The reason your room heated up so much is because you closed the door. If you had 780s in SLI or 290s in CF and left the room open, the temperature in the room would be very similar since they both consume a similar amount of power. If one card runs at 50*C and another at 95*C but they use a similar amount of power, the temperature rise in the room would be the same. It doesn't matter at all what temperature the GPU runs at since that's not what determines the temperature rise in you room.
R9 290 and 780 use a very similar amount of power in games.
This is just plain false. Heat generated is related to how much power is used (the more power used the more heat is generated), but there is no correlation to how much heat the card generates. The chips are different die sizes, have different internals (and heat spreaders) and different cooling solutions. Just because two cards use similar amounts of power do not mean they generate the same amount of heat. You cannot deny that 780s and 780ti's run at < 82 degrees and don't go above that even when overclocked. Even running Uber mode (which is more than twice as loud in real world sound) the AMD cards run over 90 degrees.
That's why I made the throttling comment, because even with annoyingly loud fan speeds they still hit 90+ and throttle sometimes. The reason a 290 can keep up with a 290x is because they changed the bios to default to a higher fan speed after the heat debacle of the 290x launch.
I am pretty sure he first mentioned 780Ti SLI because I would have never even started talking about 3x 780s vs. 2 780Tis as a better alternative or recommending the much more cost effective Gigabyte 780 Ghz Edition. I am pretty sure he edited his post. If he mentioned 780 SLI, I wouldn't have disagreed as I think it's an gppd setup for the $ before after-market R9 290s come out.
If he edited it I didn't see it. All I saw was you quoting his post that had no mention of the 780ti...
Not really. Is that why it takes 780TI SLI to keep up with 290X CF at 4K and multi-monitor gaming and why 780 SLI can't beat R9 290 CF? If you looked at recent reviews, there is no evidence at all that SLI works better than CF when it comes to 780 SLI vs. R9 290 CF. In fact, CF tends to scale better. If you were talking about CF issues on 4K and Eyefinity for HD7900 series, I would agree.
Nope. The reason it takes a 780ti to keep up with 290x is because NVIDIA had awful 4k and multi-monitor drivers for their high end setup. The reviews stated the issue was acknowledged and being addressed. 4k is also hugely bottlenecked at this point so it doesn't really matter what you throw at it, nothing runs 60fps. I care about 1080p and 1440p. Most gamers run at 1080p and the future is 1440p, not 4k. There isn't even native 1440p content...it's all upscaled. It will be over 5 years before a game engine is updated to render in 4k and game companies start producing native 4k content. Mention 4k to me when that happens, until then it's meaningless.
You must be talking about running R9 290X in quiet mode, yet your statement is implicit in general for all NV vs. AMD cards. R9 290 doesn't throttle if you set a manual fan curve and neither do any of the other 5870/6970/7970 cards.
If you want your cards to be more than twice as loud and sound like an aircraft taking off, feel free. I'd rather quiet performance, thanks. My dual 780ti system is whisper quiet, you can't even hear it from a couple feet away.
Also, you statement that "NV cards run much cooler" is ambiguous at best. How cool a card runs is largely determined by the combination of the cooler and its power usage. Slap an Accelero on an R9 290X or GTX480 and temperatures will drop to mid 60s.
There is nothing ambiguous about it. The NVIDIA cards run at 80~ degrees, the new AMD cards run 90-95~ degrees. I'm not putting a custom cooling solution or a water block on a card just to make it reasonably cool. That's extra money for no better performance.
False. You do not need an uber case to achieve high airflow. A
$100 case and a motherboard with 3 slot spacing would work perfectly. 2 Asus DCUII 780s or 2x Gigabyte Windforce 3x 780s would offer superior temperatures, noise levels, overclocks than 2 reference 780s in this style case (2 intake fans, 1 rear exhaust fan and a solid top exhaust fan).
