AMD Nano Blacklist Situation

[VulgarDisplay]

I don't think any of these sites should be given hardware for free. It automatically creates a situation where they feel beholden to the company providing them said free crap.

 

[Elixer]

Coil whine is sometimes very hard to predict. It is often not connected to your main frequencies (your GPU frequency) but to secondary periodic events (your GPU moving between power stages within a single frame for example). Those large load differences will result in a momentum on any coil in your system.
Now, if this creates an audible whine or not is influenced by the weight of said coil and the fixation on the board. Simply improving the fixation of a singing coil (preferrably with something that has no electric or magnetic influence, like hot glue) can be a permanent fix.

What is also interesting is, the same card can behave very differently on different systems.
If the PSU output is slightly higher or lower, that can change the harmonics as well.
This is why you can make the whine be out of the human range of hearing, by doing overclocking (or underclocking) and/or putting more/less volts in.

We know that AMD validates the GPU in question, but, we don't really know if AMD tests the boards themselves, or it leaves it up to their partners. I wonder if Ryan can get that info from his contacts?

 

[Attic]

One thing on the coil whine is that from what i've read and from direct experience it begins at very high frame-rates.

For me turning on vsync (which I always do) has been the easiest solution on cards that have coil whine.

 

[BFG10K]

One thing on the coil whine is that from what i've read and from direct experience it begins at very high frame-rates.

For me turning on vsync (which I always do) has been the easiest solution on cards that have coil whine.

These have been my findings too.

Almost every card will have coil whine if it generates a high enough framerate. Some are louder than others. Some need a higher FPS to cause it. But virtually every card I've ever used had it - from both camps.

Obviously a combination of the CPU + game + settings plays a role, but if you cap the framerate to a reasonable value, you should never have the problem.

Even this Titan of mine has coil whine when I uncap the framerate, but it's never there when I cap it. So if your card's squealing, ask yourself: do you really need a game menu to be rendered at 1000FPS?

Capping the framerate also saves power, makes the card cooler & quieter, and lets the card boost more often when it needs to.

 

[bononos]

All the above is BS.

1. AMD certainly does include reviewer guides with their review units. Do you need me to post one?
.......

This is from pg.1 in the thread, do you have a sample?
Ryan Smith replied later in the thread to someone else (I think) that he does not receive any.

 

[Phynaz]

This is from pg.1 in the thread, do you have a sample?
Ryan Smith replied later in the thread to someone else (I think) that he does not receive any.

"It was only after we saw AMD’s reviewer’s guide that we decided to go ahead and include it, because quite frankly we didn’t believe the numbers AMD had published"

http://www.anandtech.com/print/5261/amd-radeon-hd-7970-review

"With all these differences between Kabini and Bay Trail, including price, power and cores/threads, it is hard to find Intel parts that accurately match each other. AMD has put the following in front of reviewers to provide a guide:"

http://www.anandtech.com/print/8067...2-athlon-53505150-and-sempron-38502650-tested


Or you could try Google, here's the first hit;
http://videocardz.com/56728/amd-radeon-r9-fury-x-reviewers-guide

And then there's the famous mobile GPU that AMD shipped with an Intel system for benchmarking

In recent times the guides are less "demands" and more explanations.

 

[3DVagabond]

Well, just glanced briefly at TPU's Nano review and if anyone wonders why AMD gets annoyed with certain review sites it should be clear here.

1) We are back to having a retail card in the comparison along with a review sample. Something that is only reserved for AMD. Never have I seen TPU (or anyone else) do it with nVidia cards to see if retail samples boost the same as the review samples.

2) Alright we have another new metric thrown in as well. Another way of testing only AMD cards. It's called Clock frequency analysis. Obviously info we don't need for any other card.

3) And making it's return, especially for Nano we have Thermal camera and fan noise recording. It was used in the past but went away and has been resurrected special for Nano.

 

[ShintaiDK]

Press card outperforms retail card. Sounds like a rerun.

Oh, and coil whine again. Both cards.

 

[Phynaz]

Well, just glanced briefly at TPU's Nano review and if anyone wonders why AMD gets annoyed with certain review sites it should be clear here.

1) We are back to having a retail card in the comparison along with a review sample. Something that is only reserved for AMD. Never have I seen TPU (or anyone else) do it with nVidia cards to see if retail samples boost the same as the review samples.

2) Alright we have another new metric thrown in as well. Another way of testing only AMD cards. It's called Clock frequency analysis. Obviously info we don't need for any other card.

3) And making it's return, especially for Nano we have Thermal camera and fan noise recording. It was used in the past but went away and has been resurrected special for Nano.

More information isn't a bad thing, assuming the information is accurate.

 

[Stuka87]

More information isn't a bad thing, assuming the information is accurate.

True, but would be better if the same tests were run on everything, as opposed to picking and choosing.

 

[RussianSensation]

Press card outperforms retail card. Sounds like a rerun.

Oh, and coil whine again. Both cards.

Didn't you keep repeating for the least 4 years that AMD will never catch NV in perf/watt on a 28nm node, not with HBM or anything. I remember you are one of the most vocal members here repeating ad-nauseum how AMD is doomed in perf/watt and more so that AMD will never make a large die GPU (aka 550mm2+) or be able to improve performance much beyond 290X's on the same 28nm node? You kept claiming how the only way for AMD to achieve tangible increases in performance was with a shift to a new 14/16nm node and/or a new GPU architecture.

