AMD Polaris Thread: Radeon RX 480, RX 470 & RX 460 launching June 29th

[Armsdealer]

Time to add to the board "AMD faking marketing slides"...

Well if they aren't the product is a definite dud. WCCF and the marketing slides / Computex demo are telling two diametrically opposing stories about its performance.

Take your pick.

 

[Silverforce11]

The 51% GPU usage thing was pretty stupid I agree. I understand what he was trying to do, but it did come out wrong.

When their rep was asked on Reddit why they did that, their response is they cannot show real performance benchmarks because out of respect of tech journalists and all the tech sites.

They can only show obtuse or "partial" performance numbers to give hints only. Their reasoning was they didn't want to steal the thunder from the tech press.

I disagree with that though because gamers always take results from hardware vendors with a huge grain of salt.

Just like "Pascal is 10x faster than Maxwell"... later we find out only in very unique compute application. Then JHH hyped it, "GTX 1080 is 2x as fast as Titan X".. only in their imaginary VR game that does not exist.

So gamers always will read the launch reviews from the tech press for validation. Even if AMD provided a full benchmark suite, people will still give clicks to tech sites to confirm.

My thought on this, they are holding back because they want to wait til the final launch, giving their driver team more time to *optimize* as much as possible. When they want the tech press to start reviews on it, they'll send a link to download the final bios & driver so these guys can run benchmarks. Prior to this, samples have different bioses and running on non-ready drivers, so any leak will be misleading.

* We all know how AMD improves performance of a new architecture over time via drivers.

 

[Silverforce11]

Time to add to the board "AMD faking marketing slides"...

You obviously have some logic.. what do you think about AMD's slides showing results for a new architecture on 6 months old drivers? Back during that time, they just showed Polaris 11 running Battlefront, on what they said was ES and immature drivers. Yeah, that's the same driver.

 

[Armsdealer]

My thought on this, they are holding back because...

Let's be clear - AMD's #1 problem is an image problem. It's not a technology problem, and even if it were, there are plenty of companies that sell even inferior products with great marketing (Mercedes Benz).

Making sure people are excited about your product at launch is definitely one key way to improve your image. Fiascos like "overclockers dream" are detrimental to brand perception. Having a launch where people's expectations were managed, and then reached a climax with reviews is a very positive thing.

New management seems to be doing good things. Knock on wood.

 

[sirmo]

When their rep was asked on Reddit why they did that, their response is they cannot show real performance benchmarks because out of respect of tech journalists and all the tech sites.

They can only show obtuse or "partial" performance numbers to give hints only. Their reasoning was they didn't want to steal the thunder from the tech press.

That might be a part of the reason but it's not the main reason. They say stuff hardware journos will test like Zen's 40% IPC uplift all the time.

 

[Silverforce11]

That might be a part of the reason but it's not the main reason. They say stuff hardware journos will test like Zen's 40% IPC uplift all the time.

The thing with saying IPC for CPU is it's a ballpark figure that doesn't mean much unless you know the clock speeds.

If AMD says Polaris has X IPC increase, we would already know since we have the clockspeeds for many months now.

Those SiSoft leak are spot on, correct shader count and clocks 1266mhz. I have to think that it was AMD who deliberately leaked those since they were from a long time ago.

 

[sirmo]

The thing with saying IPC for CPU is it's a ballpark figure that doesn't mean much unless you know the clock speeds.

If AMD says Polaris has X IPC increase, we would already know since we have the clockspeeds for many months now.

Those SiSoft leak are spot on, correct shader count and clocks 1266mhz. I have to think that it was AMD who deliberately leaked those since they were from a long time ago.

I run a small hardware manufacturing business. It takes a long time to design and ship an electronics project. The turnaround time between different factories is usually weeks if not longer. AIB's probably had access to Polaris GPU samples long time ago, and it was probably leaked by one of those.

