AMD Confirms RX 480 At $199 USD, Other APU & Polaris Announcements

[Stacc]

Right, and Intel is also not coming up sooner because of Ryzen/Naples.
We're all believers.
There might be no actual "evidence" of the rush but do we believe if any of the involved parties could milk a design a bit further they would just launch something else ahead of time?
I'm counting on the usual suspects to refute with their infinite wisdom of course, we know how it goes. Maybe keep OEMs happy, since rebranding wouldn't cut it? Or maybe some other excuse? Sorry, reason.
Don't really want to pick a(nother) fight here with the usual people but come on, there's always a reason behind moving a schedule, in either direction.
I won't discuss this anymore, I foresee another long exchange of nasty posts.

Of course companies shift schedules depending on their competition, but product develop cycles are very long. All of this "Intel has been caught on their heels by Ryzen" or "Nvidia has pushed up their schedule because they found out how good Vega is" and these sorts of storylines are ridiculous. For example pushing up a chip release from 1 year out to 6 months out is usually impossible, simply because designs have to be given to fabs roughly a year in advance, and then there is usually a relatively set schedule of preparing mass production.

Not only are development cycles long, but usually companies have a pretty good idea of what their competition is up to. If you, random message board poster, has known about Vega (and its older codenames) for years, so has Nvidia. Also, Nvidia has tended to be ahead of AMD when it comes to predicting market trends and targeting new markets (see developing CUDA for HPC before there existed a market for it or embedded GPUs for robotics) so suggesting they are caught off guard goes against the last 10 years of GPU development.

 

[ManuelGomes]

Twice is highly doubtful but who knows?! Would be nice indeed.
NVidia will continue to dominate, have no doubts about that, at least until AMD has their complete act together, and maybe even then. Resources available to them are in another level I guess.
But one thing seems to be true though. AMD is trying to come back and shake things a lot. And they're getting everyone on their toes when everyone took them for dead, or dying. An eye opener no doubt, for those that take stuff for granted.

 

[cube]

What happens if Volta is twice as fast as the 1080 Ti?

Even the 1070 is too expensive.

I am looking forward to a faster and cheaper Vega.

 

[koyoot]

What happens if Volta is twice as fast as the 1080 Ti? Perhaps NVIDIA has moved the schedule up because it smells blood in the water (i.e. Vega delayed by 6+ months) and wants to continue dominating at the high end. Maybe Vega will be able to squeak out a few wins versus the 1080 Ti, maybe it'll be faster across the board, or maybe it'll lose across the board. We still don't know, because the release keeps getting pushed further and further back. I continue to be amused at koyoot's posts claiming that AMD is coming from a position of strength here, or that NVIDIA is in trouble.

I love when people are reading way too much in my posts.

Nvidia has a brand to protect, especially in HPC market. Vega is threat to it, regardless of what naysayers will tell you.

Two things you guys do not know. Already Vega GPUs are powering cloud computing . Secondly, AMD confirmed that Vega is on track to Q2 release, with very wide Q3 availability. It happened in yesterdays conference call.

Manuel, Intel did not pushed forward the releases of their HEDT CPUs. They knew that Ryzen will be 8C/16T design, so they have added 10 core design to their Broadwell-E lineup. They have thought that AMD is coming with their HEDT CPUs, but they did not know it was actually mainstream offering. Right now, Intel, to maintain high margin on HEDT must release 12 core part with Skylake HEDT CPUs. That is how this market works. I am still baffled that you guys do not know how this works. Each of the companies is constantly spying over their competitors, to avoid brand damage.

Funniest part is this. AMD does not have anything to lose right now when brand is considered. Nvidia and Intel have A LOT to lose. Think about this guys.

Of course companies shift schedules depending on their competition, but product develop cycles are very long. All of this "Intel has been caught on their heels by Ryzen" or "Nvidia has pushed up their schedule because they found out how good Vega is" and these sorts of storylines are ridiculous. For example pushing up a chip release from 1 year out to 6 months out is usually impossible, simply because designs have to be given to fabs roughly a year in advance, and then there is usually a relatively set schedule of preparing mass production.

Product development cycles are long if you are building brand new architecture. For consumer GPUs Nvidia will reuse the GP100 chip layout with 64 cores per SM.

It will bring 30% improvement core for core, clock for clock. In essence. Expect that 3840 CUDA core with 1.2 GHz, will be 40% faster than Titan Xp, which is clocked at almost 1.6 GHz.

 

[ManuelGomes]

Right, but everyone seems to be having trouble with their roadmaps lately.
Anyway, Lisa confirmed Vega is on schedule. I am curious as to the performance of Vega really, with so much said I hope AMD delivers this time around.

 

[Asgorath]

Even the 1070 is too expensive.

I am looking forward to a faster and cheaper Vega.

What makes you think Vega will be cheaper than a 1070? You can get 1070s in the $350-380 range from a quick search.
[doublepost=1493750539][/doublepost]

Two things you guys do not know. Already Vega GPUs are powering cloud computing .