You also said that HBM is a complete waste of time and money as it provides no benefit at all over GDDR5 for either performance or perf/watt. You never acknowledged that the use of HBM1 allowed AMD to reduce the complexity of the GPU's memory controller and allocate that space to the GPU resources instead -- something that many people pointed out as a major benefit of HBM1 over GDDR5. You never acknowledged that the use of HBM1 allowed AMD to use the saved TDP from not using GDDR5 to increase the TDP/GPU clocks because normally the extra power usage of GDDR5 is in conflict with the GPU's TDP. You also said that we've had this level of performance in the Nano in this form factor (i.e., in the GTX970 mini/980 cards for the last 12 months). This is absolutely not true. I also remember you never believed that AMD's 2nd fastest card the Fury non-X (aka HD7950/R9 290 successor) would offer 87-89% of the performance of the Titan X for $550.

Boy, you were literally wrong on pretty much everything relating to HBM1 vs. GDDR5, AMD's Fury/FuryX/Nano, AMD's perf/watt, AMD's performance increase on 28nm, the performance of the Fury non-X relative to the Titan X. The only thing you ever got right was HBM1 being limited to 4GB but this is ultimately hardly relevant as neither a 980Ti nor a single Fury/Fury X/Nano is fast enough for 4K on its own. You even claimed how AMD's 30% faster than GTX970 mini performance was pure marketing imagination and BS.

So now instead of admitting how wrong 98% of your predictions have been on the Nano, Fury and Fury X, HBM1, you sit here discussing Nano's poor price/performance vs. cards like 970/980 or 980Ti, its coil whine, which is btw a common occurrence in many NV and AMD cards. Furthermore, correlating coil whine to actual choke quality has not been substantiated because very high quality Cooper Bussmann chokes used on Gigabyte's X99 boards, the same ones uses for servers because of their rock solid stability, can coil whine as well.

In TPU's BF3 test, the Nano is whipping the 980 by 33%.

In TPU's testing which uses 3 highly questionable tests in its averages (Project CARS, broken/incorrect Wolfenstein NWO bench, and WOW that hardly represents real world raids in the game), despite that the Nano still manages to match the 980Ti in perf/watt.

Interestingly enough, at TechSpot, the Nano actually supersedes a stock 980Ti in perf/watt:

"Forgetting about the Nano's tiny size and focusing squarely on price and performance, it's the $650 GTX 980 Ti that we're interested in. This comparison sees the Nano trail the GTX 980 Ti by a 13% margin at 1600p and 10% at 4K, though the Nano consumed 16% less power on average, so it looks to be the more efficient graphics card."

Any objective person can clearly see all the pros and cons of the Nano but the usual haters who constantly focus only on the negative aspects of the Nano just further keep digging their already tarnished reputation.

Really where AMD dropped the ball with the Nano is its price -- it's priced way too high. Had AMD dropped this card to $449, it would have completely wiped the floor with the 980.

It's actually remarkable that AMD matched NV's perf/watt on more or less the same underlying GPU architecture that underlines the nearly 4-year-old HD7970. It took NV Fermi, then Kepler and then Maxwell - 3 distinct GPU architectures to be able to actually beat AMD GCN in perf/watt but with HBM1 AMD caught up on the fundamentally re-balanced (8 ACEs) 2011 architecture.

So positive points for the future is that AMD has closed the perf/watt gap with NV and now they can focus on using the advantages of HBM2, GCN 2.0 and 16nm node shrink to stay competitive. For us PC gamers who desire competition, this can only be viewed as a great development after all the trash talking online by AMD haters who wish for nothing but AMD to disappear.

The beauty of the Nano is that it foreshadows the next generation of AMD/NV GPUs with HBM2. Sure, the Nano is horribly overpriced, no doubt about it, but it's a proof of concept, and it's indisputably the absolute fastest+smallest miniITX card in the world right now.

What happens when AMD/NV bring out even faster GPUs with HBM2 on 16nm in a 6-7 inch form factor? At the very least it's giving miniITX case manufacturers a lot to think about.

 

[RussianSensation]

Almost every card will have coil whine if it generates a high enough framerate. Some are louder than others. Some need a higher FPS to cause it. But virtually every card I've ever used had it - from both camps.

Obviously a combination of the CPU + game + settings plays a role, but if you cap the framerate to a reasonable value, you should never have the problem.

This is a very solid and fair post that describes the issue of coil whine very accurately because it's not vendor specific. Thank you for that! One can manufacture a solid core choke with the exact same components (i.e., quality) as a traditional choke. Whether a card has coil whine or not doesn't necessarily mean that there is a scientific correlation between the quality of the actual choke and the noise it emits.

We know that some GPUs can coil whine despite what appear to be good components on the PCB. For example, MSI Gaming GTX970 with Military Class 4 components, PowerColor PCS+ R9 290 with the Gold Power Kit components, Gigabyte GTX980Ti Gaming, etc. Many GPUs can coil whine but it doesn't necessarily mean that the quality of the chokes is poor, at least in my opinion. The Military Class 4 and Gold Power Kit on paper seem like much higher quality parts than what's used on typical reference cards.

However, it is disappointing for the GPU industry as a whole that $600+ graphics cards exhibit these issues regularly. I definitely see your point from a practical point of view - capping FPS in game menus - however, at the same time the coil whine issues date back many many years ago and I think some quality control should be exercised by the AIBs or NV/AMD in case of reference card component selection. I suppose, it's possible that if NV/AMD QC personnel received 10 newly produced cards before they go into full production that they wouldn't exhibit coil whine (i.e., statistically speaking it's possible to get 10 random cards with no coil whine) but perhaps they need to expand their sample size for testing purposes?

