[Possibly fake] AMD losing 100+ dollars on every vega sold

[Coaxialgamer]

Quote

Our industry sources have confirmed to Fudzilla that AMD loses at least $100 on every Vega 64 card it sells at its $499 Suggested Etail Price (SEP).

The pricing of the HBM 2.0 memory, the packaging and substrate cost are simply too high to have a sustainable price of $499

Quote

Vega 64 and 56 will definitely put a dent in the Nvidia dominated higher end GPU market. There are people who are willing to buy AMD, no matter what. Frankly the performance of Vega is enough to get a lot of people excited. The only downside of the Vega architecture is that the TDP power is too high, compared to the Geforce GTX 1070/1080 competition. Despite that, the performance and price ratio are quite balanced and are gaining a lot of sales for AMD.

The real manufacturing price or BOM (Bill of Materials) price of Vega is a well kept secret. The Vega GPU is being manufactured by GlobalFoundries (GloFo) and AMD has a sweetheart deal with this chip fab. It even has a five year wafer supply agreement with GlobalFoundries.

This is where AMD saves some cost, but it currently cannot really do much about the high HBM 2 memory prices. So when AMD lets its Etailers sell Vega at the higher prices than SEP, it is actually making some money.

The pricing leaves a bitter taste as traditionally companies are very strict in controlling that no one really goes over the board with Suggested Etail Prices

http://fudzilla.com/news/graphics/44401-amd-is-losing-100-on-every-vega (source)

 

This is bad . Very bad . Worse than i imagined t, given pre-launch information .

The chip is large , power hungry , expensive to manufacture and performance is largely underwhelming ; competing against a gpu 65% it's size.

I'm going to get hate for saying this , but i dare say that this is worse than Fermi .

 

Edit:Also this

Quote

So when AMD lets its Etailers sell Vega at the higher prices than SEP, it is actually making some money.

  Amd sells all their vega cards at the same price , regardless of retail price. Amd isn't making money on vega price gouging. Retailers and re-sellers are.

 

Edit: Raja Koduri has recently made a tweet suggesting that the bill of materials provided by these "industry sources" might be incorrect. 

 

[Bouzoo]

This shouldn't matter in the end because I've yet to see a Vega card that is selling below $600 (or $649 for that matter), and that's in the US. God forbid you look on other continents. You can bet AMD is not losing money. 

Someone will correct me if the unicorn exists. 

[PCGuy_5960]

2 minutes ago, Coaxialgamer said:

Our industry sources have confirmed to Fudzilla that AMD loses at least $100 on every Vega 64 card it sells at its $499 Suggested Etail Price (SEP).

So this explains why you can't find Vega for $500

[DocSwag]

It's known that HBM2+interposer is around $175; VRMs are probably around $30-$40 ish and say PCB is $20 or something.

 

That would indicate that, according to this estimate, that the die itself costs around $350!!! Isn't that a bit high?

 

I feel like this estimate is pretty far off. I don't think it's likely that the die itself costs that much...

[Coaxialgamer]

11 hours ago, DocSwag said:

It's known that HBM2+interposer is around $175; VRMs are probably around $30-$40 ish and say PCB is $20 or something.

 

That would indicate that, according to this estimate, that the die itself costs around $350!!! Isn't that a bit high?

 

I feel like this estimate is pretty far off. I don't think it's likely that the die itself costs that much...

You're forgetting the actual cooler and retail+AIB cut, which is significant.

Plus , bad yields can actually drive die cost way up . Early in the gtx 480's producion , nvidia had such terrible yields on gf100 that early chips cost them over 500$ to produce , according to semiaccurate . thankfully , they managed to rectify that before retail launch .

 

 

[Phentos]

When the last Capsaicin event failed to really attempt to demonstrate Vega except for Raja saying that it would be called RX VEGA, I immediately changed my mind and soon placed an order for my 1080 TI once I finalized my new build. I have not looked back with an ounce of regret and the underwhelming results from Vega have confirmed my mindset on that.

 

Honestly it's time for AMD to shed some dead weight (RTG) before the whole ship sinks. The GPU market is almost a lost cause, and AMD can't afford to have the goodwill generated from Ryzen's release marred by Vega's failure.

[VagabondWraith]

So this means that AMD fanboys shouldn't buy Vega since they're causing AMD to lose money then.  

