Leading up to the Radeon R9 290X launch, I had two cards in the lab, both supplied by AMD. They ran at slightly different core frequencies in single-card configurations. No big deal, right? That was just PowerTune doing its job. But paired up in CrossFire, even with plenty of space between them, heat build-up forced the pair to drop to even lower clock rates. Right away, I was able to establish that AMD’s Hawaii GPU operates within a range of clock rates, determined by a number of variables. For R9 290X, that range starts at 727 and ends at 1000 MHz. As an aside, I do have an issue with vendors simply advertising this as a 1000 MHz GPU.
The discussion that so badly requires clarification, however, is a derivative of Hawaii’s inherent behavior. Because the GPU is always trying to run as fast as possible, and then adjusting down to obey certain power, temperature, and fan speed settings, every card is going to be just a little bit different. This is expected, and applies also to Nvidia’s GPU Boost-capable cards. But when I got my hands on a retail card purchased from Newegg, it was way slower. Like, 20% slower in many cases.
So that left me wondering: were the press boards just lower-leakage parts able to sustain higher clocks? Without a definitive answer, and with AMD insistent that my results couldn’t be right, I approached the Radeon R9 290 review much more cautiously, presenting data from both the sampled and store-bought 290X cards.
A Break In The Story
Then, on Tuesday, I received word from AMD that it had a smoking gun.
All of my numbers were generated using Quiet mode—AMD’s name for its 40% PWM ceiling, designed to keep acoustics in check. That’s a signal representing a percentage of maximum voltage into the cooling fan. So, from one card to another, 40% should yield roughly the same fan speed. Variance in fan speed is limited to the mechanical components, and should be very small.
But as it turns out, 40% on the cards AMD sent and 40% on the retail cards do not equal the same fan speed. To confirm, I compared the press card to a retail boards from Sapphire and Asus, using our Battlefield 4 benchmark at 2560x1440 and Ultra detail settings.

Right out of the box, the Asus card tops out around 1850 RPM, the Sapphire card peaks just under 1950 RPM, and the press board is sitting around 2050 RPM. “Big deal,” you say. “It’s a couple-hundred rotations per minute.” Well, let’s see how that corresponds to operating frequency.

At 1850 RPM, the Asus card cannot keep Hawaii cool enough, and the GPU hovers at exactly 727 MHz. In fact, at times, 40% isn’t even enough, which is why you saw it bump up in two spots on the previous chart—the card has to violate its 40% ceiling in order to stay within spec at 727 MHz.
Just 100 RPM more makes a massive difference. Rather than getting stuck at 727 MHz, Sapphire’s board averages 809 MHz through our 100-second Battlefield 4 run. Add another 100 RPM, and the press card yields an average of 917 MHz.
How does this relate to average frame rate?

And there we have it. The card that AMD sampled to press is 18% faster than the slowest retail sample in our lab, an Asus board. It’s nine-percent faster than the Sapphire card, which falls roughly in the middle.
The next step was to start fiddling with the Catalyst Control Center’s OverDrive applet to see if we could manually dial in equivalent fan speeds and solve the issue ourselves. On Asus’ card, a 43% maximum duty cycle got us as close as possible to AMD’s press board. Sapphire’s R9 290X needed a 42% override to get there. The outcome isn’t exact, but with 1% granularity in AMD’s driver, it’s impossible to get any closer.

The press and Asus cards are almost identical, but Sapphire’s fan is spinning about 30 RPM quicker. In theory, that only means it should perform better than AMD’s sample, if anything.

Asus’ R9 290X comes off of its 727 MHz floor with an extra 200 RPM of fan speed to help clock rates. Its new average is 792 MHz. Adding 100 RPM to Sapphire’s ceiling also helps, increasing that board’s average from 809 to 852 MHz. But neither retail-purchased product is able to catch the card we first received from AMD, which averages 917 MHz.

Even when we bring the retail cards up to the press board’s fan speed (and beyond), they are not as fast. In fact, our press board is still more than 11% quicker than Asus Radeon R9 290X. It’s 7% faster than Sapphire’s.
Shortly after AMD Radeon R9 290 Review: Fast And $400, But Is It Consistent? went live, AMD issued the following note to media:
“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.”
The official explanation for the numbers we reported in our R9 290X and GeForce GTX 780 Ti reviews is that Radeon R9 290X uses PWM to control fan speed. The fan it uses purportedly has a +/-300 RPM margin. So, 40% does not necessarily mean the same rotational speed on every card. But AMD claims it was not prepared to see variance as severe as what we saw. Company representatives say R9 290X should be spinning at 2200 RPM (although we've seen no evidence of any 290X at 2200 RPM/40%). Already, that puts Asus’ board beyond the fan’s specification at 1850 RPM.
Here in Bakersfield, CA, it still gets fairly warm during the day. I have to keep a thermal probe in the lab to ensure consistent-enough temperatures for testing. As I was taking measurements for this piece, I was seeing between 78 and 80 degrees Fahrenheit. One morning, I walked into the office and it was closer to 74 before benchmarking commenced. At that temperature, Sapphire's Radeon R9 290X behaved almost as well as AMD's press card in a 78- to 80-degree room. The reference thermal solution is operating very close to its limit, making R9 290X very sensitive to environmental changes. That's just one of those things you have to consider when you buy a card with such a wide operating range. Relief is only going to come in the form of more effective cooling from partners.
Now, all of this was to describe what we were seeing up until today (and what other outlets were commenting on). Hours ago, though, AMD issued this follow-up:
We’ve identified that there’s variability in fan speeds across AMD R9 290 series boards. This variability in fan speed translates into variability of the cooling capacity of the fan-sink.
The flexibility of AMD PowerTune technology enables us to correct this variability in a driver update. This update will normalize the fan RPMs to the correct values.
The correct target RPM values are 2200 RPM for the AMD Radeon R9 290X ‘Quiet mode’, and 2650 RPM for the R9 290. You can verify these in GPU-Z.
Roughly translated, this means that retail and press cards are being altered via software to operate at the same rotational speed under load. This would be a great thing if AMD was standardizing on the press card's fan, since that's the board we reviewed and praised. However, it's using this opportunity to put an additional 150 RPM through the 290X's fan, presumably to sustain higher clock rates for longer, improving performance. The side effect, of course, is higher power consumption and increased noise.
So, let's get Catalyst 13.11 Beta 9.2 installed and try again:

