Pricing Discussions around AMD VEGA *over-flow*

[Deleted member 13524]

Digi those bean counters you mention are sometimes too stupid to see the big picture and usually only care about short term income, but they're rarely irrational.


Something is making them bullish to try to increase the price. Be it for mining or for gaming, something seems to be bound to make Vega 10 more attractive to someone.

The bean counters are not stupid enough to try to sell Vega 10 at over $600 pre-tax in its current state.

 

[silent_guy]

except we aren't past the first few days of availability so we have pitch forks out for no reason. The 1080 FEs were all over priced for months after release. The majority of 1070s are over priced compared to launch price

Did Nvidia use rebates to be able to claim to reviewers that the MSRP was the MSRP?

All I have seen is people talking about rebates but everything else is speculation.

Retailers selling above MSRP is common and out of control of AMD. Using rebates to influence price/$ stats in reviews (why else?) is the point where "unethical business practices" enter the equation.
I can't imagine that reviewers will be looking favorably on this in the future.

 

[Rootax]

Already some reviewers are pissed. I hope AMD will learn from this (not only the pricing thing, but the whole launch).

 

[eastmen]

Yeah.... That must be it.

Right but it must be that AMD is using rebates. Whats more its that AMD is using rebates and didn't put in some legal stipulation that the rebates were not allowed to be discussed. Because AMD is this dumb. But no lets just attack AMD with just the websites that are making a profits word

No comparable. There were lots of Vega available in bundle SKUs still.

There were bundles of switch available on launch day also for awhile. The only thing the bundle sku's prove is that the bundle is to high. Which means the bundles will drop in price as they continue to sit there and it will put downward pricing pressure on the unbundled sku

Did Nvidia use rebates to be able to claim to reviewers that the MSRP was the MSRP?


Retailers selling above MSRP is common and out of control of AMD. Using rebates to influence price/$ stats in reviews (why else?) is the point where "unethical business practices" enter the equation.
I can't imagine that reviewers will be looking favorably on this in the future.

I don't know did AMD ? The whole Rebate is just a hit piece at this point. When more 64 unbundled cards hit the market will all of you guys say "See horrible AMD is at it again putting rebates on a few more cards" when the third wave hits will you do it again ? and again which each new wave ?

It seems to me that the unbundled SKU will always be used negatively against AMD from this point on. It will either be well they are still rebating , oh the quanities are too low so we were right they can't make them cheap enough or look you still can't get them at MSRP ! they must have been rebating them. All the while nvidia cards will sit with a mark up on the market and no one will say a peep.

 

[tongue_of_colicab]

I've only been casually reading all the topics but the whole Vega launch seems like one big cluster fuck to me.

First the weird FE launch. Then that strange tour which served to do what? Then giving reviewers only a couple of days to do their reviews (which I cannot see any reason for other then AMD knowing it wouldn't be all that pretty) and finally the pricing debacle which is essentially a 100 bucks price hike. I can only imagine AMD did this so reviewers would quote a lower price than what the majority of buyers will be able to buy the card for. I mean usually early adapters are willing to pay more. I can't really think of any electronics that become more expensive after the launch (excluding scalpers).

I don't see what AMD has to gain from all this.

 

[eastmen]

I wish they'd make the bundle better for gamers. Include some more monitors with a lesser rebate, but at least for some of the more affordable freesync monitors, and mebbe go back to the old "offer a selection of games and you get to choose 3" or something with some other titles tossed in.

I also think they'd be best off taking the hit on the next shipment and offering a whole lot less bundles and a lot more individual cards, and I'm not sure if the mining lust for this card is gone or not but as much as I know Raja loathes the idea I think they need to plant some miner bombs in the drivers for at least a while to discourage miners from sucking 'em all down.

I get his position on it and I actually respect it, he genuinely doesn't want to ever intentionally cripple performance or limit the openness of their system. I on the other hand am not quite so inflexible and think that it would do them a WHOLE LOT of long term good with the community/gamers if they would just do it. I'd even recommend they be open about it and their reasons why, that'd give Raja some comfort with the whole openness thing and it could REALLY solve their problems of how to make this card available for gamers.

That'd be my advice, but I don't think anyone anywhere really knows what is going to happen yet so I also highly agree with the holding off of the pitchforks and torches for a bit longer until we get a bit more clarification. I'm not sure I want them to just make a quick call on this, I'd prefer they really think about what they are trying to achieve in the gaming market place and how their actions in the immediate future are going to drastically affect the public's perception of them and the Vega launch versus the hit they'll take financially for it. I don't know the numbers and I suck with finances, but I do know gamers and the gaming community and there are a whole lot of rumors/half-truths/crazy shit going on and there are a lot of people who are viewing this as just how genuine AMD's commitment to gamers is.