 

[JDG1980]

If the base clock is 1080 MHz and the boost clock is 1266 MHz as recent leaks indicate, that explains something that had confused me for a while: why AMD would say ">5 TFlops" in their presentation for RX 480 instead of giving a more specific number.

2304 shaders at 1080 MHz: ~4.97 TFlops (close enough - in fact, it's probably why they chose that specific clock rate)
2304 shaders at 1266 MHz: ~5.83 TFlops

The most taxing loads, like mining or Furmark (and maybe even AotC on crazy settings), may cause it to hit the power limit and throttle to 1080 MHz. But most games will probably run at the full boost clock without hitting 150W. AIB cards with extra power connectors and better coolers should be able to boost nearly 100% of the time with no ill effects, though (as always) perf/watt may take a hit compared to reference.

Apple cares about GPGPU more than anything else, so this chip will fit very well with the Mac Pro and 27" Retina iMac. Tonga did less than 4 TFlops, and that was at desktop speeds; the laptop version in the iMac probably was closer to 3.5. Even GTX 1070 only does 5.78 TFlops at base and 6.46 TFlops on full boost, and that's a larger and far more expensive chip.

 

[VulgarDisplay]

When their rep was asked on Reddit why they did that, their response is they cannot show real performance benchmarks because out of respect of tech journalists and all the tech sites.

They can only show obtuse or "partial" performance numbers to give hints only. Their reasoning was they didn't want to steal the thunder from the tech press.

I disagree with that though because gamers always take results from hardware vendors with a huge grain of salt.

Just like "Pascal is 10x faster than Maxwell"... later we find out only in very unique compute application. Then JHH hyped it, "GTX 1080 is 2x as fast as Titan X".. only in their imaginary VR game that does not exist.

So gamers always will read the launch reviews from the tech press for validation. Even if AMD provided a full benchmark suite, people will still give clicks to tech sites to confirm.

My thought on this, they are holding back because they want to wait til the final launch, giving their driver team more time to *optimize* as much as possible. When they want the tech press to start reviews on it, they'll send a link to download the final bios & driver so these guys can run benchmarks. Prior to this, samples have different bioses and running on non-ready drivers, so any leak will be misleading.

* We all know how AMD improves performance of a new architecture over time via drivers.

The silly thing is, if amd just released benchmarks from their lab as a reference point to test review sites against they could nip a lot of shady sites in the bud.

Like a timed exclusive kind of deal. If I was in charge of amd marketing the first thing I would do would be to post YouTube videos of amd cards running benchmarks and giving results. Release 100% accurate tests with video proof and let people use your information as a qualifier to what the tech press is saying.

That whole crossfire scaling part of the presentation just seemed really odd like raja meant to say more but the teleprompter failed on him...

 

[casiofx]

The silly thing is, if amd just released benchmarks from their lab as a reference point to test review sites against they could nip a lot of shady sites in the bud.

Like a timed exclusive kind of deal. If I was in charge of amd marketing the first thing I would do would be to post YouTube videos of amd cards running benchmarks and giving results. Release 100% accurate tests with video proof and let people use your information as a qualifier to what the tech press is saying.

That whole crossfire scaling part of the presentation just seemed really odd like raja meant to say more but the teleprompter failed on him...

Amd's marketing always failes so hard, huang's presentation of he performance of nvidia cards are miles better.

 

[grimpr]

Amd's marketing always failes so hard, huang's presentation of he performance of nvidia cards are miles better.

Nvidias presentations are a direct copy of Apples with Steve Jobs, they both use tactics long ago invented by evangelists and preachers to create suspense, hype and suspension of disbelief as to create true believers in their corporate image and products.

 

[casiofx]

Nvidias presentations are a direct copy of Apples with Steve Jobs, they both use tactics long ago invented by evangelists and preachers to create suspense, hype and suspension of disbelief as to create true believers in their corporate image and products.

It's working well, miles better than amd's
For amd users are dying to aee the performance of the parts only to be presented with scraps of info.

 

[grimpr]

It's working well, miles better than amd's
For amd users are dying to aee the performance of the parts only to be presented with scraps of info.