The coil whine issue has been somewhat consistent and it's disappointing to see that it continues in late 2015, especially on very expensive products. However, I don't personally agree that coil whine is directly correlated to the quality of the actual chokes. The presence of coil whine on cards like the Nano or Fury X is more detrimental for consumers and AMD. Firstly, because AMD spent engineering & financial resources on incorporating AIO CLC with a Scythe Gentle Typhoon, but what benefit does it have if the coil whine overshadows the quality of these parts? Secondly, with the Nano, they focused their design on a product for a specific purpose (Nano for small-form factor PCs <- many gamers who build these focus on quiet operation) but again if the Nano coil whines, then it's less likely that someone who wants a quiet small-form factor PC/HTPC would consider it.

I'd still like to see more evidence that shows a correlation between coil whine and actual choke quality before ascertaining a relationship.

 

[Enigmoid]

It's actually remarkable that AMD matched NV's perf/watt on more or less the same underlying GPU architecture that underlines the nearly 4-year-old HD7970. It took NV Fermi, then Kepler and then Maxwell - 3 distinct GPU architectures to be able to actually beat AMD GCN in perf/watt but with HBM1 AMD caught up on the fundamentally re-balanced (8 ACEs) 2011 architecture.

So positive points for the future is that AMD has closed the perf/watt gap with NV and now they can focus on using the advantages of HBM2, GCN 2.0 and 16nm node shrink to stay competitive. For us PC gamers who desire competition, this can only be viewed as a great development after all the trash talking online by AMD haters who wish for nothing but AMD to disappear.

Its a gpu with very aggressive power algorithms with a number of other caveats so saying that it beat maxwell is disingenuous at best. The fury nano underperforms the 980ti (cut down GM200) at approximately the same die size (perf/mm^2 takes a hit) using highly binned not cut down dies on a product optimized for perf/W above all else. The equivalent GM200 comparison would be a downclocked full GM200 with power optimizations (ie sort of like a mobile optimized card).

Your link shows the nano barely more power efficient than the R9 Fury. Nano uses about 9% less power than the R9 Fury for slightly less performance.

Why are you throwing Fermi in there? Fermi was never designed to compete with GCN. Double check your eyes and your charts again. Where are kepler and previous GCN cards on the chart you posted? Oh yeah, better than or equivalent to gcn in terms of perf/W.

Fury might have been an engineering milestone for AMD but from a financial perspective it really is looking like a complete disaster.

 

[Elixer]

It is not solved by any stretch of the imagination.

A repair tech told me that they started to use paraffin and hot glue, (and some other stuff that I don't recall what he called it) and dip the coils in it to mitigate the issue.
It does help, but it doesn't eliminate the issue.

He also went on to say that you can use the most expensive parts out there, and it can still whine, so there is no direct correlation between cost and having parts that whine, in that, you can have bottom of the barrel parts, and those don't whine, yet, you have a $1200 video card, and it whines.

There are tons of cards (heck, it isn't limited to video cards either) out there, from all camps that have whine, some are much more pronounced than others, but it is still there.

This is a very solid and fair post that describes the issue of coil whine very accurately because it's not vendor specific. Thank you for that! I am honestly disappointed that some people on this forum believe the first thing they read and actually started correlating coil whine with choke quality. One can manufacture a solid core choke with the exact same components (i.e., quality) as a traditional choke. Whether a card has coil whine or not doesn't necessarily mean that there is a scientific correlation between the quality of the actual choke and the noise it emits.
...
Therefore, we have to question the purpose of such propaganda (a member who is paid by a competitor to spread negativity? A person who is a delusional fan of the competitor?) and further actually question if there is any correlation between coil whine and quality.

Is coil whine annoying? It sure is but if a GPU coil whines does it mean it's made from lower-grade level of materials? If so, is there scientific data that backs this up?

Some people just don't like facts, they get in the way of their agenda.
What that agenda is, is open for interpenetration, but, it is clearly evident at how they keep posting the same stuff over and over, screw the facts.
If they repeat it often enough, some, uneducated people will believe it.
As for the TPU piece, well, at least they bought a card, but, having different 'tests' for different cards is laughable, as was already pointed out.

 

[rgallant]

now this would piss me off


http://forums.evga.com/Titan-X-Coil-wine-m2326262-p2.aspx post #37

same story with my Titan X ! I live in ?Aus, paid $1600 for it. and the noise drives me nuts, im going to get it watercooled, and im worried the noise will be more annoying

 

[Sabrewings]

now this would piss me off


http://forums.evga.com/Titan-X-Coil-wine-m2326262-p2.aspx post #37

same story with my Titan X ! I live in ?Aus, paid $1600 for it. and the noise drives me nuts, im going to get it watercooled, and im worried the noise will be more annoying

No coil whine here with a 980 Ti and the EK Titan X block with backplate.

 

[frozentundra123456]

Its a gpu with very aggressive power algorithms with a number of other caveats so saying that it beat maxwell is disingenuous at best. The fury nano underperforms the 980ti (cut down GM200) at approximately the same die size (perf/mm^2 takes a hit) using highly binned not cut down dies on a product optimized for perf/W above all else. The equivalent GM200 comparison would be a downclocked full GM200 with power optimizations (ie sort of like a mobile optimized card).

Your link shows the nano barely more power efficient than the R9 Fury. Nano uses about 9% less power than the R9 Fury for slightly less performance.

Why are you throwing Fermi in there? Fermi was never designed to compete with GCN. Double check your eyes and your charts again. Where are kepler and previous GCN cards on the chart you posted? Oh yeah, better than or equivalent to gcn in terms of perf/W.

Fury might have been an engineering milestone for AMD but from a financial perspective it really is looking like a complete disaster.

I agree. Nano is a very special card, highly binned and optimized for low power consumption. It is hardly a valid case to say AMD has matched nVidia in overall performance per watt. I guess if RS wants to prove somebody wrong badly enough, you can use it as a corner case, but it really is the exception rather than the rule.

