Friday, September 4th 2015
AMD Radeon R9 Nano Review by TPU...Not
There won't be a Radeon R9 Nano review on TechPowerUp. AMD says that it has too few review samples for the press. When AMD first held up the Radeon R9 Nano at its "Fiji" GPU unveil, to us it came across as the most promising product based on the chip, even more than the R9 Fury series, its dual-GPU variant, and the food-processor-shaped SFF gaming desktop thing. The prospect of "faster than R9 290X at 175W" is what excited us the most, as that would disrupt NVIDIA's GM204 based products. Unfortunately, the most exciting product by AMD also has the least amount of excitement by AMD itself.
The first signs of that are, AMD making it prohibitively expensive at $650, and not putting it in the hands of the press, for a launch-day review. We're not getting one, and nor do some of our friends on either sides of the Atlantic. AMD is making some of its tallest claims with this product, and it's important (for AMD) that some of those claims are put to the test. A validated product could maybe even convince some to reach for their wallets, to pull out $650.Are we sourgraping? You tell us. We're one of the few sites that give you noise testing by some really expensive and broad-ranged noise-testing equipment, and more importantly, card-only power-draw. Our reviews also grill graphics cards through 22 real-world tests across four resolutions, each, and offer price-performance graphs. When NVIDIA didn't send us a GeForce GTX TITAN-Z sample, we didn't care. We didn't make an announcement like this. At $2,999, it was just a terrible product and we never wished it was part of our graphs. Its competing R9 295X2 could be had under $700, and so it continues to top our performance charts.
The R9 Nano, on the other hand, has the potential for greatness. Never mind the compact board design and its SFF credentials. Pull out this ASIC, put it on a normal 20-25 cm PCB, price it around $350, and dual-slot cooling that can turn its fans off in idle, and AMD could have had a GM204-killing product. Sadly, there's no way for us to test that, either. We can't emulate an R9 Nano on an R9 Fury X. The Nano appears to have a unique power/temperature based throttling algorithm that we can't copy.
"Fiji" is a good piece of technology, but apparently, very little effort is being made to put it into the hands of as many people as possible (and by that we mean consumers). This is an incoherence between what AMD CEO stated at the "Fiji" unveil, and what her company is doing. It's also great disservice to the people who probably stayed up many nights to get the interposer design right, or sailing through uncharted territory with HBM. Oh well.
The first signs of that are, AMD making it prohibitively expensive at $650, and not putting it in the hands of the press, for a launch-day review. We're not getting one, and nor do some of our friends on either sides of the Atlantic. AMD is making some of its tallest claims with this product, and it's important (for AMD) that some of those claims are put to the test. A validated product could maybe even convince some to reach for their wallets, to pull out $650.Are we sourgraping? You tell us. We're one of the few sites that give you noise testing by some really expensive and broad-ranged noise-testing equipment, and more importantly, card-only power-draw. Our reviews also grill graphics cards through 22 real-world tests across four resolutions, each, and offer price-performance graphs. When NVIDIA didn't send us a GeForce GTX TITAN-Z sample, we didn't care. We didn't make an announcement like this. At $2,999, it was just a terrible product and we never wished it was part of our graphs. Its competing R9 295X2 could be had under $700, and so it continues to top our performance charts.
The R9 Nano, on the other hand, has the potential for greatness. Never mind the compact board design and its SFF credentials. Pull out this ASIC, put it on a normal 20-25 cm PCB, price it around $350, and dual-slot cooling that can turn its fans off in idle, and AMD could have had a GM204-killing product. Sadly, there's no way for us to test that, either. We can't emulate an R9 Nano on an R9 Fury X. The Nano appears to have a unique power/temperature based throttling algorithm that we can't copy.