I didnt argued that it doesnt work, these shows are designed for the masses anyways, they work because they use techniques at the subconsious level, suspension of disbelief is the trick at play here. AMDs presentations dont belong in this category, that doesnt mean they are crap, just that they dont play the marketing game that Apple and Nvidia does.

 

[Silverforce11]

The silly thing is, if amd just released benchmarks from their lab as a reference point to test review sites against they could nip a lot of shady sites in the bud.

Like a timed exclusive kind of deal. If I was in charge of amd marketing the first thing I would do would be to post YouTube videos of amd cards running benchmarks and giving results. Release 100% accurate tests with video proof and let people use your information as a qualifier to what the tech press is saying.

That whole crossfire scaling part of the presentation just seemed really odd like raja meant to say more but the teleprompter failed on him...

If that's the case why don't NV and Intel also just release 100% accurate benchmarks, why bother with the tech press at all?

I mean in this day and age, social media, streams and youtube >> websites for PR in terms of reaching the masses.

The problem always comes down to doubt. If IHVs show benches for their own products, the consumer is always wary and untrusting so it's a pointless exercise.

 

[torlen11cc]

the power is split into 6+1 phases, which is actually more than GTX 1080 (5+1).

http://videocardz.com/61091/amd-radeon-rx-480470-pcb-pictured-up-close
Does that make sense? what is the meaning of more phases?

 

[trane]

Amd's marketing always failes so hard, huang's presentation of he performance of nvidia cards are miles better.

While there's no doubt Nvidia does marketing far better than AMD, I disagree about the presentations. Raja killed it this time, with the right poise and humour, while Jensen's Pascal announcement was predictable and messy.

Granted, that Ashes benchmark was weird, but his presentation style was far more sophisticated than anything Jensen does.

 

[sirmo]

I dunno, Raja wasn't very good personally. I think he could have delivered it better, he looked tired and was doing too much slide reading. Jensen is a used car salesman shyster though. My grandma would do a better job.

2100Mhz @67C FE cooler! "Irresponsible amounts of performance!" Nvidia's presentation was the tackiest product showcasing I've ever seen.

After that presentation I felt like I got conned into buying a bridge. And I didn't even spend any money.

 

[trane]

I dunno, Raja wasn't very good personally. I think he could have delivered it better, he looked tired and was doing too much slide reading. Jensen is a used car salesman shyster though. My grandma would do a better job.

2100Mhz @67C FE cooler! "Irresponsible amounts of performance!" Nvidia's presentation was the tackiest product showcasing I've ever seen.

After that presentation I felt like I got conned into buying a bridge. And I didn't even spend any money.

I think that's what I liked - he was honest and precise. It was all understated, there was no big drama. He's of course not a good speaker - he's an engineer after all. But I loved his many potshots at Nvidia. They were subtle, and that's what made them hilarious.

But I get why most wouldn't like it - it may seem too understated. I don't think he was tired, that's just how he is.

 

[Magee_MC]

I think that AMD is following the old premise of underpromising and overdelivering. They have been pushing perf/watt with polaris and that is what they will deliver. P10 will come in at around 100 watts and be a solid performer, and the reference blower will work effectively to keep the chips cool at that low of a wattage.

They're limiting the reference chips to a single 6 pin so that they don't run into the problem of hot and loud that they ran into with Hawaii. I expect that there will be plenty of OC headroom to push P10 out of its optimum perf/watt range with aftermarket cards.

Doing it this way allows AMD to show that the chips can run cool and effectively, while allowing people who want maximum performance to get a chip with either a 6+6 or 6+8 power setup and push them as hard as the cooling will allow, without any hot and loud backlash splashing on AMD.

 

[Riek]

You are using the 480 stats of >5TFlops not 470.

I used the rumoured core count of 470 ~2048 and clocks of 1200. which gives: 4,9Tflop.

