AMD Radeon R9 Fury X Power And Pump Analysis

[FormatC]

While AMD introduced its highly anticipated Radeon R9 Fury X in June, we’re taking a deeper look at its power consumption and pump noise.

AMD Radeon R9 Fury X Power And Pump Analysis : Read more

 

[Cryio]

I'm seriously curious to see what 2015's Omega Driver does to Fury X's efficiency and overall performance.

 

[coolcole01]

Way to go amd 650 for less performance and manufacturing defects bravo.

 

[thor220]

More pages are spent on the whine then the actual performance of the card. It's an issue that's already been addressed, tom's really didn't have to dedicate so much to such a small issue.

Comparatively, at least it doesn't reduce performance like the last 0.5 GB on the 970, and I don't remember tom's being on Nvidia's case about that much more serious issue.

 

[i7Baby]

The ISO standard is more about making sure you settle on a quality standard and stick to it rather than improving quality. You can get certification by having a system in place that churns out crap - systematically.

Somehow I don't think a lot of PC stuff is rigorously specified. In this case I think samples were signed off by sales and marketing. This is how much of the PC industry is run.

 

[crisan_tiberiu]

With AMD, many users are going to be using first time in their PC a liquid cooler, so many wont have previous experiences to compare and they just gonna say that the sound is normal. These sound issues i am sure that they were known issues i am sure of it. Still, i would buy the Fury, i am sure that the tiny pump on the card has less than 0 chance on beating in noise levels my aquarium pumps that sit next to my desktop PC (i have 1 air pump and 1 filter motor..., so yeah i know what loud means)

 

[hannibal]

http://forums.anandtech.com/showpost.php?p=37527347&postcount=43

They are really changing the pump. Someone tried to upper their margins by cutting quality. But all in all even the older pump is very quiet and with the right version situation gets even better.
The FuryX is very good card. Luckily so is 980ti even in bigger measurements! Interesting to see how air coolet Fury will change the situation. All in all 980ti will make FuryX cheaper and FuryX forced the Nvidia to make something else than TitanX to customers. Competition is good!
980ti is very good card and FuryX allmost gets there. Because of AMD problems with DX11 the situation may be reversed in DX12 games so in longer run the situation is even as it should be for the customers.
Hopefully we will soon see some retake when Toms will get the upgraded pump to the test. Also it seem that some parts that are not water cooled can get quite hot, so Air cooled basic Fury will be very interesting card to be get tested.
The situation in GPU part is better than for long times! The next year will be even more interesting, because we may see first finvet based GPUs and also second generation of HBM.

 

[FormatC]

More pages are spent on the whine then the actual performance of the card. It's an issue that's already been addressed, tom's really didn't have to dedicate so much to such a small issue.
Comparatively, at least it doesn't reduce performance like the last 0.5 GB on the 970, and I don't remember tom's being on Nvidia's case about that much more serious issue.

This is a follow-up, not the launch review. In the first part of this follow-up with power consumption measuring I have tested both cards in a lot of games and applications - together with the specific power draw for each card in each benchmark. This is more than other sites published. 10,240,000 single values - this wasn't done in a few hours.

Simply use the slider pics to compare power consumtion, performance and watts/fps.

I don't remember tom's being on Nvidia's case about that much more serious issue

This is only one example:
http://www.tomshardware.com/news/MSI-GTX-660-670-overvolting-PowerEdition,18013.html
(the original, detailled review was in German)

 

[synphul]

It's unfortunate that they continue to slap the cheapy aio pump combos on everything. Rather than improving it sounds like quality just keeps dropping. Stating that the fan is so noisy it will probably drown out the annoying budget pump isn't really much consolation. In addition to performance charts, notations like these are important in a review. If a product has undesirable quirks it would be much nicer to know before dropping close to $700 on it. If this was a $200-300 card it might be more tolerable, not at those prices though. People paying top dollar should receive top quality. I'd rather they charge $10 more and do it right. Especially considering aio's have been around long enough they're no longer in their quirky 'infant' stage.

 

[Mousemonkey]

http://forums.anandtech.com/showpost.php?p=37527347&postcount=43

They are really changing the pump. Someone tried to upper their margins by cutting quality. But all in all even the older pump is very quiet and with the right version situation gets even better.
The FuryX is very good card. Luckily so is 980ti even in bigger measurements! Interesting to see how air coolet Fury will change the situation. All in all 980ti will make FuryX cheaper and FuryX forced the Nvidia to make something else than TitanX to customers. Competition is good!
980ti is very good card and FuryX allmost gets there. Because of AMD problems with DX11 the situation may be reversed in DX12 games so in longer run the situation is even as it should be for the customers.
Hopefully we will soon see some retake when Toms will get the upgraded pump to the test. Also it seem that some parts that are not water cooled can get quite hot, so Air cooled basic Fury will be very interesting card to be get tested.
The situation in GPU part is better than for long times! The next year will be even more interesting, because we may see first finvet based GPUs and also second generation of HBM.

