Retail Radeon R9 290X cards may be slower than press samples

[PedantOne]

http://techreport.com/news/25609/upd...-press-samples

http://www.fudzilla.com/home/item/33...ns-much-slower

Interesting, remember the same articles after GTX 680 launch?

[m_jones_]

http://translate.google.com/translat...0%26bih%3D1332

http://forums.overclockers.co.uk/sho...73&postcount=9

With the results in hand, the picture is clear. The performance is basically identical between the press copy and graphics card from the shelf, at least in Uber mode. Any single frame per second is different, which is what may be considered normal as bonds or uncertainty in the measurements.

In the quiet mode, where the dynamic frequencies to work overtime, the situation becomes slightly turbid. A minor performance difference can be seen in some titles, and even if it is not about considerable variations, the trend is clear. In the end, it does an average variance tion of only a few percent, ie no extreme levels. The reason may include slightly less contact with the cooler, or simply easy changing ambient temperature.

[BababooeyHTJ]

My Titan throttled down to 1v in normal usage. There was a large thread about the issue on Nvidia forums and it seemed like most Titan owners at launch on OCN had the same issue. Where were the articles about that?

I was never a believer that Nvidia pays off some of these review sites to give amd bad press but an article like this on the day that 290 gets released about a single sample without any similar complaints on the forums has got to make you wonder. Especially after never questioning Nvidia for doing the same or holding back drivers for older products in order to make the newer product look better.

[BababooeyHTJ]

Looks like sweclockers is seeing similar results with a retail card on the quiet setting compared to their review sample.

[FlanK3r]

i think, the reality is different, better for 290x. Because new drivers, no slower

[AliG]

I don't get why this is a big deal. Companies always bin the review samples they hand out, and so it makes sense that uber performance would be the same but quiet mode might go down due to worse power consumption.

This really is no different than when Intel hands out review samples that overclock better than the samples the consumers get. No one questions that CPU companies will bin their review samples, so I don't see why it should be a big deal with graphic cards

[MaddMuppet]

I don't get why this is a big deal. Companies always bin the review samples they hand out, and so it makes sense that uber performance would be the same but quiet mode might go down due to worse power consumption.

This really is no different than when Intel hands out review samples that overclock better than the samples the consumers get. No one questions that CPU companies will bin their review samples, so I don't see why it should be a big deal with graphic cards

Totally agree. I'm sure all hardware makers do exactly the same.

[Olivon]

Looks like sweclockers is seeing similar results with a retail card on the quiet setting compared to their review sample.

Seems not.

With the cards in Quiet Mode and under full load by Metro: Last Light over up 60 minutes time levels stabilize. The difference in frequency is clear - the reference card from AMD is most of the time up to 100 MHz higher. The specimen from PowerColor Park after a while at 748 MHz, and refuses more or less to move.

http://translate.google.com/translat...ers-undersoker

[tajoh111]

So swedclockers, Hardware.fr and tomshardware are showing funny results. The swedclockers one is kind of a shocker as people were using this as the proof that funny cards weren't being sent out. Or cards with a funny bios.

[mars]

With another day of data at hand, it is clear that differences between graphics cards available. Specimens from PowerColor include a slightly slower fan combined with a slightly higher voltage to the GPU, which together provide lower frequencies as the temperature puts a stop.

This, however, only limited Quiet mode where the fan speed is locked to a maximum of 40 percent. In more ?ber released, which is what SweClockers use reviews, the difference in performance nonexistent - both press copy and PowerColor card can operate at peak frequency even for long periods.



Please spin this to state that AMD has sent out golden samples to reviewers

[BababooeyHTJ]

Seems not.

Thats what I meant, similar results to Tom's.

I still don't get the big deal. If everyone else is cherry picking review samples is AMD really creating some sort of unfair advantage?

Although, the cooler is probably even crappier than most reviews will let on.

[eric66]

Amd made stupid mistake by not allowing partners to launc with their own custom coolers. all issues so far have one source which could have been fixed since day one without creating so much drama

[tajoh111]

Thats what I meant, similar results to Tom's.