You misunderstand what I'm saying. I'm not saying you need an uber case to keep dual gpu temps reasonable. I'm saying that if you overclock your cards sync their speeds to the lower of the two. So if your top card is hotter (it always will be) then you don't want an open air (ACX or windforce) cooler because all that heat from the bottom card will make the top card even hotter. If you top card is 10 degrees hotter it won't overclock as far, that means your bottom card will match the lower speed. SLI overclocks are MATCHED. You can't have one running at 1250 and one at 1150, it doesn't work that way.
So if you intend to go SLI, you get blower (titan) coolers so the hot air is vented outside the case. That will keep the two cards temps closer together and allow you to overclock higher. It also allows you to achieve positive airflow easier because you don't have video cards mucking up your airflow by venting air in a weird direction. I have a Corsair 750D and even though it's not a very good case for airflow my temps are very low even with a 4770k and two 780ti's overclocked.
Reference blower coolers cannot compete in a modern case in SLI. People who keep saying they do keep perpetuating the myth of blowers repeated from the 90s.
Wrong again. They do compete, I have a setup to prove it. AMD blower coolers can't compete because they suck. The NVIDIA Titan coolers are robust and work very well, there have been several reviews and benchmarks showing temps with reference coolers to be better for SLI. I don't have any links handy but if you google it I'm sure you can find some.
$250 more but EVGA Classies overclock to 1.3-1.35ghz consistently, something that can't be said of 780TI. Once you account for the fact that at 1.25Ghz a 780Ti SLI is only 10-12% faster than 1.15Ghz oveclocked 780Ti SLI (
GameGPU review), it's evident that even if scaling is just 50% on the 3rd EVGA Classy card, it will cream 780 TI SLI setup.
First of all, they don't all clock to 1.3-1.35. You need a golden chip to go over 1.3, it's a safer comparison to say the classified will clock to 1.25. Your logic is also flawed for a couple reasons: 1) the 780ti will overclock just as well as the 780. If you want to look at the classified, wait until the 780ti classified hits. The 780ti at 1250 is considerably faster than a 780 @ 1250, to the tune of 20% because of the CUDA cores...and that's not even counting the faster memory speeds.
When you combine that with the fact that scaling on the third card isn't very good, I'll take two overclocked 780ti's over 3 780's any day. Also, don't forget, adding a third card adds considerably more heat to the case, so you're not going to get 3 780 classified cards clocked to 1.3...since they sync to the lowest overclock you'd be lucky to even hit 1.25. I'm not talking about professional benchmarkers who get engineering samples either, I'm talking about Joe Schmo consumer with 3 retail cards. Feel free to buy 3 780s and I'll gladly run my system vs your's and we'll see who comes out ahead.
Benchmarks already show that 780Ti SLI oc is not worth it over 780 SLI oc. So how can 3x GTX780 Tri-SLI OC not destroy 780Ti oc SLI? They would unless we are talking about CoH2 or some game with broken SLI scaling. :thumbsup:
And what benchmarks are these? Are you talking about meaningless 4k ones again? Every benchmark I've seen shows that the 780ti crushes the 780, the 290x and the Titan at everything but compute...and that's because it's neutered intentionally for that. In SLI it's an even bigger performance gap.
And if you really want to talk about value, R9 290 in Tri-fire will pay for itself with litecoin mining in less than 1.5 months and after have no problems beating 780Ti SLI. So, if you went there talking about performance/$, then NV isn't even an option if the OP doesn't mind setting up LTC.
I don't litecoin or bitcoin mine, I don't care to waste my bandwith, my electricity or my time on that [redacted]. I have a real job, that makes me real money. If you're really serious about making money with "coin farming" buy a server and pro cards and go to town, none of this amateur maybe it will work crap. If you were able to easily, no problem, 100% guaranteed make money back then everyone would do it. But you can't, it isn't guaranteed and most people don't even make money doing it.
You're looking at all the wrong things and drawing the wrong conclusions. I've made my recomendations and I will stick by them. I still think if you're spending $3-4k on a system then jumping from $550-$700 is nothing and I'll take the 15% + gains in performance for 7-10% total system cost any day.
Profanity isn't allowed in the technical forums.
-- stahlhart