[TR] Retail Radeon 290x cards may be slower than press samples - 11/7 AMD SAYS FIXED

[OCGuy]

http://techreport.com/news/25609/up...9-290x-cards-may-be-slower-than-press-samples


Update 11/7 - AMD says it has fixed the variance between review/retail cards (note this is different than the no base clock issue) Just a small quote, I don't know where to find the public source so to be fair to [H] you'll have to click through for the rest:

Hello, We've identified that there's variability in fan speeds across AMD R9 290 series boards.....

http://www.hardocp.com/news/2013/11/07/update_on_amd_responds_to_r9_290_series_performance_variance






AMD has issued a response, and TR mentions they could not run these cards at the advertised speed of 1GHZ.

This seems to be its own issue, so I did not bury it in the review thread:


"A media outlet has uniquely reported instances of AMD Radeon R9 290X boards purchased in retail that have exhibited an uncharacteristic level of performance variance as compared to press samples issued by AMD. We’re working to secure the board(s) in question for further analysis. Boards purchased by other media outlets have not exhibited similar characteristics that we’re aware of. In the meantime, we’ve identified areas where variability can be minimized and are working on a driver update which will minimize this variance. We will provide an update shortly"


More information and background sources:

http://www.tomshardware.com/reviews/radeon-r9-290-review-benchmark,3659.html (original article that started it all)

http://www.fudzilla.com/home/item/33066-toms-finds-retail-r9-290x-card-runs-much-slower (Fudo confirming his card has variances, but no retail comparison and recapping TH)

http://arstechnica.com/gadgets/2013/11/amd-stomps-nvidia-with-r9-290-at-least-in-reviews/ (Recapping TH)

http://translate.googleusercontent....rsoker&usg=ALkJrhg-r-jkd4AxvySnp0pSXy_51G32Ww (Reviewer countering TH's claim, saying their cards are within acceptable range of a retail card)

http://hardforum.com/showthread.php?t=1789735 (Thread where I first saw AMD's statement, as well as a pretty good discussion on the subject, including some insight on the world of future GPUs)

 

[BrightCandle]

So presumably they are going to make the fan speed vary on cards that have much higher thermal output and hence clock lower by default?!

The other choice part of that statement:

Interesting. We should note that TR reader JohnC has encountered clock speeds well below 1GHz on his retail 290X. You can read about his experiences in the forums.

Forum link: http://techreport.com/forums/viewtopic.php?f=3&t=90345

This guy has the "normal" problem of the card throttling at 40% but he was expecting to get 1Ghz (not surprisingly). Its not the same level of problem THG had but its certainly surprising some users that they loose 20% clock speed when a lot of review samples a full 100 Mhz better than that.

 

[BallaTheFeared]

Image from the user they referenced:

Doesn't seem to be related to vsync, based on the gpu usage so I'm going to redact the thought that was here instead of this sentence.

 

[OCGuy]

As more of these cards make their way into PCs, the backlash appears to be growing.

Anandtech, Toms, and TR have addressed the issue, and I can only imagine more will follow.

I think AMD should have just kept a lower guaranteed base clock instead of muddying the waters by providing a best-case boost clock only. They may have lost a few FPS, but still would have had good price/performance, without the last minute driver scrambling and the bad PR.

 

[BrightCandle]

He is seeing a lot more throttling than the reviews, most of which saw 900+ and certainly more than 800.

Actually thinking about it this guy is seeing the same performance problem that THG did, that is a full 20% worse putting his 290X into 770/7970 performance levels. He doesn't have a Titan beater unless he increases his fan to near Uber levels, something all the reviewers has said is really loud.

AMD's statement to THG was that the card needed RMA, now its a driver fix, either something is really wrong in their drivers OR its to do the same thing to the 290X as they did to the 290, increase the fan speed to 50% or so.

Need to wait and see how this plays out over the coming days.

 

[DarkKnightDude]

This throttle nonsense is pretty funny to watch. Why can't games just run at a fixed clock?

 

[Silverforce11]

If you have a bad card, RMA it. A few bad samples so far, lets beat it up more guys!!

 

[itsmydamnation]

this could have something do to with it:
http://forum.beyond3d.com/showpost.php?p=1802697&postcount=1887

The current driver is controlling the fan by PWM set point. Electrical to mechanical can present some variance; we're in the process of changing the control to an RPM set point.

given that 7% fan change has such a big performance difference a few % either way +/- could have a bit to do with the issue. Could also explain some of the noise level variance being seen.


but im sure faux rage is far more satisfying for the usual suspects.

 

[BrightCandle]

If you have a bad card, RMA it. A few bad samples so far, lets beat it up more guys!!

Question is really is this card faulty? By spec from AMD its not faulty, its behaving as its described to do. Its not an obvious failure case because AMD hasn't defined what clock speeds you will get. I don't its that clear cut at all.

 

[GaiaHunter]

Question is really is this card faulty? By spec from AMD its not faulty, its behaving as its described to do. Its not an obvious failure case because AMD hasn't defined what clock speeds you will get. I don't its that clear cut at all.

In the EU you shouldn't have much trouble getting a replacement from the retailer claiming the performance of the product doesn't meet the expectations.

 

[ThackerSS]

Guys I can almost guarantee this is vsync or framerate limiting happening. I saw it happening in the latest Batman specifically due to the fact that it is vsynced and can't run faster than 60fps anyway.


You can reference my other thread, but any game that puts a cap on the framerate or if you cap the framerate or vsync yourself you'll see this exact same thing happen in both the GPU usage graph and the clock graph with the 290x. Anything that is "uncapped" will use all the mhz and max out the GPU usage. I spent a day trying to figure this out myself

 

[Silverforce11]

Question is really is this card faulty? By spec from AMD its not faulty, its behaving as its described to do. Its not an obvious failure case because AMD hasn't defined what clock speeds you will get. I don't its that clear cut at all.