 

[RussianSensation]

Its a gpu with very aggressive power algorithms with a number of other caveats so saying that it beat maxwell is disingenuous at best.

Does it matter how AMD achieved it? No one precludes NV from binning GTX980Ti chips and making a GTX980Ti mini Nano competitor. That's not the point of my previous post -- clearly some people ignored all possibility of AMD matching Maxwell in perf/watt on 28nm, regardless of HBM1, binning, you name it. Don't tell you that you missed 4 years of posts that claimed how AMD is done for in perf/watt and will never come back on 28nm?

The fury nano underperforms the 980ti (cut down GM200) at approximately the same die size (perf/mm^2 takes a hit) using highly binned not cut down dies on a product optimized for perf/W above all else. The equivalent GM200 comparison would be a downclocked full GM200 with power optimizations (ie sort of like a mobile optimized card).

1. NV currently doesn't sell such a product so your comparison while valid is not practical. Out of the box, the Nano matches 980Ti's perf/watt at TPU and beats it at TechSpot. I personally couldn't care less about perf/watt metric in a vacuum but given all the FUD I've been reading online for the last 3-4 years about how AMD will never match Maxwell in perf/watt has been proven wrong. Now where are all those people who kept making those statements? Oh right, they are now discussing Nano's poor price/performance and coil whine. Figures. They are only interested in hating anything about an AMD GPU, with 0 intentions of ever buying an AMD card in the first place. If the Nano beat the 980Ti in every metric and had HDMI 2.0, then the discussion would shift solely on 6GB >>>> 4GB. You want to bet on it?

2. For its intended function --- the smallest mini ITX cases, NV has no competitor to the Nano, at any price. That reason alone makes the comparison of the Nano to the 980Ti irrelevant. Why? Because anyone who wants top performance and overclocking is buying the 980Ti anyway. It seems people are having the hardest time in the world understanding that the type of gamers who are buying GTX980Ti and Nano do not cross-shop these 2 videocards. This should be common sense because even the Fury X >> than the Nano too.

Your link shows the nano barely more power efficient than the R9 Fury. Nano uses about 9% less power than the R9 Fury for slightly less performance.

My link shows the Nano tying GTX980Ti in perf/watt at TPU at 4K and beating 980Ti at TechSpot. I already explained above how perf/watt is something I don't care about in a vacuum but everyone else who trash talked that AMD will never match Maxwell in perf/watt on 28nm node and how HBM1 had no advantage over GDDR5 (aka AMD should have just made a larger 290X with GDDR5) has been proven wrong.

Why are you throwing Fermi in there? Fermi was never designed to compete with GCN.

But it did in reality. It took NV a whopping 2.5-9 months to roll-out the entire GT620->690 line-up of cards.

Did you forget how HD7750/7770/7850/7870 wiped the floor with Fermi cards for months before sub-GTX670/680 cards ever showed up?

Do you remember any objective PC gamer shouting how NV will never match GCN in perf/watt? Of course not, because they are not stupid. But clearly then when NV started leading in perf/watt, not only where most of the people who couldn't care less about it during GeForce 8, GTX200, GTX400, GTX500 eras started claiming how perf/watt was THE most important metric in the world, but they insisted that AMD will never catch NV in perf/watt going forward. :hmm:

Double check your eyes and your charts again. Where are kepler and previous GCN cards on the chart you posted? Oh yeah, better than or equivalent to gcn in terms of perf/W.

Oh what a surprise -- a crippled architecture, with crippled compute, horrible double precision compute, VRAM gimped is outperforming GCN in perf/watt.

That's awesome; and now let me know how Kepler's GTX670/680/770/780/780Ti are doing in modern games.

Let's try high textures at 1080P. Oh, what's that, it's impossible on cards with less than 3GB of VRAM.

Ultimately, Kepler's GTX600/700 desktop perf/watt was a marketing gimmick over HD7000/R9 290 series of cards. Over the 2-2.5 year period of GTX600/700 vs. HD7000/290 series, for the most part NV offered only 2 options: 670/680/770 2GB VRAM crippled products OR ovepriced $500-650 780 3GB, $700 780Ti and $1000 Titan. OMG, perf/watt FTW. That's a nice marketing bullet-point.

Fury might have been an engineering milestone for AMD but from a financial perspective it really is looking like a complete disaster.

Easy to be an arm-chair CEO. So let's you are AMD's management with limited financial resources that have to be allocated between their CPU, APU and GPU divisions. You also have limited time -- aka you cannot just produce a brand new GPU architecture that takes 3-4 years to design (NV has sighted on various occasion that's how long it takes for their teams to design a new architecture). What would you have done in AMD's shoes because there sure are A LOT of engineers / expert managers here who think they could have designed a far better GPU than the Fury with HBM1. I am ready to hear suggestions.

This reminds me of all the PS4 bashes and to this date not a single person online ever outlined an alternative engineering design that would be $399, not incur major financial losses to Sony/MS and yet be way more powerful than PS4. No product is perfect but what's the better alternative?

Furthermore, what makes you think AMD will not use what they have learned on HBM1 and Fury/FuryX/Nano for their future GPU designs? Do you honestly think AMD will scrap everything they have learned on the Fiji chip and HBM1?

What prevents AMD from shrinking Fiji to 16nm node, and using HBM2 and cards like the Fury/Fury X and Nano become next generation's mid-range $300-400 cards?

Also, when AMD basically wiped the floor with NV in price/performance for 5 consecutive generations (HD4000->R9 290) and AMD failed to gain market share overall or make substantial profits, any new CEO would try to a new strategy. Will the new strategy work? Maybe or maybe not but we know that the old AMD GPU strategies more or less failed long-term.