"Fiji" is a good piece of technology, but apparently, very little effort is being made to put it into the hands of as many people as possible (and by that we mean consumers). This is an incoherence between what AMD CEO stated at the "Fiji" unveil, and what her company is doing. It's also great disservice to the people who probably stayed up many nights to get the interposer design right, or sailing through uncharted territory with HBM. Oh well.
759 Comments on AMD Radeon R9 Nano Review by TPU...Not
Driver issues are nothing new. They're not unique to NVIDIA nor AMD. WDDM 2.0 exposed problems neither side was prepared for. NVIDIA just happened to be less prepared than AMD this time around.
It was hard for reviewers to get excited about Mantle when only a few games support it. The technology ending up in Direct3D 12 and Vulkan was AMD's doing. AMD could have milked that cash cow for a while but elected not to. Personally, I applaud AMD for it but that doesn't help AMD's bottom line.
And yes, I know how to spell. Welcome to Canada, Eh?
Go back through the past 5 years of motherboard reviews I've done, and see both AMD and NVidia cards used. I dropped AMD cards from my personal rigs because NVidia didn't have the same driver problems. I've seen far less problems on NVidia than AMD (and yes, I still use a 7970 daily, before you even go there). The 7970 is nothing but problematic, my 780 Ti's work great.
That's the thing, most driver problems are highly dependent on complete system configuration, and installed software. You can't exactly pin every problem in video rendering on the videocard. I still have issues with video rendering on the 7970, the same problems I've had for the past 3-4 years that I've had the card.
Ashes has nothing to do with Mantle, what it does show is problems Nvidia have with DX12 and the fact that despite all the noise about having better DX12 support than AMD they don't have Hardware ASync while AMD do. and that actually matters.
One argument is: given AMD own all the consoles which all have their GCN ASync hardware devs will use it.
Another argument is given that AMD only have a 20% PC market share (and falling) with Nvidia owning 80% devs are not going to bother.
This however, does.
Anyone who has even the faintest idea what Mantle is and does would not test it in the way 90% of reviewers did, the conclusion they came to was utterly inevitable with their testing methodology.
More straw man stuff. Nobody is going to fall for your misdirection.
Also, reviewer's won't kill AMD, their lies in PR and product availability will. I used to get AMD samples direct from AMD, they fired the staff that I dealt with, then never contacted me again, and getting in touch with the right people is damn near impossible.
Then they fired my contact, and getting a new one is nigh on impossible. See no AMD board reviews? Ever wonder why? It's because I won't buy chips to review them, and I shouldn't have to go to board makers to get one. AMD can afford the $25 to send me a CPU to have their boards covered here, but they won't. That's all it would cost them.. $25.
AMD pays a dude to sit on twitch and stream, and you claim they are broke? lol. They simply spend their marketing dollar the wrong way.
Anyway. i need to get to bed.
Not reviewers. It's their CEOs, and their marketing.
Nice try though. Still doesn't explain why they decided to not send me chips, and thereby minimizing their exposure...
AMD CPU reivews here on TPU, done by me:
www.techpowerup.com/reviews/AMD/FM2_APU_Preview/
www.techpowerup.com/reviews/AMD/FM2_APU_Review/
www.techpowerup.com/reviews/AMD/FX-8350_Piledriver_Review/
www.techpowerup.com/reviews/AMD/A10-6800K/
If I wanted to take your way of saying things... oh look at the date since I've done a review... oh, look, that time period matches when you say they have been having issues... maybe they should have kept ME happy? Maybe their lack of sending me samples fucked them over, and now they are paying the price?
NOw here again we have an issue of not sending samples... whose fault is that?
A..M...fukin..D.
BE that as it may, abundantcores has been trying to say that AMD isn't sending samples out because reviewers fucked them over, so much so that they are now bleeding millions. Reviewers made them fail.
I think not.
They don't send you a chip and they hope you go out and buy one - and they get a review. They're hoping for the sale and the review....probably the same strategy they're hoping for with Nah,No! It couldn't have anything to do with AMD having to service a $2bn debt that AMD's own BoD, including Hector Ruiz, managed to saddle the company with? The same debt servicing that's crushing the life out of their R&D budget?