Source?

 

[Stacc]

Funniest part is this. AMD does not have anything to lose right now when brand is considered. Nvidia and Intel have A LOT to lose. Think about this guys.

AMD has a lot to lose. Both Intel and Nvidia are doing much better financially as companies. If AMD can't compete in HPC with Zen and Vega then they could go out of business.

Product development cycles are long if you are building brand new architecture. For consumer GPUs Nvidia will reuse the GP100 chip layout with 64 cores per SM.

It will bring 30% improvement core for core, clock for clock. In essence. Expect that 3840 CUDA core with 1.2 GHz, will be 40% faster than Titan Xp, which is clocked at almost 1.6 GHz.

Once again you seem to be unable to grasp the fact that GP100 is the same pascal as the reset of the lineup but with HPC features. Volta of course is a new architecture, and it is not the same as GP100. We should learn more about Volta when Nvidia talks about it next week.

 

[cube]

What makes you think Vega will be cheaper than a 1070? You can get 1070s in the $350-380 range from a quick search.

I know. Those are American prices and still quite high.

AMD brought very competitive $100 and $200 cards with Polaris. If Vega is much faster, I expect a $300 entry.

Ideally, it would be GTX1080-level performance for $300.

 

[Stacc]

I know. Those are American prices and still quite high.

AMD brought very competitive $100 and $200 cards with Polaris. If Vega is much faster, I expect a $300 entry.

Ideally, it would be 1080 performance for $300.

Don't expect AMD to deliver performance for free especially if they feel it adds value over the competition. If an AMD card offers the same gaming performance as a GTX 1080 but has better compute performance expect very similar pricing to the competition.

 

[cube]

Don't expect AMD to deliver performance for free especially if they feel it adds value over the competition. If an AMD card offers the same gaming performance as a GTX 1080 but has better compute performance expect very similar pricing to the competition.

Most people are using these cards for gaming and not compute, and AMD does not have an stranglehold on the market allowing it to charge exorbitant prices. For $350, it would still accomplish the milestone of offering GTX1080 performance for GTX1070 price.

GTX1070-level performance for $300 would not be competitive enough. I could see this priced at $250 given that I do not expect much change in what they charge for the RX5, though.

 

[Asgorath]

Most people are using these cards for gaming and not compute, and AMD does not have an stranglehold on the market allowing it to charge exorbitant prices. For $350, it would still accomplish the milestone of offering GTX1080 performance for GTX1070 price.

GTX1070-level performance for $300 would not be competitive enough. I could see this priced at $250 given that I do not expect much change in what they charge for the RX5, though.

If they do this, they'll be out of business fairly quickly. Vega is a huge GPU that is expensive to make. They can't just magically make all of that go away and be profitable with $300 GPUs. Remember, AMD has not been profitable in a very long time, and so you shouldn't be expecting more performance for significantly less money than the equivalent NVIDIA GPUs. The latest NVIDIA offerings are very aggressively priced (e.g. 1080 Ti) and that's going to make it hard for AMD to make a lot of money on Vega.

 

[Stacc]

Most people are using these cards for gaming and not compute, and AMD does not have an stranglehold on the market allowing it to charge exorbitant prices. For $350, it would still accomplish the milestone of offering GTX1080 performance for GTX1070 price.

GTX1070-level performance for $300 would not be competitive enough. I could see this priced at $250 given that I do not expect much change in what they charge for the RX5, though.

This is the problem with a lot of people loyal to AMD. They buy AMD because its cheap and provides value. If AMD can compete at the high end, they are going to charge like its a high end product. They tried to do this with the Fury line, marketing it as a premium brand just like they are going to market Vega as its own "RX Vega" lineup.

Currently, the RX 480 provides roughly the same performance as a GTX 1060 for 10% less cost. So if we apply this to the GTX 1080 then expect AMD's competitor to cost $450, not $250. Like Asgorath pointed out, AMD is a business and has to turn a profit. Vega 10 is likely more expensive to produce than the GTX 1080 Ti, so there is a reasonable floor for how low they can price it.

 

[cube]

If they do this, they'll be out of business fairly quickly. Vega is a huge GPU that is expensive to make. They can't just magically make all of that go away and be profitable with $300 GPUs. Remember, AMD has not been profitable in a very long time, and so you shouldn't be expecting more performance for significantly less money than the equivalent NVIDIA GPUs. The latest NVIDIA offerings are very aggressively priced (e.g. 1080 Ti) and that's going to make it hard for AMD to make a lot of money on Vega.

Who said there's only one Vega chip?

NVIDIA likes to overcharge. They are using similar process technology so of course there's margin for AMD to match year-old NVIDIA chips for a much lower price.

 

[Asgorath]

Who said there's only one Vega chip?

NVIDIA likes to overcharge. They are using similar process technology so of course there's margin for AMD to match year-old NVIDIA chips for a much lower price.

