Nvidia fell for an orchestrated pricing trap, says AMD exec

Cal Jeffrey

Posts: 4,174   +1,422
Staff member
Bottom line: The GPU wars are on, and AMD is playing hardball. Its latest move in cutting the prices on its RX 5700 cards was not a reaction to Nvidia’s Super 20 series launch, but rather a sneaky pre-planned pricing strategy... or so AMD says.

As you may recall, when AMD unveiled the Radeon RX 5700 and 5700 XT, it listed them with an MSRP of $379 and $449 respectively. When Nvidia launched its GeForce RTX 2060 and 2070 Super cards in an apparent attempt to torpedo AMD’s RX release, AMD responded by slashing its prices to $349 and $399.

However, in an interview with Hot Hardware (below 26:30), AMD Vice President and General Manager of the Radeon Division Scott Herkelman said the price cuts were a pre-planned maneuver to get Nvidia to overprice its Supers.

“We know that they [Nvidia] have a tremendous burden on their gross margins, because their die size is big, and so we just game-played suggested retail prices,” said Herkelman.

However, it was more than just a play to get its rival to bump its pricing. Herkelman says that they saw that the regular RTX 2060 and 2070 were struggling with slow adoption. AMD wanted to cause a “logjam” there too.

“We made the appropriate move not only to deposition their Super series but also to logjam their 2060 and 2070 because we knew that they’re having slower success,” Herkelman admitted. “And we wanted to do a double fake, which was not only to block their Super strategy but also slow down their 2060 and 2070.”

Was it an underhanded move? Sure, but all is fair when it comes to beating out your competitor, and it ultimately results in a win for the consumer. As far as Herkelman is concerned, it’s a win for AMD as well.

“It's been so much fun these last couple of weeks just playing that out,” said Herkelman. “It's just fun finally to win.”

Indeed, it is just as fun for the consumer watching the GPU makers duke it out to the tune of lower prices.

Permalink to story.

 
Can I be convinced we are in a place where we were before crypto destroyed the market? I may consider building a new rig when a descent CPU with overclocking headroom is released.

This is a great PR move, but after over a decade of trying to stay with AMD and only getting driver problem after driver problem, it would be like driving a stake through my hand to buy AMD again. It would only be with a full refund warranty, which would drive the price up. I am not seeing making the switch for quite a while. :/
 
Can I be convinced we are in a place where we were before crypto destroyed the market? I may consider building a new rig when a descent CPU with overclocking headroom is released.

This is a great PR move, but after over a decade of trying to stay with AMD and only getting driver problem after driver problem, it would be like driving a stake through my hand to buy AMD again. It would only be with a full refund warranty, which would drive the price up. I am not seeing making the switch for quite a while. :/
Driver problems? Where?
 
Nvidia may have large die sizes, but the previous generation of AMD, before RDNA came along, had to use stuff like HBM2 to get performance comparable to Nvidia's, as the CEO of Nvidia was so kind as to point out.
 
Driver problems? Where?

Where? At my house, and many other's houses as well.
I supposed you meant when. From around 2004 through 2016, so I was wrong. They are worse than I have been saying. 12+ years. And I still have the cards in computers laying around. As of their last boot, the problems were still there.
 
Half the time I hear about price reductions and then go look on the sites I actually shop on ie Amazon, Newegg, etc, the prices are the same. In rare cases, higher, but that happens enough for me to be skeptical. The only valid 'news' regarding lower prices is the total in my cart before I hit checkout/buy.
 
Nvidia may have large die sizes, but the previous generation of AMD, before RDNA came along, had to use stuff like HBM2 to get performance comparable to Nvidia's, as the CEO of Nvidia was so kind as to point out.
And while nVidia was looking the other way and laughing, AMD is sneaking up on them. The effect nVidia is currently experiencing is "just a flesh wound" I am sure!
 
AMD claiming they meant to is pure bs. The CEO saying they meant to do that is no different than someone falling down, then saying I meant to do that.

more like AMD trying to throw a punch but for some reason they keep hitting themselves over n over.

nvidia has no reason to look down to see that they just stepped on, afterall its just AMD products.
nvidia has felt zero effects from AMD. Hell if it wasnt for AMD even saying anything no one would know AMD still made video cards.
 
This is typical AMD advertising tactics. You sling the sh!t at Nvidia, (or whomever),before you have the product to back it up.

(Although I have heard that AMD's pro line "FirePro" cards are quite good). Anybody want to sound off on that, or call me a troll? C'mon dive right it..
 