With the exception of the press card shooting up about 60 RPM for no clear reason, AMD's new driver successfully gets all three cards running around 2200 RPM. I don't have the same mic setup as Igor here in my lab, so I can't give you a good noise comparison. However, the card is noticeably louder than it was at 2050 RPM.

Clock rates are higher across the board, as you'd expect after increasing cooling capacity. AMD's own press board now averages 944 MHz, up from 917 MHz. Asus' troubled retail card settles at 863 MHz. Meanwhile, Sapphire crests 906 MHz in our test, on average.

Frame rates improve, too. The 11% variability we saw previously is down to 6%. That's great news. Remember that we're still dealing with the same reference cooler though...and it's getting louder. Fan speed as a percentage creeps up as high as 46% in the case of Asus' card, turning Quiet mode into Not Quite Quiet mode.
AMD is basically doing two things. First, it's fixing an issue that should have never made it past quality control and tightening up the variance between Radeon R9 290X cards in the wild. Simultaneously, it's pushing average clock rates of all cards higher using a faster fan speed setting that no board we've seen used previously. Our Battlefield 4 benchmark illustrates this nicely. By default, 190 MHz separated the average core frequencies of our fastest and slowest samples. After AMD's newest driver build, that number shrinks to 81 MHz.
If you saw our R9 290 coverage on Tuesday and decided that a $400 card delivering almost as much performance as the $550 Radeon R9 290X sounded tasty, then you were probably disappointed that we refrained from drawing a definitive conclusion about the press board’s representativeness. Let’s see if that trepidation was justified.

At launch, using Catalyst 13.11 Beta 8, Radeon R9 290 would spin up to about 2550 RPM under load, corresponding to a 47% PWM value. The same 47% on Sapphire’s R9 290 yielded just over 2300 RPM.

It’s no wonder, then, that Sapphire’s retail card bounced off of the board’s 662 MHz floor after just a few seconds of gameplay. By the end of a 100-second run, we recorded an average core clock rate of 726 MHz.
Manually scaling the card’s fan speed up to match our press board required a 50% PWM setting. But even at the same 2550 RPM, we were only seeing an average of 891 MHz compared to the press card’s 939 MHz.

Plotting out average performance shows us that the press-sampled R9 290 is 23% faster than our retail card. That’s exactly what I was afraid of, and the reason we held off on a judgment.
However, the new beta driver also imposes a new fan speed target on R9 290 as well. Instead of aiming for 40% PWM, like AMD originally planned, or 47%, which it later updated through software, it’s now aiming for 2650 RPM (more than 300 RPM faster than our retail card and about 100 RPM faster than our press board).

With Catalyst 13.11 Beta 9.2 installed, the press card maintains a fairly flat fan speed, while the retail board encounters strange spikes above 2700 RPM before tapering off and repeating itself.

Clock rates certainly do come closer together, though. AMD's press board has zero trouble maintaining its 947 MHz peak clock rate. Meanwhile, the retail R9 290 from Sapphire averages 904 MHz. That's not quite as good, but it beats the hell out of the 726 MHz it was averaging right out of the box and the 891 MHz it managed after we manually goosed its fan to 50% PWM.

As frame rates go, the retail 290 pulls within 3.5% of our press board, which is the sort of variance AMD originally told us to expect.
Making Things Right
Am I at least satisfied that whatever oversight resulted in retail cards shipping out at varying degrees of lower performance is being actively addressed by the team at AMD? Yes; now gamers are getting the performance that got them so excited when these cards first launched.
Catalyst 13.11 Beta 9.2 makes Radeon R9 290X a little louder as it adds about 100 RPM to the "Quiet" mode we originally reviewed. The difference is noticeable, but I don't think gamers lacking headphones will be perturbed by the sound. But the R9 290 was already loud. So picking up an additional 100 RPM, even after AMD cranked the PWM from 40 to 47% prior to launch, is detrimental.
The bad news is that I really couldn't imagine buying an R9 290 equipped with AMD's reference cooler, particularly in light of today's update that adds even more fan speed and noise. The good news is that I have now have higher hopes for third-party 290s. With Catalyst 13.11 Beta 9.2, our Sapphire Radeon R9 290 is just as fast as Asus' Radeon R9 290X, tested on the previous page. If we could just get our hands on more aftermarket cooling solutions, I'm pretty sure we could chip away at the most compelling reasons not to buy these boards today.