I still don't get the big deal. If everyone else is cherry picking review samples is AMD really creating some sort of unfair advantage?

Although, the cooler is probably even crappier than most reviews will let on.

A 200mhz drop on performance is enough to completely change the tone of reviews(this is what swedoverclocks got and similarly toms hardware). Now you basically have a quiet mode that gives you a card that is basically badly priced, as it makes the card 12 percent faster than a 280x but costing between 33% to 83.33% more. It completely changes its lustre vs a gtx 780 now.

Uber mode is now the only real mode people should use if they buy this card or their better off buying a 280x, gtx 770.

The magnitude of the difference is what is particularly bad about this. With nvidia or Intel(I don't remember intel needing to do this) even cherry picked, I don't remember cherry picked cards running 18 percent faster.

Now that more evidence is coming about, are we going to justify it because we say everyone does it? Where before we were saying no need to worry, AMD wouldn't do this.

It seems to me, everything becomes okay when AMD does it like rebranding.

[Hornet331]

Amd made stupid mistake by not allowing partners to launc with their own custom coolers. all issues so far have one source which could have been fixed since day one without creating so much drama

its also quite funny when you look at the 280X, there isn't a single one with reference cooler and most of them are already compatible with the 290s...

snip

this whol ranting gets boring, this "issue" will be a none issue in less then 2 month.

Just look how many HD7xxx reference designs are out there... good luck finding one. the first 3rd party designs for them came out 2-3weeks after the release of the HD7xxx series. After this point basically noone bought reference cards anymore.

The only ones that will be hit with this "issue" are early adopters and most of them will yank the stock cooler off anyway.

[eric66]

whoever came with idea of dual mode while using this craptalistic cooler and not allowing custom ones should get fired immediately. this thing is currently pissing on their parade even though cards are great people are mostly focused on heat/noise department

[nossy23]

I don't get why this is a big deal. Companies always bin the review samples they hand out, and so it makes sense that uber performance would be the same but quiet mode might go down due to worse power consumption.

If it is true that the difference is really 20% in performance like Tomshardware is reporting, it is a big deal. We'll see.

[kuroikenshi]

there must've been some decision made at amd to halt the release of after-market coolers,

It could be a strategic move to ensure prices remain flat for the next two months, after all we've already seen that these cards are superb performers when cooled accordingly. the fact that 290x are going out of stock globally means that they've already sold tens of thousands of units.

and if by ensuring flat pricing on high demand products is good, I honestly see that AMD has made a sound economical and strategic move by doing what they've done.

also, there remains the strategy to increase market range for GNC, which is something mantle will really need, best way to win developers' hearts to use Mantle by being the architecture of choice by gamers.

I see flat pricing and Mantle the two prongs out of 3 prong strategy (third being console market).

[Hornet331]

the fact that 290x are going out of stock globally means that they've already sold tens of thousands of units.

how can something go out of stock, when most retailers didn't have any stock to begin with.

I had a look at many retailers here and most of them wont get any stock before mid november, many not even before last week of november.

Hell i wouldn't be surprised if all the manufacturers keep the sent cards at a minimum so they can ship there custom cards as soon when they are allowed to do.

[BababooeyHTJ]

Uber mode is now the only real mode people should use if they buy this card or their better off buying a 280x, gtx 770.

Yup, pretty much. AMD botched the quiet mode fan profile from the looks of things. Either way that shouldn't be a surprise. Even in the review samples. How do you think that they'll perform in high ambient temps?

At least every fan profile adjustment is available right in CCC. Its not like you need to download a third party tool or flash a bios.

You either find a fan profile that you like or flip a switch. Either way "problem solved" sort of.

At the end of the day the cooler on this card is a pos. I think that eric66 hit the nail on the head. AMD should have allowed aib partners to release non-reference models at launch. Sort of like Nvidia did with GTX 670. Even if it meant a delay.

Let me just say this. At least I don't have to wait a few weeks for a third party tool to edit and flash the bios on this card to stop it from throttling like I did with Titan.