Their last few launches have been terribly problematic, and I personally blame most of it on miners. I don't get mining, it seems like the whole stock market bubble to me before everything crashed. I kept wondering how they were selling all these packaged junk bonds that were obviously not worth what they were sold, it felt like a weird get rich quick scheme or money for free thing...and I always remember the quote, "If it sounds too good to be true, it probably is", and figured it'd end badly.

I was shocked when I found out I hadn't misunderstood it at all, there were just a lot of greedy slimeballs out there and a lot of people who didn't understand it and fell for the scam. Since then I've decided to sort of trust my instincts on weird stuff like that, and mining just seems like people are making monies with nothing really backing it but crunching numbers fast and I think it's gonna fall apart when people realize that...but I could very, VERY easily be wrong on that as I don't know very much about mining except that I HATE HOW IT MAKES IT SO HARD TO GET THE BLOODY VIDEO CARD I WANT TO GAME ON!!!!

Sorry, it's late and I'm a bit medicated..but that's how I'm feeling on this one. They're doing great with CPUs, but no matter how good they are or what a great value they are AMD ain't gonna get back much graphics market share unless they can get the cards in to the hands of gamers. I think the only way to do that is to declare open war with miners on a driver level, because I just don't see any other way to do it. Price won't stop them, bundles won't stop them; they want the best and they'll get it.

Just make it not the best for that, the problem will take care of itself. No one in the community will fault you, heck they'll be screaming your praises to the gods of gaming themselves for finally doing something substantial to face the problem.

I know you don't want to AMD, but unless you can come up with a better idea to take the miners out of the equation the prices are going to be moot. The discounts on early cards are going to just be advantages to the miners, they always seem to find a way to get them first.

It would also cut down on the incentive for retailers to mark up prices. If they have to actually put the cards on the shelves and have people buy them it might make them reconsider jacking up the price 50-80%. That would be nice too, wouldn't it?

Please AMD, do the something good. You're on such a roll, the Ryzen kicks so much ass and everyone still loves it..but to be so close and blow it over miners?

Please, don't.

I don't get the mining spin your putting . Vega is only getting about 40 MH/s at 200+ watts. The 1070 can do in the low 30s at 100w . So yes vega cards are faster but also require more power. So unless miners can get them cheaper than the nvidia cards its not worth it. Not only cheaper but enough to make the diffrence in power back since that will continue to effect profits through the life of the card vs the initial cost of the card being recouped.

The only bright spot is vega 56 because at $400 its cheaper and faster than the 1070 mostly by large enough performance deltas to make up for any extra power usage . The 64 is slower than a 1080 and while at its MSRP has a slight cost advantage its power usage is way off the charts .


These cards are going to drop quickly in price and stick at MSRP much faster than previous cards . The 4x0 and 5x0 series will be more popular for mining along with the geforce 1xx0 line up

 

[Rootax]

They gained not so bad reviews, thinking the rx vega was priced again 1080. I guess.

 

[Entropy]

Digi those bean counters you mention are sometimes too stupid to see the big picture and usually only care about short term income, but they're rarely irrational.


Something is making them bullish to try to increase the price. Be it for mining or for gaming, something seems to be bound to make Vega 10 more attractive to someone.

The bean counters are not stupid enough to try to sell Vega 10 at over $600 pre-tax in its current state.

Unless AMD has manufacturing issues that make them unable to fill any kind of demand, so they kill two birds with one rock by pricing the cards high. The price could then drop to the claimed msrp when there is supply.
Can't say I find this hypothesis likely but then again no other hypothesis makes a lot of sense, and there has to be some reason for Vega to launch so late.

 

[troyan]

Maybe AMD doesnt want to sell Vega to gamer. There is a $300 difference between the Vega 56 and GTX1080TI. The margins must be really thin for everyone involved in the chain...

 

[Entropy]

Maybe AMD doesnt want to sell Vega to gamer. There is a $300 difference between the Vega 56 and GTX1080TI. The margins must be really thin for everyone involved in the chain...

Judging by nVidias quarterly reports, the margins are juicy and dripping with fat.

 

[troyan]

When you can sell a card with a 480mm^2 die for $699+ they are very healthy.

But what will happen when you have to cut $300 from the price?

 

[Entropy]

When you can sell a card with a 480mm^2 die for $699+ they are very healthy.

But what will happen when you have to cut $300 from the price?