And BTW, what primarily gives the 980Ti the win is its overclocking performance. At stock speeds, the Fury X and 980Ti are very close in performance.

NV is primarily a graphics card/visuals company, unlike AMD whose business is split across CPUs, GPUs and APUs. NV's market cap is nearly 8.5X of AMD's and their focus on graphics allows NV to shift most of their R&D towards a specific industry segment (i.e., graphics). It would actually be mind-blowing and shocking if NV's best could not beat AMD's best graphics card. This is the same scenario we read in the CPU forum where people are bashing AMD because they can't somehow (somehow!? lol!) make a CPU as competitive as a company with 100X its market cap.

But the crazy part of all as I mentioned before, while the Nano is targeted at < 1% of the entire PC gaming market and hardly matters for the vast majority of PC gamers at its $650 price, there is a ton of HATE on the Nano. Counter to that, almost non-existent mention online from professional North American reviewers and PC gamers on how the GTX950 is an overpriced pile of garbage, 2GB VRAM gimped, how GTX960 2GB is DOA and how GTX960 4GB is yet another overpriced turd, the worst x60 series card from NV released in the last 5 generations <<===>> You know let us all focus on a card that <1% of PC gamers will actually buy and ignore the horrible products that are recommended online to most PC gamers who are shopping in the $150-250 GPU space.

This is almost like a political agenda/massive PR smear campaign tactic used to divert attention away from the masses from the REAL issues (how $160-200 GTX950/960/960 4GB are overpriced garbage) that are actually highly relevant for the vast majority and instead focus on a small niche product by blowing that issue way out of proportion (i.e., how overpriced the Nano is).

Sites like TechReport & HardOCP should be ashamed of themselves for their coverage of the GTX960 2GB and conclusions in their respective reviews on such a horrid card, but instead they are spending massive amounts of time focusing on a <1% niche card, after themselves admitting almost no one cares about it. How come these same sites aren't exploring the 2GB VRAM issues on the GTX960 2GB, or discussing the horrendous price/performance of GTX950/960/960 4GB in the $150-200 space?

Only TPU actually had the decency to state the truth about the current desktop discrete GPU market:

"NVIDIA simply cannot get the pricing of its sub-$300 lineup right and continues to offer nothing compelling until the $310 GeForce GTX 970. The company may yet make a ton of money with their mid-range line-up, but that's only because of its better sales-force. The Radeon R9 290 TurboDuo from PowerColor is a gem." ~ TechPowerup

It's no wonder AMD initially blacklisted TechReport and HardOCP - those sites have slowly started to betray the trust of PC gamers as a whole because imo and many other gamers have noted this, they both failed to provide consistently good advice to PC gamers in the most important GPU segments -- that is the $150-300 ones. Instead they praised overpriced products such as GTX750/750Ti over much faster R9 270/270X and the same for GTX960 over R9 280X 3GB/R9 290 or completely ignored the poor price/performance of the 980 against 290/290X/R9 295X2, or the horrible price/performance of the niche Titan/Titan X cards. Yet, now a $650 Nano makes no sense because it's an overpriced niche product as well? Well of course it doesn't make sense from a price/perf perspective, but why didn't they call out these other overpriced products on the same issue? Their inconsistency is what's at odds with their negativity towards the Nano. One would have to be completely clueless/tech illiterate or biased to give GTX960 2GB Silver/Gold awards for the last 9 months for gaming purposes. It's good the Nano came out because now we know what sites are actually consistent in their viewpoints.

---------

Back to topic -- I will address the blacklisting situation because I've seen a lot of defending here of TechReport and HardOCP. Here is my take on TechReport, I could write an essay 10x bigger if I started digging deeper into their reviews.

There it is again - TechReport recommending GTX750Ti 2GB and GTX950 2GB in the $160 and below segment, fully ignoring the 2GB VRAM limitations. Then, for the Sweet Spot, TR recommends GTX960 2GB, once again ignoring its 2GB VRAM limits, and its horrid price/performance. Then, more shocking:

"While the 280X is the fastest of the three by a smidgen in raw FPS terms, it's based on older hardware that lacks support for FreeSync and AMD's TrueAudio DSP. If you really want an AMD card, we think you'll be better off with the R9 380 over the long term."

^ TrueAudio is a useless feature while anyone who can afford a FreeSync monitor could easily spend extra from a $200 R9 380 towards a $240 R9 290. Once again, TR completely ignores the key benefits of R9 280X over 380 having 3GB of VRAM and more horsepower.

In TR's Asus Z170-A review, they never mentioned the budget audio solution on that motherboard, never discussed how if you use the Ultra M.2 slot, that the SATA Express gets disabled, never discussed how motherboards in ate $135, $150, $170 price range from Asrock or Gigabyte wipe the floor with the Z170-A on features and component quality.

Then they went on to recommend the most absurd CPU for a budget gamer - the overpriced i3-6320, fully ignoring that a budget gamer is better off saving $40 and buying the i3-6100 OR spending just $30 more for a quad-core i5-6400. I mean seriously, how can a professional review site make mistakes of this magnitude?

Then there is this -- the most gross FAIL from a professional review site of all time:

"If you can't swallow the Core i7-5930K's cost but still want six Haswell cores in your system, we conditionally recommend the Core i7-5820K. This chip has 12 of its PCIe lanes lopped off, for a total of 28. We think Intel's decision to cripple this processor in this fashion is unfortunate, because it removes one of the key advantages of "extreme" processors based on the X99 platform. Many folks who build systems based on these CPUs will want 16 lanes going to two different PCIe x16 slots for multi-GPU configs. With a 5820K installed, though, an X99 system can't deliver. It effectively has no more PCIe bandwidth for SLI and CrossFire than a quad-core Skylake chip based on the much more affordable Z170 platform."