Interesting that you see AMD somehow forcing Nvidia's hand, maybe Nvidia is just doing it's own thing and AMD are the one's playing catch up and having to release untested products in order to try and stay in the game. The GDDR5 bit is amusing enough as it indicates a design change that could explain why there was mention of the cards RAM using a fair bit of juice, it's also good to know that it can use Mantle but I didn't see mention of Freesync?

 

[Nuckles_56]

My question is why Total War: Attila was causing such a spike in watts/fps on the 980ti?

 

[FormatC]

My question is why Total War: Attila was causing such a spike in watts/fps on the 980ti?

The power consumption is more or less ok, but the performance not. It's not a Nvidia optimized game

 

[army_ant7]

Is it just me or have the individual performance and efficiency conclusions basically change here vs. the launch article?

Did those testing method changes listed at the beginning really have that much of an effect or was it a difference in software and/or settings used to benchmark?

I'd like your insight on this Igor, and I'll have to do an article comparison myself.

 

[FormatC]

..Did those testing method changes listed at the beginning really have that much of an effect or was it a difference in software and/or settings used to benchmark?
I'd like your insight on this Igor, and I'll have to do an article comparison myself.

I've measured since years in FHD because the most gamers prefers this resolution. But especially the Fury X is very limited in FHD. It was the goal to find now the best and more plausible method and game selection to show the difference in various situations. In UHD you get higher numbers and I will use Thief in UHD as reference in the near future for the not so important custom cards.

This review shows only a small part of all 21 tested games and apps. I've selected the most representative for this suite and I will use is this for all upcoming launches. Only one game is not enough to give a fair and objective description / conclusion. The range is too large.

 

[AS118]

Well this is all well and good, but I'm more interested in the air-cooled Fury and the Fury Nano myself.

I feel like the Nano, with its Fiji / HBM architecture with small size and high power efficiency may be the breakout product of this generation. As for the air-Fury, if it has 980Ti levels of performance at $100 cheaper, that'd be quite interesting, and it'd be easier to Crossfire as well.

Anyhow, I'm done giving Nvidia my money when it has such a big sales lead and there's only 2 GPU manufacturers on the market. It's not like it is with cars or fridges where there's lots of competition. I want AMD to not only stay in business, but have enough money to give Nvidia and Intel a run for their money.

So that means I'll keep buying AMD to make sure they have funds for R&D. Their products are good value and perform well enough (within spitting distance of Nvidia) so it's all good to me.

 

[random stalker]

From our point of view, Cooler Master and the OEMs it hired are responsible for this failure. Norms need to be adhered to and executed, especially if you proudly display certificates. We’re sure that AMD will get its money back from these companies, since these kinds of things are always subject to air-tight contract clauses in this business.

That is not how it works. From quality management point of view: "outsourcing a process/manufacturer does not relieve a subject of the responsibility of ensuring a product/process meets all required standards as specified by some sort or an internal engineering norm/drawing, etc..."
So it is entirely AMDs fault - outsourcing the pump to CM doesn't make AMD magically immune to all CM may cause.
Thus there are now two options:
- AMD needs better quality control so they wouldn't allow this kind of card on the market,
- AMD knew about it and released the bad card anyhow.

The ISO standard is more about making sure you settle on a quality standard and stick to it rather than improving quality. You can get certification by having a system in place that churns out crap - systematically.

Somehow I don't think a lot of PC stuff is rigorously specified. In this case I think samples were signed off by sales and marketing. This is how much of the PC industry is run.

It is usually the quality department and the R&D that approve the samples; but yes, in a broad sense it works like that.

 

[FormatC]

The OEM business is a bitch. You get a few samples from your OEM and all looks well. You start the mass production and the first batch is around 1000 pieces. This is and sounds not so much, but it is all what you need to start into a disaster without a final control.

 

[random stalker]

The batch size always depends on a part. Basically if you want me to make you some stuff, you need to give me all the norms and all the specifications you deem to be important, so I can set up my production and quality control.

After that, there is a part approval process, which is a hardcore version of the following:
- I make some paperwork that everything is OK on my side.
- I make initial samples, send them to you, you take them apart and give me green, yellow or red light.
- I set up the production tools and start making beta samples.
- Then you send your people to audit my processes - and you will send the most fierce evil people you can find so they find any nonconformity even the smallest one.
- I ship you another batch or two/three/four, you set up your processes based on this second/third/fourth... batch. During this series I set up additional and tune my production equipment.
- After that comes the 2 day test production, where I demonstrate the capability of manufacturing required amount of part of the required quality. You also send some of your people to observe that I don't cheat.
- Then, after a while, you audit me once again - to check that I've fixed everything.
- After that I send you some parts with increased quality control both yours and mine.
- And after a while (which can be several tens of thousand parts) you allow me to remove some obsolete quality control measures.