There's no doubt that competition is a good thing, but compare the profitability of NVIDIA and AMD. NVIDIA has been in the black for as long as I can remember, and continues to make solid profits on their GeForce GPUs. AMD has been losing money hand over fist for years, and there's little sign of this changing even with Vega. As I said, AMD can't just wave a magic wand and have all the money they've invested in developing Vega suddenly be refunded so that they can make profits on a $300 Vega GPU.

 

[Stacc]

They are using similar process technology so of course there's margin for AMD to match year-old NVIDIA chips for a much lower price.

Right, this logic doesn't follow. Nvidia has been getting basically all high end GPU sales for a year, which means they have recuperated a lot of the development cost. If anything it means that Nvidia would have an easier time undercutting AMD if they chose.

 

[Asgorath]

Right, this logic doesn't follow. Nvidia has been getting basically all high end GPU sales for a year, which means they have recuperated a lot of the development cost. If anything it means that Nvidia would have an easier time undercutting AMD if they chose.

Exactly. This is probably why the 9-month-old GP102 in the 1080 Ti was priced aggressively at $700 instead of $1000 as most people were expecting. Even if Vega solidly beats the Pascal GPUs, NVIDIA has already made so much money from the architecture that they can drop the prices if needed until Volta comes out. It's a hard sell for Vega at $800 or $900 (which is probably what AMD would like to be able to charge) unless it's a huge knock-out win for them across the board (which I suspect is unlikely to happen).

 

[cube]

Right, this logic doesn't follow. Nvidia has been getting basically all high end GPU sales for a year, which means they have recuperated a lot of the development cost. If anything it means that Nvidia would have an easier time undercutting AMD if they chose.

The logic follows. NVIDIA will lower their prices to better match Vega, in the same way the 1060 was released to fit RX480.

 

[Asgorath]

The logic follows. NVIDIA will lower their prices to better match Vega, in the same way the 1060 was released to fit RX480.

NVIDIA charged more for the 1060 because it beat the RX 480 while consuming much less power? I don't follow your point.

 

[cube]

NVIDIA charged more for the 1060 because it beat the RX 480 while consuming much less power? I don't follow your point.

The RX480 also beats the 1060 and they are priced similarly.

Look at the performance of the RX480 vs GTX970 and their respective launch prices.

 

[Asgorath]

The RX480 also beats the 1060 and they are priced similarly.

Look at the performance of the RX480 vs GTX970 and their respective launch prices.

Except that it doesn't, the 1060 was better on average (yes there were outliers where the RX 480 wins, but those were in the minority). That's why NVIDIA was able to charge $10 more for the 1060.

I'm also not sure what the 970 has to do with any of this?

 

[cube]

Except that it doesn't, the 1060 was better on average (yes there were outliers where the RX 480 wins, but those were in the minority). That's why NVIDIA was able to charge $10 more for the 1060.

I'm also not sure what the 970 has to do with any of this?

You must have looked at benchmarks with old drivers.

The Vega equivalent to GTX1070 would be a similar case to RX480.

Do not forget that Vega needs half the VRAM, and that architectural improvements will unleash latent power, so that it will be possible to do more with less.

 

[Stacc]

You must have looked at benchmarks with old drivers.

The Vega equivalent to GTX1070 would be a similar case to RX480.

Do not forget that Vega needs half the VRAM, and that architectural improvements will unleash latent power, so that it will be possible to do more with less.

The RX 580 does slightly beat out the GTX 1060, but this is at a significant efficiency cost since the RX 580 is essentially just an overclocked RX 480. These are also done with recent drivers and includes AMD favorable tests like Vulkan and DX12 benchmarks.

I'm going to take all the Vega architecture claims AMD has thrown out there with a grain of salt until we get to read reviews and see what the performance actually looks like.

 

[Asgorath]

Right, the RX 580 uses slightly more power than the GTX 1080 (185W vs 180W) and loses by 58% on average. I really don't understand how people can claim AMD has the most efficient architecture, based on data like this.

 

[ActionableMango]

One of the funniest things about AMD for this example is that sometimes people are intentionally misinformed about the products and the details about them, to avoid leaks. Even Apple is not doing anything like this.

Few companies, indeed, are more secretive than Apple, or as punitive to those who dare violate the company’s rules on keeping tight control over information. Employees have been fired for leaking news tidbits to outsiders, and the company has been known to spread disinformation about product plans to its own workers.​

http://www.nytimes.com/2009/06/23/technology/23apple.html

 

[cube]

Right, the RX 580 uses slightly more power than the GTX 1080 (185W vs 180W) and loses by 58% on average. I really don't understand how people can claim AMD has the most efficient architecture, based on data like this.

I would not buy the RX580. To increase a little the clock, you have to feed it quite a bit more power.

RX480 can indeed already beat the 1060 in some games.

Anything that includes Rise of the Tomb Raider is useless, as this game seems not optimized at all for AMD.

And 3GiB VRAM is not enough for some games at 1080p Ultra.