It won't be nearly as profitable, obviously.
There are a number of details we just don't know. (Vega should be cheaper than Fiji to produce though.) Also, depending on corporate strategy and internal accounting, profitability of Vega 10 may not be a major concern as it is the first implementation of a design that will get used both in future discrete GPUs as well as in APUs that may target much greater sales volumes. It serves both to fulfill their WSA agreement with GloFo, and as an architectural trailblazer.
(Also, the Vega 56 is to some extent a salvage product, increasing product yield, so it makes more sense to use Vega 64 as the base line for hypothesizing.)

At the end of the day, if a product isn't attractive at at a given pricepoint it won't sell, recouping none of its design costs, and yielding little to no impact on the market/industry.

 

[manux]

It won't be nearly as profitable, obviously.
There are a number of details we just don't know. (Vega should be cheaper than Fiji to produce though.) Also, depending on corporate strategy and internal accounting, profitability of Vega 10 may not be a major concern as it is the first implementation of a design that will get used both in future discrete GPUs as well as in APUs that may target much greater sales volumes. It serves both to fulfill their WSA agreement with GloFo, and as an architectural trailblazer.
(Also, the Vega 56 is to some extent a salvage product, increasing product yield, so it makes more sense to use Vega 64 as the base line for hypothesizing.)

At the end of the day, if a product isn't attractive at at a given pricepoint it won't sell, recouping none of its design costs, and yielding little to no impact on the market/industry.

Are you sure vega is cheaper than fiji?

It's not just the size of the chip but also more expensive manufacturing process, the heavy circuitry needed for power input, cooling capable of dissipating 500W+ and the extra expensive to buy(not necessarily to manufacture) HBM2. All the time when competing in price racket where competitor has significantly smaller chip, bulk memory and lesser cooling/power requirements. Also the r&d cost of the chip needs to be amortized. Selling at +-0 would mean r&d cost is never recovered and there is more pressure in future r&d. Dust particle on 28nm chip and 14nm chip causes very different kind of error, 14nm is more expensive per transistor and there is plenty more transistors in vega compared to fiji.

[SK Hynix: Customers Willing to Pay 2.5 Times More for HBM2 Memory]

http://www.anandtech.com/show/11690/sk-hynix-customers-willing-to-pay-more-for-hbm2-memory

 

[Deleted member 2197]

Since Vega true price less retailer voucher is $499, the breakeven cost should not be far from this point if AMD is unable to profitably maintain $499 as the long-term MSRP.

"£449 is not possible, $499 is below what they cost us direct from the board partners by a large chunk of cash, AMD rebated us to hit $499 on a set amount of units.

https://forum.beyond3d.com/posts/1996665/

 

[mrcorbo]

It seems to me that the unbundled SKU will always be used negatively against AMD from this point on. It will either be well they are still rebating , oh the quanities are too low so we were right they can't make them cheap enough or look you still can't get them at MSRP ! they must have been rebating them. All the while nvidia cards will sit with a mark up on the market and no one will say a peep.

And this proves that you are either completely clueless about this or deliberately missing the point. Unbundled cards being available at the MSRP would be great, however they get there. It would solve the problem. They should continue to issue rebates as needed to insure a continuing supply of them instead of just doing it for a few hundred cards for a few retailers during the launch. Otherwise, the only difference between their marketing being misleading and an outright lie is a semantic one.

 

[DavidGraham]

The only bright spot is vega 56 because at $400 its cheaper and faster than the 1070

Nope, it is not faster, it trades blows with the 1070, depending on the title. It may be generally faster in DX12 titles, but it lags behind in DX11 titles.
https://www.techspot.com/review/1468-amd-radeon-rx-vega-56/page8.html

Several reviews have the 56 faster than 1070 indeed, but that's because of their choice of games. It's a really annoying trend during this launch. Several press outlets populated their charts with old DX12 titles (like Ashes and Warhammer), while ignoring some of the best and popular DX11 ones. One of the notable examples is hardware.fr who populated their Vega 64 review with AMD friendly DX12 titles, ignoring every DX11 titles as if the market is only DX12 games. Anandtech and Toms did the same to some degree (their lineup was mostly AMD friendly games). Despite these measures Vega 64 lost to the 1080 or came to a tie. Once you diversify your lineup (like TechSpot did), the correct picture emerges. 56 trades blows with 1070, Vega 64 is slower than 1080, Liquid Vega 64 trades blows with the 1080.

It's time for these sites to upgrade their charts with modern games: Prey, Mass Effect Andromeda, Hellblade, Battlegrounds, Watchdogs 2, Titanfall 2, For Honor, Ghost Recon Wildlands, Rainbow Siege .. etc. Sites that diversify their line up include PCGamesHardware and ComputerBase. That's why these sites remain the defacto for game benchmarks.