Nvidia may have large die sizes, but the previous generation of AMD, before RDNA came along, had to use stuff like HBM2 to get performance comparable to Nvidia's, as the CEO of Nvidia was so kind as to point out.

And RDNA has to use optical shrink to get comparable performance (still got nothing on RTX 2080 Ti) to Nvidia.

Is it finally 2020 so we can real, actual, meaningful 7nm GPU? Why the heck 3 years later we still haven't doubled GTX 1080 performance comfortably?
 
I thought AMD makes either useless cards or mining cards. What do those have to do with GPUs that nVidia makes?
 
I'm happy AMD have something good on the GPU front, but realistically Nvidia still hold most of the cards.

You're looking at an AMD GPU on TSMC 7nm not quite half the die size of say a TSMC 12nm 2060 Super with similar performance and similar power metrics.

Problem with that is if Nvidia take that 2060 Super, shrink it on Samsung 7nm EUV it will then be basically the same size as AMD's Navi parts. If not smaller. But it'll also be about 25 percent faster, based on easy clock speed gains. While using less power.

Nvidia won't just die shrink Turing though, will they? They will tweak their architecture as well. It'll be even faster above and beyond mere clock speed gains.

So the fact remains whilst Navi is an advance for AMD's GPU architecture, it isn't enough if they go head to head with Nvidia on process parity. Nvidia's arch is just ridiculously superior.

I just am hoping some of that tasty Zen 2 cash this year is flooding into AMD's pockets and right to their R&D for future GPUs.
 
Initially Nvidia just kinda matched them price vs performance with the Supers. Really though I looked back on the 570 pricing and damn 169$... prices are still nuts, Apr 18, 2017. How about MINER cards and GAMER cards priced separately? I was hoping for at least a sub 300$ vega 64 or 250$ used, but nope looks like eol. But will get that 5600 soon?
 
This is typical AMD advertising tactics. You sling the sh!t at Nvidia, (or whomever),before you have the product to back it up.

(Although I have heard that AMD's pro line "FirePro" cards are quite good). Anybody want to sound off on that, or call me a troll? C'mon dive right it..
AMD doesn't really need to win at every price point and we'll most likely see the 5600 series release this year too. They just need to continue offering good prices.
 
Can I be convinced we are in a place where we were before crypto destroyed the market? I may consider building a new rig when a descent CPU with overclocking headroom is released.

This is a great PR move, but after over a decade of trying to stay with AMD and only getting driver problem after driver problem, it would be like driving a stake through my hand to buy AMD again. It would only be with a full refund warranty, which would drive the price up. I am not seeing making the switch for quite a while. :/

There is a saying that usually people who have nVidia have the most issues with AMD's drivers, I'm not sure if that's the case here but I've had only AMD cards in the last 5 - 6 years and had no problems, especially the last few years they have really upped their game in that department :)
 
I'm happy AMD have something good on the GPU front, but realistically Nvidia still hold most of the cards.

You're looking at an AMD GPU on TSMC 7nm not quite half the die size of say a TSMC 12nm 2060 Super with similar performance and similar power metrics.

Problem with that is if Nvidia take that 2060 Super, shrink it on Samsung 7nm EUV it will then be basically the same size as AMD's Navi parts. If not smaller. But it'll also be about 25 percent faster, based on easy clock speed gains. While using less power.

Nvidia won't just die shrink Turing though, will they? They will tweak their architecture as well. It'll be even faster above and beyond mere clock speed gains.

So the fact remains whilst Navi is an advance for AMD's GPU architecture, it isn't enough if they go head to head with Nvidia on process parity. Nvidia's arch is just ridiculously superior.

I just am hoping some of that tasty Zen 2 cash this year is flooding into AMD's pockets and right to their R&D for future GPUs.

You can't assume they will get 50% reduction in size and clocks won't necessary go up that much, I will be surprised if we ever get 3Ghz GPU since we are reaching silicon limitations now. But yes AMD still has a lot of work to do the thing that saves them right now it the lack of RT that saves them space but if nVidia wants it to be popular they have to go big to make it work on mainstream cards
 
There is a saying that usually people who have nVidia have the most issues with AMD's drivers, I'm not sure if that's the case here but I've had only AMD cards in the last 5 - 6 years and had no problems, especially the last few years they have really upped their game in that department :)
Well, I have an ancient P-45 / Wolfdale E7300 rig, w/ XP SP-3.