[mars]

im still flabbergasted that people would pay > 500 for a high end graphics card, then moan and groan about the stock cooler (lol)

whats the 50-100 bucks (200 for a block/bp/fittings) compared to the 200-300 more I would pay for a 780?(before the AMD-caused Nvidia price drop)

then you would have the need to get some nice 4k monitors still...who cares about the small cost of a decent cooling solution for your card?

Here in Australia most retailers have the card though most are sold out or on single digits

[nossy23]

im still flabbergasted that people would pay > 500 for a high end graphics card, then moan and groan about the stock cooler (lol)

whats the 50-100 bucks (200 for a block/bp/fittings) compared to the 200-300 more I would pay for a 780?(before the AMD-caused Nvidia price drop)

then you would have the need to get some nice 4k monitors still...who cares about the small cost of a decent cooling solution for your card?

Here in Australia most retailers have the card though most are sold out or on single digits

Why moan and groan about it? How about you lose your guarantee on your card if you replace the cooler? It's not only about the 50-100 bucks you know.

[Hornet331]

Why moan and groan about it? How about you lose your guarantee on your card if you replace the cooler? It's not only about the 50-100 bucks you know.

then wait for a card with another cooler.. is that so hard?

Also its not like ocing wont void your warranty...

[BababooeyHTJ]

im still flabbergasted that people would pay > 500 for a high end graphics card, then moan and groan about the stock cooler (lol)

whats the 50-100 bucks (200 for a block/bp/fittings) compared to the 200-300 more I would pay for a 780?(before the AMD-caused Nvidia price drop)

Come on, that initial investment into water cooling is very expensive. A decent loop for two cards will cost you more than you paid for one of the cards itself.

[Ace123]

I work for a company that manufactures parts and peices for aircraft, some civillian, and some military. I work in an FAA repair station part of the facility, but I often work closely with our prototype department and also manfucaturing as well. For manufacturing im often utilized as a consultant to identify problems with newly manufactured products. Because I spend my days repairing the very products they are building new, it makes me very adept at recognizing the cause of those problems because the units I receive almost always have identical problems, but usually because of wear and not a newly created but faulty part. I help protype often, but in a different capacity. For them I help complete steps on prototype parts that I regulary perform on our repaired parts, but they rarely perform due to the fact that the trial, and final stages of assembling a prototype build is a small fraction of the entire process for them. As you can imagine most of what they do is identifying parts that they currently have that they can utilize, and modify to serve the initial function of the prototype, but also layout what they will require, and how they will obtain it to build a prototype, then begin to assemble a part out of nothing but paper drawings. Part of this knowledge will help me explain to you what I know about the manufacturing process from protype to product.

When a product is nearing the end of the prototype phase, a series of qualification tests will begin to identify any possible problems that may arise as the final prototype (hopefully) is assembled. Once satisfied, a final prototype model is built as though it were to be manufactured, but a extreme level of scrutiny is used at every step to ensure exacting standards. The final prototype will be perfect, or as close to it as possible. This part should be superior in quality unless an inherent design flaw is present. Once qualified, the part will be manufacturable. In AMD's case, their graphics cards are manufactured by third parties. Those companies will then all create identical prototypes until AMD is satisfied that their quality standards meet or are within tolerance the quality standads of AMDS original. During this process, those third parties will go to extreme lengths to assure that the quality of their prototype samples to AMD are as near perfect to the design as possible as well. Once AMD is satisfied they will release manufacturing athority and the manufacturing process will begin. The only problem is, once granted the authority to manfucture, the quality standards are often relaxed a bit from the prototype phase to cope with delivering product at an acceptable pace.
I imagine the parts the reviewers recieved were scrutinzed by AMD for exacting standards and quality was high as AMD was to make sure reviewers received good parts that meet their specifications.
Retail cards are of manfacturing quality and meet standards with variances allowed by the manufacturing process. These parts function as intended by AMD, but often first runs of them do not meet perfecting standards of the prototypes until the manufactures become adept at building them.

[GoldenTiger]

Come on, that initial investment into water cooling is very expensive. A decent loop for two cards will cost you more than you paid for one of the cards itself.

Not to mention that most people don't put a watercooling loop into their pc's, let alone for 24/7 use.