Not only does this "professional reviewer" fail to mention that the performance hit between PCIe 3.0 x8 and x16 is basically non-existent because the types of gamers buying X99 with dual GPUs will be running MSAA and high resolutions, but he didn't even acknowledge that there are X99 motherboards for $225 that allow 5820K to run Tri-Fire / Tri-SLI in 8x/8x/8x configuration. Neither did he discuss the long-term benefit of 6-cores vs. 4-cores and that 5820K OC performs exceptionally well against i7 6700K OC.

I think a lot of PC builders forgot what a good reviewer/review site is. You have to be consistently good in reviewing various PC hardware to be respected. It's not someone who tries to sell you the product and hide the downsides at all costs for the most popular brands (Asus, NV, etc.), but someone who genuinely provides great advice to PC builders, and not because Asus or NV sent you more free product samples. How some of these review sites became popular is shocking. They are totally incompetent and this incompetence extends not only to CPU but to reviews of other products as well. Since for most of us PC building is just a hobby, for a professional review site that makes $ from publishing reviews, I find such gross mistakes and omission of information unacceptable.

 

[tential]

I don't get why so many people are posting about this card like they ACTUALLY cared about it.

People cared about this card because they thought it MIGHT be cheap. And I use MIGHT heavily. They thought because of hte comparisons to the GTX 970 and 980, that it might be a great performance for the price (Which touches on why AMD is just a poorly run company for thinking this tactic of comparing it to the GTX 970/980, then pricing it against a 980Ti was a smart decision.)

The vast majority of people talking about this card have NO interest in the mini ITX system builds it will be used in.

I will buy the Arctic Islands version of this card no question or the WC version depending on the M-ITX cases out there. ESPECIALLY now that the HEDT platform has MITX options.

Yes, many people here have the traditional desktop gaming rig. Me, and for the people out there like me, we have setups build for entertaining, and my rig is just there. It's only as big as it is, because I thought it had to be. The N1 case that came out recently, the Nano, X99 MITX board, all of the pieces are coming together now. I don't need a big gaming rig. I can get something 1/5th the size now, and still get the same performance. So that's what I'll get.

This product wasn't for everyone, so why people who have no interest in the builds thiss is being used in are attempting to bash this product, and the reviews associated with it (From both sides of the field) is beyond me. Move on to something you care about.

I wish Nvidia would release an alternative and support Gsync, so that I'd have options, but I'm happy with AN OPTION, rather than nothing.

There will always be people at the high end who want something, have no viable alternative, and will pick it.

Personally, I like the Nano.
It's just not time in my upgrade cycle to pick up such a card. And AMD handled the release too poorly for me to give them money. I love business and running a business, you have to do a better job than that. I'm too much of a capitalist to give them money for handling it in that manner but that's another story.

 

[Enigmoid]

Does it matter how AMD achieved it? No one precludes NV from binning GTX980Ti chips and making a GTX980Ti mini Nano competitor. That's not the point of my previous post -- clearly some people ignored all possibility of AMD matching Maxwell in perf/watt on 28nm, regardless of HBM1, binning, you name it. Don't tell you that you missed 4 years of posts that claimed how AMD is done for in perf/watt and will never come back on 28nm?

You are talking about architecture. My point is that the products are not designed as to make architectural comparisons based solely on performance and power without looking at other factors.

1. NV currently doesn't sell such a product so your comparison while valid is not practical. Out of the box, the Nano matches 980Ti's perf/watt at TPU and beats it at TechSpot. I personally couldn't care less about perf/watt metric in a vacuum but given all the FUD I've been reading online for the last 3-4 years about how AMD will never match Maxwell in perf/watt has been proven wrong. Now where are all those people who kept making those statements? Oh right, they are now discussing Nano's poor price/performance and coil whine. Figures. They are only interested in hating anything about an AMD GPU, with 0 intentions of ever buying an AMD card in the first place. If the Nano beat the 980Ti in every metric and had HDMI 2.0, then the discussion would shift solely on 6GB >>>> 4GB. You want to bet on it?

Which is fine as a practicality check. But insufficient to make claims about the architecture.

2. For its intended function --- the smallest mini ITX cases, NV has no competitor to the Nano, at any price. That reason alone makes the comparison of the Nano to the 980Ti irrelevant. Why? Because anyone who wants top performance and overclocking is buying the 980Ti anyway. It seems people are having the hardest time in the world understanding that the type of gamers who are buying GTX980Ti and Nano do not cross-shop these 2 videocards. This should be common sense because even the Fury X >> than the Nano too.

There are plenty of competitors. They simply dont have the same performance. You try telling people looking at the nano that a 970 mini with 1.6x the perf/$ isn't a valid competitor for most users (unless money is no objection).

My link shows the Nano tying GTX980Ti in perf/watt at TPU at 4K and beating 980Ti at TechSpot. I already explained above how perf/watt is something I don't care about in a vacuum but everyone else who trash talked that AMD will never match Maxwell in perf/watt on 28nm node and how HBM1 had no advantage over GDDR5 (aka AMD should have just made a larger 290X with GDDR5) has been proven wrong.

Their product matches yes. Their architecture may or may not but this hasn't been tested on a level playing field.

But it did in reality. It took NV a whopping 2.5-9 months to roll-out the entire GT620->690 line-up of cards.

Did you forget how HD7750/7770/7850/7870 wiped the floor with Fermi cards for months before sub-GTX670/680 cards ever showed up?