If we would go by the 1536 we would have around 1536*1200 ~3,7Tflop which would make alot more sense going by the difference in performance between the 270x and the 470x.

 

[renderstate]

I didnt argued that it doesnt work, these shows are designed for the masses anyways, they work because they use techniques at the subconsious level, suspension of disbelief is the trick at play here. AMDs presentations dont belong in this category, that doesnt mean they are crap, just that they dont play the marketing game that Apple and Nvidia does.

I'd argue there is nothing redeeming in giving bad presentations and RTG latest presentations are really bad on all fronts and I disagree they don't play the marketing game. They try very hard but they fail at that. Another awful presenter is Elon Musk but that doesn't seem to damage his companies

 

[renderstate]

Raja Koduri needs to improve his presentation skills. Understated doesn't mean bad, his problems are others. First of all he doesn't know what to say, showing he hasn't prepared well and/or there is no logical connection between a slide and the next slide, ultimately leading him to reading the slide content line per line, which if the worst thing a presenter can do. Second he delivers without any confidence. Third he says a lot of words but delivers little new content. His RX480 presentation was pretty terrible.

 

[3DVagabond]

What are you talking about? disappointment? We're still at the confusion/hype stage.

http://wccftech.com/amd-rx-480-faster-than-nano-980/
http://videocardz.com/61005/new-amd-radeon-rx-480-3dmark-benchmarks

The wccf article is only 15 hours old!

Two weeks to launch, that's a very long time.

To AMD's benefit, they haven't been over promising or hyping at all. They've been talking about perf/watt, VR and low prices for the last 6 months. That's it. No overclocker's dream, no anything. They have been very consistent with their message.

Most of the hype is from people refusing to accept that.

Most of the hype is non existent being exaggerated by overstating people's expectations just to make ridiculous claims against them.

Some poster says he heard an unconfirmed report that Asus is releasing a 1600MHz part and to take it with a grain of salt. Now it's supposedly everyone's expectations and somehow it's AMD's fault.

 

[grimpr]

I'd argue there is nothing redeeming in giving bad presentations and RTG latest presentations are really bad on all fronts and I disagree they don't play the marketing game. They try very hard but they fail at that. Another awful presenter is Elon Musk but that doesn't seem to damage his companies

For you everything AMD does is either bad or failed, your biased assumption and the context you communicate is AMD = failure, so its no point arguing about it any further, enjoy Apples and Nvidias presentations.

 

[3DVagabond]

Indeed. Pretty sure AMD would make a big deal if this card was 'VR-ready', but it looks like it's below the minimum required (R9 290/GTX 970).

VR ready for $199 and 2.8X perf/W are the claims AMD has made. You obviously don't want to talk about what they've actually said. Instead you want to talk about what they haven't said. Why is that?

If they can't compete on perf/mm² and perf/watt at the bottom they might have some trouble matching GP102 when it hits the market, HBM2 or not. At this point I don't think NVIDIA needs much more than a high-clocked 1/2 GP104 to remain competitive with Polaris 10 (estimated at ~260mm²). How much 260mm² worth of 14nm LPP translates into 28nm planar in terms of size? Shouldn't it blow Hawaii out of the water?

What is it nVidia has that the 480 can't compete with? Have you been given the actual size and power consumption of any of these cards that you are lining up against each other and claiming AMD can't compete??

He also suggests Radeon RX 470 / RX 460 won't be available this month. I hope he's wrong, because if these cards get further delayed AMD is missing an opportunity to regain some marketshare, as Inno3D is already teasing a Pascal-based Geforce GTX 970/980 replacement from the competitor.

AMD has said the 29th for all of them, I believe. Also, please inform us of the 1st delay for them so we can decide if there is a further delay. Again, AMD said months ago 2nd Q and the 29th meets that deadline.

What is this card from Inno3D that replaces the 980/970? Wasn't that the GM104? Isn't the GP104 the replacement? You must be talking some other Maxwell replacement. When has nVidia announced the release of this GPU?

What are the facts or references to back up any of your conclusions?