[AresKrieger]

This is why you do not use HBM on consumer cards, way to much money for little to no benefit

[Guest]

Meanwhile some people have to pay nearly or over 1000$ for the 64 LE alone.

[Light-Yagami]

I've watched the Gamers Nexus video on HBM2 pricing. Here's the video. 

All you need to know is there.. I believe all of you've watched it already at this point..

IMO, the whole HMB2 thing is a disaster for AMD. Many were speculating that AMD was making positive zero at best with Vega GPUs, I didn't know it was that bad tho..

[Coaxialgamer]

Just now, VagabondWraith said:

So this means that AMD fanboys shouldn't buy Vega since they're causing AMD to lose money then.  

Doesn't work that way . AMD will lose money on vega regardless of if the cards are bough or not. They'll lose even more on unsold inventory .

[Luxanna]

Imo Vega is their biggest mistake since the release of bulldozer. 

[captain cactus]

HBM2 is a necessity for Vega, power wise and bandwidth wise. The arch needs every mb/sec it can get, especially with almost all of the bandwidth saving features still not working due to drivers. SK Hynix/Samsung just didn't deliver on the original spec of 1GHz HBM2 at a lower cost compared to HBM1 so we're left with lower performance HBM2 stacks until the kinks in HBM2 are sorted out. 

[captain cactus]

2 minutes ago, Luxanna said:

Imo Vega is their biggest mistake since the release of bulldozer. 

It isn't. It's a very nice arch, if everything falls into place. That means everything's enabled in drivers, and the HBM2 supply/cost issues are a big part of Vega's delay and cost. AMD's limited cash pretty much forced them to make a one-size-fits-all architecture. It's amazing at compute (V56 eating 1080s and coming darn near close to 1080Ti) and does pretty well in gaming (V56 is beating a 1070 more often than not, on immature drivers with stuff turned off). Just at a higher power draw (which could be fixed by undervolting).

[Coaxialgamer]

3 minutes ago, Phentos said:

When the last Capsaicin event failed to really attempt to demonstrate Vega except for Raja saying that it would be called RX VEGA, I immediately changed my mind and soon placed an order for my 1080 TI once I finalized my new build. I have not looked back with an ounce of regret and the underwhelming results from Vega have confirmed my mindset on that.

 

Honestly it's time for AMD to shed some dead weight (RTG) before the whole ship sinks. The GPU market is almost a lost cause, and AMD can't afford to have the goodwill generated from Ryzen's release marred by Vega's failure.

I'm going to say AMD shouldn't shed RTG .

They spent billions on ATI . 

It's behind their whole APU strategy . Besides , the importance of GPU will only grow over time . GPGPU and neural networking are gaining traction , and both benefit hugely from very parallel processors such as a gpu .

Without their Graphics IP , it's very likely Sony , microsoft and nintendo would have looked elsewhere for their 8th gen consoles .

Between 2012 and 2016 , graphics also accounted for nearly half of AMd's revenue. I dare say that RTG is the reason AM is still alive today.

[Luxanna]

Just now, lots of unexplainable lag said:

It isn't. It's a very nice arch, if everything falls into place. That means everything's enabled in drivers, and the HBM2 supply/cost issues are a big part of Vega's delay and cost.

What's nice in an arch that's slower than it's main competitor while outputing more heat and consuming more energy? Let's be honest: HBM is the future but Fury X and Vega only prove that the consumer GPU world is not ready for it. 

 

[captain cactus]

Just now, Luxanna said:

What's nice in an arch that's slower than it's main competitor while outputing more heat and consuming more energy? Let's be honest: HBM is the future but Fury X and Vega only prove that the consumer GPU world is not ready for it. 

 

Updated my reply to you, it explains it pretty well.

[AresKrieger]

Just now, valdyrgramr said:

Faster memory, less needed for higher res gaming.  But, the cost doesn't justify that.

Well if the memory is a limiting factor it helps quite a bit but DDR5x wouldn't be the limiting factor for any consumer application so the benefit is minimal, and given that the HBM is increasing their costs by about  $100 a card (HBM cost about 125 a card compared to 20ish for ddr5x from the info I had) it would have made perfect sense to go with the DDR5x even if it did affect performance a bit

[PlayStation 2]

"Kids, you tried your best and you still failed. The lesson is never try."