It's a pity AMD's board seem less interested in the company than the people who rally behind it - still, there isn't much incentive to succeed I suppose, Lisa will pick up her $11.5 million compensation either way - and with a raft of built-in excuses and apologists, they always have mitigating circumstances : Nvidia, Intel, reviewers, Microsoft, GlobalFoundries, TSMC, game developers, the buying public...basically anyone except the people who saddled them with a debt burden that is grinding the company down to dust.
May I pose the question as such. Is AMD trying to market their brand in an entirely new way, or better yet are they competently marketing the brand?
I'd like to offer the following. I occasionally (upon release of new product lines) get e-mails from Nvidia about their latest and greatest cards. I never seemed to get those e-mails from AMD, until the 3xx series cards had a date. In the last three months I've been inundated with requests to follow AMD on Twitch, where they'd be giving away a brand new card after you were forced to watch somebody stream with it.
AMD has limited the amount of people who were offered Fury and Nano samples. You can argue this as AMD being paranoid, but it seems to be a stupid point. Nvidia have said that they won't release an HBM1 powered card, because of a severe limitation on the available stock and the memory limits. This is the company that proudly released a nearly 1600 USD card, that could be outperformed by an SLI setup with cards that were 1/2 the price and less. If you're willing to place that kind of albatross around your neck, yet still pass on HBM1, what does that say about the maturity and availability of HBM1?
Is AMD pricing themselves out of the game? It seems that way, but only because of incompetence of message. Nvidia can put the 970 (340ish USD), 980 (520ish USD), and Titan Z (1600ish USD) into a generation and "win." AMD on the other hand offers the 380 (200ish USD), 390 (320ish USD), and the Fury (650ish USD). Rather than focus on the high resolution benefits of the current product stack, or perhaps working with a B or C level studio to get a competent DX12 coded game onto the market to demonstrate their technological prowess, AMD are demonstrating the fact that they offer an excellent price to performance ratio (even if the performance per watt is not great). While that does get you something, the reality is that it doesn't translate to sales if you aren't marketing yourself as a value brand. AMD is still trying to be the king of performance, when they could do so much better this generation by selling themselves as the king of value, which will be reaped in the next several years as DX12 actually gains traction.
Is AMD actually focused on the high-end market? I have to ask the question, because it seems like everybody fails to do so. Nvidia has released a high end card with each of the last few generations, which does nothing for most people. Realistically, gaming on a console isn't as great, but when I could have two consoles for less than your single GPU, the cost to performance is pretty screwy. Likewise, the Nano demonstrates a rather unique focus on GPU size, not on massively increased performance. If, as was posed by the OP, Nano was stretched to Fury size and the pricing was reduced what would be on offer? A card that performs very well of the price per watt metric, demonstrates a new memory technology, and might well compete with the 980ti favorably on overclocking performance. Instead, what we receive is a lackluster Fury and a massively priced Nano. It really seems like AMD targeted two traditional market segments, fired everything, and forgot that there might be someone in-between.
In short, I think AMD marketing needs to be fundamentally reworked. There's such a fundamental disconnect between what they are selling, and how they are selling it, that I can't understand how their PR doesn't go home each evening and drink themselves to sleep. Social media pressure is great for low cost items. Traditional reviews are great for higher cost items (this is why Consumer Report and Car & Driver still exist). There's a fundamental stupidity demonstrated in trying to sell one with another though. You won't see Consumer Report passing judgement on deodorant, yet Old Spice has a better marketing engine through viral videos than the objective quality of their product. AMD is trying to be both a modern PR machine, and sell an expensive product. That kind of disconnect is hurting their bottom line more than objective silence on products ever could. Either support your expensive product with traditional reviews, or sell your cheap product with social media campaigns. What you are doing now is just painful to watch.