 

[mrcorbo]

It's time for these sites to upgrade their charts with modern games: Prey, Mass Effect Andromeda, Hellblade, Battlegrounds, Watchdogs 2, Titanfall 2, For Honor, Ghost Recon Wildlands, Rainbow Siege .. etc. Sites that diversify their line up include PCGamesHardware and ComputerBase. That's why these sites remain the defacto for game benchmarks.

Your "more representative" list contains several games from a single publisher who does notoriously unoptimized PC ports. Swap out 3 of those Ubisoft games for Rise of the Tomb Raider, Doom and Hitman and swap out ME:A for a Battlefield game and you might have something.

 

[Alexko]

Are you sure vega is cheaper than fiji?

It's not just the size of the chip but also more expensive manufacturing process, the heavy circuitry needed for power input, cooling capable of dissipating 500W+ and the extra expensive to buy(not necessarily to manufacture) HBM2. All the time when competing in price racket where competitor has significantly smaller chip, bulk memory and lesser cooling/power requirements. Also the r&d cost of the chip needs to be amortized. Selling at +-0 would mean r&d cost is never recovered and there is more pressure in future r&d. Dust particle on 28nm chip and 14nm chip causes very different kind of error, 14nm is more expensive per transistor and there is plenty more transistors in vega compared to fiji.

http://www.anandtech.com/show/11690/sk-hynix-customers-willing-to-pay-more-for-hbm2-memory

No idea what the respective costs really are, but Vega is smaller, uses half the number of memory stacks, and a smaller interposer. Now 28nm wafers are certainly cheaper than 14nm ones, but the respective costs at launch might be a different story, as Fiji was two years ago—and yes, in two years, AMD has not increased peak FP32 FLOPS/cycle at all.

 

[Deleted member 13524]

No idea what the respective costs really are, but Vega is smaller, uses half the number of memory stacks, and a smaller interposer. Now 28nm wafers are certainly cheaper than 14nm ones, but the respective costs at launch might be a different story, as Fiji was two years ago—and yes, in two years, AMD has not increased peak FP32 FLOPS/cycle at all.

Also Fiji was made at TSMC whereas Vega is made at Globalfoundries, which should be charging substantially less per state-of-the-art wafer.

 

[Entropy]

Are you sure vega is cheaper than fiji?

It's not just the size of the chip but also more expensive manufacturing process, the heavy circuitry needed for power input, cooling capable of dissipating 500W+ and the extra expensive to buy(not necessarily to manufacture) HBM2. All the time when competing in price racket where competitor has significantly smaller chip, bulk memory and lesser cooling/power requirements. Also the r&d cost of the chip needs to be amortized. Selling at +-0 would mean r&d cost is never recovered and there is more pressure in future r&d. Dust particle on 28nm chip and 14nm chip causes very different kind of error, 14nm is more expensive per transistor and there is plenty more transistors in vega compared to fiji.
http://www.anandtech.com/show/11690/sk-hynix-customers-willing-to-pay-more-for-hbm2-memory

Well, let me put it this way - the marginal cost for the two million and first VEGA 64 should be a bit lower than for the Fury X. Smaller die (wafer cost may be a tad higher, but then again the chip is smaller, and AMD need to purchase a certain number of wafers under any circumstances, so...?), smaller and simpler interposer, two stacks of memory, simplified cooling. The differences shouldn't be major though.
(Transistor cost of 14nm is absolutely not higher than for 28nm. Such a statement carries a lot of assumptions which simply aren't true for the devices typically manufactured. Check wafer costs instead. Even those are a bit tricky though because they are not necessarily easily accessible in the public domain for all foundries, and who knows what pricing AMD has with GloFo, never mind the intricacies of WSA. You can't use very simplistic reasoning when it comes to yields. Have some faith in AMD engineers, they design for manufacturability.)

True Cost is a hopelessly complex question. Depends a lot on volume though since the sunk costs are high. Shouldn't much of the R&D for Vega be spread over all GPUs and APUs using the architecture for instance? Let's leave that to the bean counters. That's what they do themselves, and focus on BOM even though everyone knows that doesn't paint the whole picture. And BOM should be way lower than $500!)

What can be said with some certainty though is that regardless how you choose to look at it - recouping sunk costs, impacting the industry in terms of target architectures to take into account, mindshare in their major target market, gaming - achieving a good sales volume is vital. There is always an initial buying frenzy as well as difficulties getting all the geese in a row in the supply line. Once both of these are satisfied though, we'll have a better sense of where VEGA fits price wise in the market place.