If your card is NOT performing as advertised, you can RMA it.

Otherwise what exactly is the story here, a big AMD conspiracy whereby they send "golden samples" to reviewers.. and retail cards are 20% worse.. way to kill yourself with such a move as the masses of RMA comes in. Don't be daft.

 

[DarkKnightDude]

Guys I can almost guarantee this is vsync or framerate limiting happening. I saw it happening in the latest Batman specifically due to the fact that it is vsynced and can't run faster than 60fps anyway.

If this is true, what a big fuss over nothing. Pretty hilarious.

 

[OCGuy]

If this is true, what a big fuss over nothing. Pretty hilarious.

Wouldn't AMD just state this, instead of what is being reported?

They are going so far as to autopsy Toms cards.

I think it is safe to say v-sync isn't the issue.

 

[iiiankiii]

Two cards with problems = all cards with problems. That's how it works.

 

[BrightCandle]

AMD has made a statement to TechReport (its right there in the OP) that this is real and they are working on a fix to it, but we aren't sure what "it" is. Seems kind of crazy at this point to ignore there is a problem considering the manufacturer thinks there is.

 

[EndlessWargasm]

There is going to be some variance due to fan speeds and environment.

One fan at 40% is not going to be spinning at the exact RPM as another fan at 40%, even if they are the same model.

Some peoples cases are going to run hotter than others.

Some peoples ambient temps are going to vary more than others.

Due to the nature of boost clocks, this all has to be somewhat expected.

Could potentially be a bit of a PR nightmare, the scope of which will mostly be determined by how many reference cards get out there.

Make with the custom coolers already.

 

[GaiaHunter]

http://www.sweclockers.com/artikel/...der-for-golden-samples-sweclockers-undersoker

 

[BallaTheFeared]

According to Ryan from Anandtech the base clock is 662MHz for the 290, so working as intended.


I'd like to see the FCAT from that usage, looks ugly from this end.

 

[BrentJ]

There is going to be some variance due to fan speeds and environment.

Same on NVIDIA GPUs with GPU Boost... just sayin. This is the nature of dynamic clocks based on power/temp/etc.... environment, temperature, power, voltage, all of it affects the clock speed, on both GPUs. There can be a variance between NV GPUs of the same brand, or different brand, affected by environment or power or temp. Card A can run at 1215MHz, Card B might run at 1180MHz. Hopefully fixing the control point by which the fan is set on 290 series will reduce the variance. I like the fact that a simple driver update can fix such an issue, rather than having to do a BIOS flash or something.

 

[BallaTheFeared]

Same on NVIDIA GPUs with GPU Boost... just sayin. This is the nature of dynamic clocks based on power/temp/etc.... environment, temperature, power, voltage, all of it affects the clock speed, on both GPUs. There can be a variance between NV GPUs of the same brand, or different brand, affected by environment or power or temp. Card A can run at 1215MHz, Card B might run at 1180MHz. Hopefully fixing the control point by which the fan is set on 290 series will reduce the variance. I like the fact that a simple driver update can fix such an issue, rather than having to do a BIOS flash or something.

I think the biggest difference there is Nvidia cards have a much higher base clock, or min clock speed vs what they'll boost to than what it seems AMD cards are speced for.

As stated according to Ryan from Anandtech the base clock for the 290 is 662MHz, which is a far cry for the 1GHz boost it can run during short duration benchmarking and prior to thermal equilibrium inside a case.

I don't care for either method, but the truth is so long as these cards are above 662MHz (assuming this is official base clock as reported by Ryan) they're working as intended.

 

[Erenhardt]

Someone tell this guy to peel off the foil!

In all seriousness now. If AMD drivers have so much "power" I can see the issue being caused by multiple driver installations. Now it will be more critical then ever to do proper clean driver installation, then ever before.

 

[wand3r3r]

Mountains out of molehills / mud throwing again? Even the title doesn't suggest rumor or anything although the verdict is out on what is going on, much less if there is an issue with more than 1% of cards.

Interesting how the titan downclocking significantly (was it 10%, where are the links?) passed under the radar and was quickly ignored. This appears to be the same type of thing. Remove the benchmark impressing stunts they've been doing this generation and resort to static clocks, or benchmark them at guaranteed numbers so if you get a golden sample (reviewers too) then it's luck of the draw, however if you don't you get the guaranteed clocks.

To me the whole issue appears a publicity stunt, however we'll know soon enough if there is truth, and what the cause is. If I got a card that throttled back it would go.

I just find it ironic any time AMD starts stomping on NV, new viral "issues" arise. Perhaps it's just coincidence.

 

[BrentJ]

There are two different issues on the table here, currently being muddied together.

Issue 1.) Speed variances between the same models of video card. Potentially because of electrical and mechanical differences inherent to different boards. Current fan control point is by PWM, new control point will be by RPM, and this will be fixed in a driver update. Should stabilize performance between same brand of video card.

Issue 2.) 290X throttling it's clock speed below 1GHz in default. This one is normal, this is correct, the card will throttle it's performance below 1GHz while running in Quiet Mode at 40% fan. This data is everywhere to show this. Users should know that this will happen. The card is operating normally, this is its intended operation. One way to combat this is to set to Uber Mode, or set CCC manually to 100% so that there is no fan cap. This should help stabilize the clock speed closer to 1GHz. Everyone is hoping custom cards will fix this issue completely.

 

[wand3r3r]

:thumbsup:

Thanks for your input Brent, there are too many posters (apparently) with agendas and axes to grind.