Which has absolutely nothing to do with the statement that it took nvidia 3 gens to match gcn. That statement is flat out false. Fermi was never the direct competitor to GCN (it competed only because of delays) and kepler had better efficiency.

Oh what a surprise -- a crippled architecture, with crippled compute, horrible double precision compute, VRAM gimped is outperforming GCN in perf/watt.

This is a surprise. You would think it would be the other way around.

That's awesome; and now let me know how Kepler's GTX670/680/770/780/780Ti are doing in modern games.

This game is a POS and invalidates any argument that is being made with it.

Let's try high textures at 1080P. Oh, what's that, it's impossible on cards with less than 3GB of VRAM.

I know. Look at that 6GB titan jump ahead...or not.

Easy to be an arm-chair CEO. So let's you are AMD's management with limited financial resources that have to be allocated between their CPU, APU and GPU divisions. You also have limited time -- aka you cannot just produce a brand new GPU architecture that takes 3-4 years to design (NV has sighted on various occasion that's how long it takes for their teams to design a new architecture). What would you have done in AMD's shoes because there sure are A LOT of engineers / expert managers here who think they could have designed a far better GPU than the Fury with HBM1. I am ready to hear suggestions.

You dont have to be a genius to see that AMD's board of directors is incompetent.

Also judging by how poorly the fury X does against Hawaii an improved Hawaii with gcn 1.2, colour compression, an improved front end, and something like ~3500 shaders would have done quite well. Keep DP so you can subsidize off the professional market as well. Might use a little more power but margins could be significantly higher.

Furthermore, what makes you think AMD will not use what they have learned on HBM1 and Fury/FuryX/Nano for their future GPU designs? Do you honestly think AMD will scrap everything they have learned on the Fiji chip and HBM1?

What prevents AMD from shrinking Fiji to 16nm node, and using HBM2 and cards like the Fury/Fury X and Nano become next generation's mid-range $300-400 cards?

AMD doesn't need to mass produce HBM GPU's to gain experience. They can do all that R&D in the lab cheaply (perhaps not quite as well but close) without incurring the costs of production.

<snip>

While I think you have a point the 960 and its recommendations have nothing to do with the nano.

 

[Elixer]

This product wasn't for everyone, so why people who have no interest in the builds thiss is being used in are attempting to bash this product, and the reviews associated with it (From both sides of the field) is beyond me. Move on to something you care about.

You don't have to want this product to point out the inconsistencies of the 'reviewers' based on what product they are "reviewing", and how they change their tune depending on if it is AMD or nvidia.

I wish Nvidia would release an alternative and support Gsync, so that I'd have options, but I'm happy with AN OPTION, rather than nothing.

I assume you meant freesync?

Personally, I like the Nano.
It's just not time in my upgrade cycle to pick up such a card. And AMD handled the release too poorly for me to give them money. I love business and running a business, you have to do a better job than that. I'm too much of a capitalist to give them money for handling it in that manner but that's another story.

Nano looked to be the best thing from the announcement of the Fury line, and, it could have been that, but, instead, it is reserved for a specific crowd, and thus, isn't a mainstream product that we were all hoping for.

AMD may very well be making lots of profit off of these launches, which is good for them, but, as a mainstream consumer, these cards just don't cut it.
Which is why the 380X could very well be THE card people have been waiting for.

 

[RussianSensation]

I don't get why so many people are posting about this card like they ACTUALLY cared about it.

Bingo. This is the best summary from any review I've read on the Nano that tells a prospective buyer everything they need to know really:

Performance data:

"Looking at some performance statistics we found the R9 Nano to be 5.6, 8.7 and 12.1 percent slower than the R9 Fury X at 1080p, 1440p and 4K average frame rates, respectively. The Asus GeForce GTX 970 Mini is probably the closest competitor to the R9 Nano and our testing shows the R9 Nano is 14.9, 24.7 and 31.1 percent faster at 1080p, 1440p and 4K average frame rates."


Conclusion:
"In terms of being the fastest compact graphics card AMD's R9 Nano succeeds in its ambition with ease; an ultra-compact ITX 1440p or 4K gaming system is a very real possibility, though it should also be noted than many mini-ITX cases of late are easily capable of housing full length graphics cards anyway. The price could also be a sticking point for many prospective buyers given that the R9 Nano offers around 25 percent more performance than compact Nvidia GTX 970s yet costs 85 percent more ($649 versus $349). This isn't a totally unexpected situation for the high-end market where diminishing returns are a de facto expectation. AMD's R9 Nano is a graphics card that serves its (very) niche high-end small form factor market well but its premium pricing makes it unaffordable to many small form factor system builders, and the coil whine could also prove to be another sticking point. In fact, it is this that holds the R9 Nano back from attaining our Premium Grade award. Were it not for this, it would be well deserved."
http://www.bit-tech.net/hardware/graphics/2015/09/10/amd-radeon-r9-nano-review/12

What a great review
Pros
Cons
Conclusion

TPU is also straight to the point:

"Last but not least: pricing. With $649, the R9 Nano is expensive, really expensive. The only card with a worse price/performance ratio is NVIDIA's GTX Titan X. AMD's R9 Fury X comes at the same $649 price level, which means that if you can fit the R9 Fury X into your case and have ruled out the GTX 980 Ti, the Fury X should be the card to buy. The R9 Nano really only makes sense for a small form-factor build where money doesn't matter. Here, it's the only option on the market with nothing in sight from NVIDIA because they don't have HBM memory, which allowed for such a compact design in the first place. If you plan on using 1080p resolution only, cards like the Gigabyte GTX 970 Mini and various compact GTX 960s will suffice, at much lower pricing. But for 1440p and beyond, the R9 Nano is the only choice, and it's a little technology marvel that completely owns it own niche."

