Anyone concerned or is your water cooling drowning any mention of fear? Honestly I figured with tuning software and some fan speed I would be able to override the 280 to stay at top clock anyhow but not going that way anymore anyways.
Water cooling here:
After seriously overclocking a GTX 480 24/7 for years in the same loop as a seriously overclocked I7-920, which is probably the hottest and most power-sucking CPU ever (maybe 6-cores rival it nowadays), I laugh at all mentions of downclocking due to temperature. GTX 480 is the hottest and most power-sucking single GPU card I think that there has ever been. The R9 290 series might rival it, but even they are on the lower end.
HOWEVER, as I mentioned I'm on water with some serious radiators (overall 6x120 "blocks" spread over 3 radiators), and I would NOT consider the R9 290 series if not on water. IMO if you're doing air, I would spend the extra and just get a 780. Maybe a 780 with an aftermarket cooler even. I ran my GTX 480 with the stock cooler for a few days while watercooling parts were still arriving back in the day, and I remember it NOT being a good experience. I was able to easily hear it to the point of it bothering me through closed-ear headphones in a fairly enclosed case. It might be safe for the GPU to work at 95C, but think about what all that heat does to your motherboard, memory, HDD/SSD, PSU, and everything else around your case. Even if it's mostly exhausted out, it still "radiates" in your case - even from the back side of the chip and the PCB itself, and heats everything up. With it now being winter, then think about how it would be in the hot summer next year.
Nevermind the ridiculous noise.
A good aftermarket cooler will for sure mitigate these issues, however there is no (air, with fan on it) cooler that will magically make a R9 290 perform like a GTX 780. It's fundamentally "flawed" (if you see it that way) or worse as far as how much heat the chip outputs compared to Nvidia's. Sidenote: I think Nvidia can actually make cards that perform SIGNIFICANTLY better than the cards they have now - if they increase their heat/power targets - through chip engineering or voltage increases. However, I think they're artificially "holding back" as a strategy to make their next-gen look that much better early next year, or maybe for other reasons, and in part because... they had no reason to! They had no competition for 6 months - all those 780/Titan sales with ridiculous margins, who cares if they "lose" now, 6 months later - by some % of performance, dollar for dollar. I really think they can make cards, today, that perform at 1.5x what the Titan does - maybe for close or the same price, with a higher voltage/clock or more shaders on the chip or some other engineering change - and those cards would also output 95C and sound like a R9
. But what's the point? What's the use in beating a dead horse (dead for 6 months that is)? On the other hand, I really think AMD is straining, and we can see it with the R9. Performing cards, but IMO the engineering is far more "on the edge" than Nvidia's. They're really straining to catch up. I think we'll see the same next year - Nvidia releases expensive, excruciatingly well-performing cards early in the year, AMD strains to catch up while Nvidia sells the shit out of their cards for most of the year with no competition.
If you already have a water loop (like me) and looking for a water-water replacement card, IMO the R9 290 is a ridiculously good value. I highly prefer nvidia due to more professional appearance, better engineering (IMO) and overall more stable, mature, better "feeling" drivers, however, this card has really got me thinking hard. I'll wait for the 780Ti reviews due soon enough, and I haven't bought something yet, but I'm buying a GTX 480 replacement sometime before Christmas (it's just time...), and for now I'm heavily leaning towards the R9 290. Mostly because of price. Think about it, it's HALF the price of the 780Ti for pretty much the same average performance (strictly speaking frames/second/$).
Someone mentioned Mantle. Okay, it was very highly advertised but NOBODY knows how it will perform yet, and in what situations. My gut feeling is that for most people, it will make NO difference - like, at all. It might increase the FPS in high framerate situations (looking at the ground or at a wall), but the frames that are shader-limited will not change. It's supposed to facilitate the CPU-GPU interaction, but what about situations where games are GPU limited? I don't think Crysis 1 is CPU limited or with how often the CPU "feeds" the GPU frames. It's mostly limited by the extreme utilization of shaders - something no Mantle will help with. That's just my gut feeling about mantle...
Also, if we're talking esoteric shit, consider the flipside: AMD doesn't have PhysX, so you won't see all those pretty cloth and particle effects in Borderlands 2 or Batman games. It also doesn't have GSync, even though I fail to get excited about that, given the rumors (or was it an announcement) that it'll be Asus-exclusive for over a year. I haven't had a display made by Asus, and I doubt I ever will. Also exclusivity = higher price = higher margin. There's a reason Asus probably paid up the ass for those rights, the same reason AMD paid up the ass for DICE and EA to effectively advertise their Mantle shit, whether it'll perform or not.
P.S. Speaking of esoteric shit, one other thing that's got me seriously seriously considering a R9 290 is the outstanding OpenCL performance of AMD compared to Nvidia (Nvidia has silently been intentionally crippling that performance and other features on "gaming" cards since the 4xx series to promote its Tesla for "compute" and Quadro for "professional/workstation applications" lines), i.e. Litecoin mining is usually times faster without exaggeration - in addition, it's much better in speed/watt compared to Nvidia, again in factors. I like to mine in my idle time, while I'm at work during the day or when I sleep, which the R9 290 will do MUCH better than a 780 Ti - even cgminer vs cudaminer for those who are more familiar. I haven't seen exact numbers, but my feeling is more than 1.5x better. Then again, not many people care about OpenCL... yet
.