[Coaxialgamer]

3 minutes ago, lots of unexplainable lag said:

HBM2 is a necessity for Vega, power wise and bandwidth wise. The arch needs every mb/sec it can get, especially with almost all of the bandwidth saving features still not working due to drivers. SK Hynix/Samsung just didn't deliver on the original spec of 1GHz HBM2 at a lower cost compared to HBM1 so we're left with lower performance HBM2 stacks until the kinks in HBM2 are sorted out. 

Samsung never promised 1ghz hbm2 . Hynix did , but they couldn't deliver. 

AMD relied to heavily on Hynix's roadmap , too heavily in fact . They didn't have a failsafe in case they couldn't get hbm2 at a good price.

 

When TSMC 20nm for gpu didn't come to fruition , both AMD and nvidia were able to counteract , because they had planned for it . Both maxwell and Fiji were originally intended for 20nm . Same thing happened with Cayman arch .

 

AMD took a bet , and they lost . Problem is they weren't prepared.

Even if vega gained 10% performance overnight , it still wouldn't be much better.

 

[Luxanna]

Just now, valdyrgramr said:

You seem to not understand.  When Vega was being developed it wasn't intended to tackle a Titan or 1080ti.  AMD stated they were only going after the 1080 at most then people jumped on a hype train expecting more.  Vega does what it is intended to do, but ya AMD is not the best for power consumption control.

No, no no. AMD in it's ad showed off that vega is -magically- going to compete with volta while it struggles against pascal. Show me where and when amd said that vega is going to compete with 1080 because I don't believe amd would do that, because it would kill all the hype, just like the benchmarks before the launch did. 

[79wjd]

12 minutes ago, Phentos said:

 

 

Honestly it's time for AMD to shed some dead weight (RTG) before the whole ship sinks. The GPU market is almost a lost cause, and AMD can't afford to have the goodwill generated from Ryzen's release marred by Vega's failure.

people were suggesting the exact opposite of that before ryzen.

[ravenshrike]

11 minutes ago, Phentos said:

When the last Capsaicin event failed to really attempt to demonstrate Vega except for Raja saying that it would be called RX VEGA, I immediately changed my mind and soon placed an order for my 1080 TI once I finalized my new build. I have not looked back with an ounce of regret and the underwhelming results from Vega have confirmed my mindset on that.

 

Honestly it's time for AMD to shed some dead weight (RTG) before the whole ship sinks. The GPU market is almost a lost cause, and AMD can't afford to have the goodwill generated from Ryzen's release marred by Vega's failure.

 

4 minutes ago, lots of unexplainable lag said:

It isn't. It's a very nice arch, if everything falls into place. That means everything's enabled in drivers, and the HBM2 supply/cost issues are a big part of Vega's delay and cost.

As an architecture it is too expensive for gaming, to be sure. However as a compute architecture it beats the pants off of anything cost/perf except the full up V100 in certain workloads. The gaming market for enthusiast class PCs is much, much smaller than the compute market when all is said and done and in that realm Vega should be quite successful. I suspect that RX Vega will be artificially limited to a small fraction of the number of cards produced with the rest going to Instinct and SSGs.

[captain cactus]

5 minutes ago, valdyrgramr said:

You seem to not understand.  When Vega was being developed it wasn't intended to tackle a Titan or 1080ti.  AMD stated they were only going after the 1080 at most then people jumped on a hype train expecting more.  Vega does what it is intended to do, but ya AMD is not the best for power consumption control.

Actually, the performance/watt for Vega compared to Polaris is up by a big amount. You overclock a 580 into V56 power draw yet V56 still outperforms it by a big amount (or you undervolt V56 to 580 levels and it still outperforms it massively). It's just that the Polaris power draw is so high it sets a high base power draw to work down to.

[The Benjamins]

14 minutes ago, Phentos said:

Honestly it's time for AMD to shed some dead weight (RTG) before the whole ship sinks. The GPU market is almost a lost cause, and AMD can't afford to have the goodwill generated from Ryzen's release marred by Vega's failure.

Sad thing is this is caused by the business decisions made years ago, the newly formed RTG lead by Raga hasn't release a product that was made start to finish by them. I think Navi will be the the tipping point. If it is not a amazing card that beats the competition at a reasonable time frame then it will be very bad for the future of RTG.