Nano looked to be the best thing from the announcement of the Fury line, and, it could have been that, but, instead, it is reserved for a specific crowd, and thus, isn't a mainstream product that we were all hoping for.

I actually never understood the initial hype. 175W TDP, 8-pin power connector (i.e., maximum power usage of 225W) and small heatsink suggested it wouldn't be an overclocking monster. Then we saw that Fury/Fury X were poor overclockers to begin with which dampened any possibility of the Nano somehow getting 1200-1400mhz overclocks. I was personally very surprised that the Nano was $649 instead of $449 because it implies besides the R9 390, AMD's cards above that have very poor price/performance. However, this holds true for NV too because GTX980 is also a horrid price/performance card. That basically left the $300 battleground with 290X/970/390 and $650 980Ti. At $450 the Nano would have found its customers but it would have completely cannibalized the sales of the Fury and Fury X. I think what AMD should have done is re-balanced the entire pricing of its Fiji cards. For example:

Fury = $469-479
Nano = $499 (price premium for form factor)
Fury X = $579-599 (cheaper than the 980Ti because it's slower but still a premium over the Nano)

Unfortunately, it looks like AMD cannot afford these prices at all given their financial state and they have chosen profits over market share.

It's only getting worse because 980Ti can be found for $610 and yet AMD is not lowering prices on any of its Fiji cards. That is 100% proof that AMD is prioritizing profits per unit over market share and doesn't want the 'budget' brand image associated with AMD's cards over so many years of price/performance. I do think that this generation they went way too aggressive with price increases and since they don't have the performance crown, there is no halo effect either. So I can't see this strategy working that well but we'll see in their Q3 and Q4 results.

AMD may very well be making lots of profit off of these launches, which is good for them, but, as a mainstream consumer, these cards just don't cut it.

Even if Fury X was $500, it's would still be a questionable buy. An after-market 980Ti is 22% faster out of the box at 1440P, but also comes with bonus 50% more VRAM. But then the 980Ti can be overclocked even further which extends the lead over a stock Fury X to 29% in some games. Realistically, that means even at $500, the Fury X is still too expensive, isn't it?

Seems like AMD really didn't anticipate for after-market 980Ti cards to be priced at $650 and to be as fast as they are. AMD simply cannot compete in the $500+ space this generation so it's somewhat strategic that they are focused more of their attention on R9 390/390X cards.

Which is why the 380X could very well be THE card people have been waiting for.

At least in the US, it's still possible to purchase an after-market R9 290 for $240. As long as that card is still for sale, a 380X doesn't seem compelling. AMD should have launched the 380X as the 285X in September of last year. They also should have launched R9 390/390X in January of 2015 alongside R9 290/290X and just have all these chips duke it out.

 

[3DVagabond]

More information isn't a bad thing, assuming the information is accurate.

You're not getting it. Why don't we see retail samples of nVidia cards to see if they boost like the review cards (and anyone who thinks a 1% difference over 22 games proves anything needs to go to school for statistical analysis)? Why don't we see TPU give us frequency charts of cards that we know the clocks reduce over time like Titan, 780/ti, Titan-X, 980/ti if they are solely interested in giving us more accurate info? They changed up their review routine purely trying to get a gotcha on Nano.

True, but would be better if the same tests were run on everything, as opposed to picking and choosing.

You are getting it. If you are going to compare products in what is basically a competition you use the same playing field.

 

[BFG10K]

Also, when AMD basically wiped the floor with NV in price/performance for 5 consecutive generations (HD4000->R9 290) and AMD failed to gain market share overall or make substantial profits, any new CEO would try to a new strategy. Will the new strategy work? Maybe or maybe not but we know that the old AMD GPU strategies more or less failed long-term.

Agreed. People hating the $650 price need to ask themselves: what's the alternative? Price it lower and continue the epic fail bang-for-back strategy? What will that achieve? How will that help AMD?

It's better for AMD to sell out a low-volume part for $650 rather than $400. As Apple has shown in the smartphone market, marketshare is virtually irrelevant. All that matters is profit.

Again, that might not be good for customers, but Apple/nVidia/Intel have historically engaged in many anti-customer practices. But these companies are profitable while AMD is not. Customers still open their wallets so they tacitly accept being treated this way, despite complaining on the forums.

 

[Elixer]

Even if Fury X was $500, it's would still be a questionable buy. An after-market 980Ti is 22% faster out of the box at 1440P, but also comes with bonus 50% more VRAM. But then the 980Ti can be overclocked even further which extends the lead over a stock Fury X to 29% in some games. Realistically, that means even at $500, the Fury X is still too expensive, isn't it?

The fly in the ointment is, these cards might just be DX12 titans, and once they have the product out, they very well can't up the price since they would be faster than the 980 Ti's then, but, this very well might be another "overclockers dream" type of thing. Hard to prove something when there isn't anything out yet that uses DX12. On paper, it seems feasible that their GCN architecture is better for DX12.

The more realistic picture is, they just don't have the margins to price it any lower.
They are most likely paying a premium for HBM, then add the cost of UMC & TSMC to produce everything they need turns this into one expensive piece of hardware.

At least in the US, it's still possible to purchase an after-market R9 290 for $240. As long as that card is still for sale, a 380X doesn't seem compelling. AMD should have launched the 380X as the 285X in September of last year. They also should have launched R9 390/390X in January of 2015 alongside R9 290/290X and just have all these chips duke it out.

Sure would love to be a fly on the wall to hear why the full tonga never came to be before... It had GCN 1.2 as well, so, something just doesn't sit well with why they never went full steam with tonga.
All the other 3xx parts also use up AMD's allocation from TSMC, when they should have been pushing full Fury now.