Nope. Its the cooler, it can BARELY cool the GPU, and it CANNOT keep it under the 95º threshold with the low fan speeds target. Then the GPU throttles.
I don't agree. The issue with the 290 and 290X isn't the cooler by itself, it is the new powertune algorithm which dramatically lowers performance in relation to temperatures. This did not happen with the 7970. The 7970 more or less gave you consistent performance even with a reference cooler - AND you could easily use quiet fan speeds without a performance loss.
Just FYI, I owned the 5870, 6970 and 7970, all reference. None of those cards lost 20-25% performance for quiet (~40 or less%) fan speeds - I did not mind the reference cooler ONE BIT with the 7970. Because of the new powertune, it basically necessitates aftermarket cooling or high reference fanspeeds for optimal performance. This also was not the case with the 7970 and prior generation AMD cards - all of those performed consistently without variance even with the reference cooler.
Woah woah woah gentlemen. I wanted to know what you guys made of the bios thing. This discussion went in the wrong way.
Also I don't think 6% is significant, but whatever.
What do you guys think would have changed with the bios?
Also 25%? what?
Then we can start argue about case temperatures. Specially in a loaded system. My case temperature in idle is around 27-28C. I am sure it will reach or break 33C with CPU and/or GPU load. But the retail card above might flatline before 33C. And this is only 1½ minute testing.
Thist just in, GPU's throttle when in a case with GARBAGE air flow. Film at 11.
Also 25%? what?
...the effective cooling level on the 290 is so poor that even after dropping to the base GPU clockspeed of 662MHz it still generates too much heat, requiring it to ramp up the fan speed to compensate. With that in mind, we went ahead and ran a selection of our benchmarks with the 34% maximum fan speed. The performance hit, as you’d expect, is significant.
To get down to the 34%-38% fan speed range, the 290 has to shed an average of 22% of its performance, peaking under a few titles at 25%. To be sure this makes the card much quieter – though not as quiet as a GTX 780 – but it also sacrifices much of the 290’s performance advantage in the process. At this point we’ve essentially reduced it to a 280X.
Thist just in, GPU's throttle when in a case with GARBAGE air flow. Film at 11.
Yikes - this topic really has been talked to death hasn't it? Poor cooler + throttling = poor performance. The more interesting discussion would be what is changing between the press card bios and the retail card bios. I don't think we need another thread with 10+ pages of people screaming about the reference cards being awful and everyone else saying they're fine. Custom cards are supposed to be available in 2 weeks, can we just wait until then?
LegitReviews said:After a restart we noticed the BIOS part number was now 113-C67101A0-026 on the retail card, which is what we had on our press sample from AMD.
You can accuse AMD of everything(Yes, it's their fault) but not goldensampling like tom's sugested.
Look at this:
Their investigation achieved to find the problem of variating performance.
http://www.anandtech.com/show/7481/the-amd-radeon-r9-290-review/15
The variances due to fan speed and powertune (as mentioned in the original article at LR) are simply not acceptable. Consumers are presented with FPS charts in reviews which are borderline deceptive considering powertune dynamically throttles the card depending on ambient temps (especially at quiet fan speeds) which can lower the performance dramatically.
The 290X in the above chart is throttling by 300mhz after 90 seconds. [redacted]. 90 seconds. Is this acceptable? In a case with a 33C temp. Dudes, 33C hardly qualifies as "garbage" air flow. 33C is not unreasonable - that's better than average for a chassis with components under load.
Like CD said. AMD has a capable chip and doesn't have a clue what to do with it. Powertune needs to be re-written from scratch as to not cause so much variance in performance, and the reference design should be updated as to prevent the variance in performance. Now that would be a good product. As things are? How can anyone objectively find this situation remotely acceptable? People are buying cards expecting them to "work" out of the box and they're getting performance which is less than what they'd expect. Most consumers aren't power users that scan tech news daily and search for driver/BIOS updates. In fact, most of my PC gaming buds hardly ever read PC tech related news - those types of folks would be stuck with sub standard performance because they wouldn't be aware of a newer BIOS. They wouldn't be aware of what the "uber" BIOS does because it isn't the default. That isn't acceptable.
Press cards runs at lower voltage dont they? What if retail cards are not guaranteed to run at this voltage?
It's the Off-Set Voltage Range - How about Hawaii's GPU 1.258vm range (That's my conclusion) compared to nVidia's ? - AMD can do compared to nVidia's better Reference Cooling Solution costing $150 more this time around, Spend the difference on a Cooler and bury nVidia - Don't take me wrong because I love nVidia.Press cards runs at lower voltage don't they?
I didn't think Average Joe bought more than a bargin bin PC, and a $1000 tablet/smartphone.I'm with you on this schmee, for people like us who are power users we are acclimated to tweaking things and maximizing performance. Yet one must keep in mind that there is a much larger contingent of users out there who don't scan the web daily for information like this - they expect stuff to work to it's potential out of the box, you know? And for those people who just expect it to....just work, AMD is doing a huge dis-service for those folks. Fact of the matter is that Joe average consumer will pop a card in, use the defaults, and get less performance than they should.
4-5 years ago, I was one of those people. I didn't mess with forums and check tech news all that often - I just happily used my cards while being totally oblivious to dual BIOS' and stuff of that nature. I wanted stuff to just work. I think this is the largest contingent of users on the market - most people don't read forums like this, and they don't read tech news on a daily basis. They expect their stuff to just work.
The only way to make claims like others have done is to increase sample size by a drastic amount. A sample size of......two.....means absolutely nothing due to statistical variance. It's things like that which lead to broad claims and erroneous assumptions.
With that being said, I have four R9 290X samples in-hand and NOT ONE of them performs the same. We're talking about a delta of 15% in some cases and that isn't acceptable at all. Am I ready to publish the results? No, because in order to avoid errors and post a conclusion I am happy with, I need another six cards and more time.