This 'dance' can usually take about an year, but also can be shortened (if we already collaborated on a same product), but is seldom shorter than half a year (well, I've seen a production for a new product being set up under three months, audits, paperwork, machines and everything, but those guys were crazy...)
So, there is a very small chance that AMD didn't noticed those problems before.

 

[eklipz330]

More pages are spent on the whine then the actual performance of the card. It's an issue that's already been addressed, tom's really didn't have to dedicate so much to such a small issue.

Comparatively, at least it doesn't reduce performance like the last 0.5 GB on the 970, and I don't remember tom's being on Nvidia's case about that much more serious issue.

if that's how you perceive things sure. this just shows how much of a fanboy you are towards AMD. they've spent a total of one article on the noise, and they've already done a full blown review.

and yes, they did address the .5gb issue on the gtx 970, in lengthy detail. a simple search would have saved you(and everyone that gave you a TU) from looking like a tool:

http://

http://

http://

it's people like you who give TH a bad name.

 

[FormatC]

@random stalker:
I'm very often in Asia because I'm not only an editor. The main problem is the typical man-in-the middle story:
AVC -> Cooler Master -> Sapphire -> AMD. Competence splitting. And no one feels responsible. But I see Cooler Master as guilty.

If you see all things working well, from mockup to the pre-production samples / small series, it is common to give the green light for the mass production. Cooler Master must control AVC, Sapphire must control Cooler Master, AMD must control Sapphire. I'm sure that AMD saw the first real samples only after shipping to branch offices and distributors. And exactly this was too late. The whole launch was a disaster.

 

[random stalker]

Agreed. But the one, that is responsible for the shipping of a NOK product is always the end of the line manufacturer.
Thus, he is responsible to make sure, everything, that is sent to the consumers, meets all the required criteria (and there is a lot of them, not only functional, qualitative, regulatory...). So without the final OK from the AMD, no AMD cards can leave the factory.

But here is the catch - there are various levels of nonconformity/problems/faults (name it as you like):
-minor - general annoyance, which some sensitive people may see as a fault (f.e. logo is dark red and not bright red)
-major - component may not work correctly and/or regular customers may rma the part (aka rma if a customer finds out)
-critical - component will not work and/or can be hazardous to health without warning (100% rma and/or a lawsuit impending)

Based on the level of a problem, there is a general guideline of what to do with such a part. The first and most important part, when handling major and critical problems, is to ensure that a part with said defective part can not be used in production (and especially not be shipped to customer). Which means not only sorting out the stock, but also marking all the OK parts and first (or first few) batch after said sorting action.

Based on the AMD reaction, it seems, AMD sees such problem as 'only an annoyance'.
And it wouldn't be the only time a faulty product was released, because the company releasing it was in a bind

 

[FormatC]

Agreed. But the one, that is responsible for the shipping of a NOK product is always the end of the line manufacturer.
Thus, he is responsible to make sure, everything, that is sent to the consumers, meets all the required criteria (and there is a lot of them, not only functional, qualitative, regulatory...)

This cards were made by Sapphire, not AMD. This is a typical outsourcing and I'm sure that all was shipped from PC Partner, not AMD

AMDs fault is at the end the missed final control in Asia. May be it was too hot in China for the pinstripe yuppies

 

[random stalker]

Oh sorry, must overlooked that in the article. Thought the AMD sells the cards under their own brand
If Sapphire makes them and sells them under the Sapphire brand, then Sapphire is at fault (the distributor doesn't matter as he is only moving the stock from point a to point b) unless all other manufacturers suffer from the same problems. Then the product design should be reviewed and the quality requirements should be rechecked - especially criteria for sound conformity. If the design is faulty, you can not blame manufacturers for producing a faulty product and the ball is again in the AMD court. If it is only Sapphire that has those kind of problems, then we all should ditch Sapphire cards until the problems are sorted out, and go with MSI or Gigabyte cards

 

[FormatC]

Sapphire produced ALL this cards for AMD and shipped it to other brands to label it. PC Partner is AMDs OEM and Sapphire is a brand of PC Partner. All cards are reference and were only labeled later. The same with Nvidias TitanX. Not one AIB has an influence. Companies like Gigabyte bought at the begin only 100 of this cards from AMD to put it in their own boxes

 

[Garrek99]

I've been reading reviews here for well over a decade and obviously really appreciate the info that you guys output.
What I'm unhappy about is the new image viewer control that has the left and right arrows so opaque.
Please make these arrows transparent so that they stop blocking content when the images are viewed on small screens.
Thanks.