year late, horrendously overpriced, one of the worst releases by NV in years. Yet, it wasn't ripped apart on the forums and still isn't. It's ludicrous for a 2-4GB 770s to cost $380-$450 when a 1Ghz 7970 is $280-290 and R8 280X is $299.

That's his point exactly, 660Ti/760/770 are all worse than R9 290X. Having said that, I still think by now a more reasonable price is $550 because 780 isn't exactly new.