I put an Nvidia GT-710 in it, to replace an old 9500 GT card. My logic was, the 710 will use much less energy, and has a lot (2 GB) more VRAM.than the old card.

Naturally, Nvidia wouldn't let me run the old 180 something series driver, and forced me to install, a 329. dot sumpin' sumpin' driver.

Well, every time I'm working the machine a bit, and fire up the VLC player for a movie, if I pull the scroll bar a bit to quickly. or to a place the video card doesn't want it, the driver crashes. "The video driver has stopped working", blah, blah, blah.

This isn't exactly the first issue I've had with this card. I had it in my internet box, (G-41 Pent 2 core), and went back to the Intel IGP.

All I can say is, if AMD drivers are worse than the GeForce drivers I've experienced, all I have to say is, "god be with you my sons"..
 
Great AMD, you can talk the talk.

Now how about walking the walk? Your pricing shenanigans only hurt the 2070 super, arguably. The 2080, 2080 super, and 2080ti are still printing a mint for nvidia. Playing silly pricing games doesnt really matter if you still cant get past the performance of Vega 64.

Driver problems? Where?

Where? At my house, and many other's houses as well.
I supposed you meant when. From around 2004 through 2016, so I was wrong. They are worse than I have been saying. 12+ years. And I still have the cards in computers laying around. As of their last boot, the problems were still there.
Perhaps you should stop judging modern AMD cards by nearly 3 year old graphics drivers? It isnt 2004 anymore, and it isnt 2016 anymore. It's 2019, and AMD's drivers are right behind Nvidia's in quality. Hell, in 2016 the RX 480 was out, and AMD's drivers were already rapidly approaching Nvidia quality and were a far cry from the drivers of just a year earlier.
 
Last edited:
Well, I have an ancient P-45 / Wolfdale E7300 rig, w/ XP SP-3.

I put an Nvidia GT-710 in it, to replace an old 9500 GT card. My logic was, the 710 will use much less energy, and has a lot (2 GB) more VRAM.than the old card.

Naturally, Nvidia wouldn't let me run the old 180 something series driver, and forced me to install, a 329. dot sumpin' sumpin' driver.

Well, every time I'm working the machine a bit, and fire up the VLC player for a movie, if I pull the scroll bar a bit to quickly. or to a place the video card doesn't want it, the driver crashes. "The video driver has stopped working", blah, blah, blah.

This isn't exactly the first issue I've had with this card. I had it in my internet box, (G-41 Pent 2 core), and went back to the Intel IGP.

All I can say is, if AMD drivers are worse than the GeForce drivers I've experienced, all I have to say is, "god be with you my sons"..

Hmm in the last few years I've had R9 280X >> R9 390X >> RX Vega 64 >> Radeon VII and had 0 problems with VLC media player ;-P
 
With Control (DXR), Borderlands 3 and The Outer World (UE4 engine) imminent release, 2060S is looking rather good for the people who are upgrading their PC to play those games. I'm also quite interested in these game also, the sad thing is they all release on Epic game store.
 
Last edited:
You can't assume they will get 50% reduction in size and clocks won't necessary go up that much, I will be surprised if we ever get 3Ghz GPU since we are reaching silicon limitations now. But yes AMD still has a lot of work to do the thing that saves them right now it the lack of RT that saves them space but if nVidia wants it to be popular they have to go big to make it work on mainstream cards

Based on what is known about 7nm EUV my 'assumptions' were conservative estimates.

Even a large die, 'bad' architecture on a technically inferior (TSMC 7nm DUV) process hit these targets.

Radeon VII being a prime example, achieving ~ 1.6 times the density of Vega 64. With the aid of better memory, minor tweaks and higher sustained clocks it is about 30 percent faster. Despite being binned and having less shaders enabled.

Two main reasons why Nvidia might have chosen Samsung 7nm EUV. It's cheaper for them or it's better than TSMC DUV. Quite likely both. On paper it's equivalent to TSMC's EUV, which is 10-15 percent (realistically- TSMC claim 20%) more dense than their current DUV process.

In other words taking a 12nm TSMC 475mm² die down to ~250mm² is pretty much the sort of optical shrink you would expect from 7nm EUV. The other factors are pretty much the performance increase you should also expect with minimal enhancements to the core design itself.

It could be slightly worse, it could be slightly better, but it's ballpark. There are no guarantees in this business but the targets are there and they're realistic.
 
Last edited:
Back