AMD Radeon R9 295X2 - Page 2

kladner · 2014-04-11, 03:33

As far as the gtx560 goes, I was guessing until I looked it up on James' chart. Maybe you've said; is your 560ti a 448 core? That certainly ups the power draw a lot. I was basing my estimate on what my GTX 460 draws, and figuring that efficiency has usually gone up from one generation to the next.

I would also guess that the fan on the AMD is dissipating all the heat of the 500 W infrastructure of the card, while the closed loop is doing the chips themselves. With the middle fan design, some hot air goes out the rear, but a lot spills in the case. While I'm sure the VRMs, etc. get quite hot delivering that kind of power, the chips' dissipation has to be the greater heat output; and that will presumably be going straight out, taking some of the ambient heat with it. This can bode well for the amount of load added to your case.

The other question is, how does your 560ti arrange its cooling? Is it the single-blower-exhausting-out-the-back type, or is it multiple fans which dump mostly inside? If the former, you could be in very good shape with both cards running hard. The latter might not be terrible, either, especially if you aim for positive pressure in your case. This encourages more hot air to go straight out through the rear grills. However, there will still be more heat spilled inside, which will reduce the effectiveness of the liquid cooling, as well as that of your CPU.

Your thought to try a new dual card alone is probably a very good one. It would probably also be good to wipe the nVidia driver to the best of your ability, get the AMD running right, and then take on adding the nVidia back in. Besides taking it slow on the driver front, this would also give you a good read on the effect the 295X2 has on your system and its cooling.

EDIT: On the other hand, a single 120mm radiator for 500 W might still run pretty hot on the GPU itself. I tried a similar liquid cooler on my fx8350 (125-140 W), and promptly went back to the huge-ass air cooler I had been using. The direct exhaust did not compensate for the reduced size of the heat exchanger. I did not keep the liquid cooler long enough to look into the other cooling issues which LaurV mentions.

tapion64 · 2014-04-11, 07:52

Quote:

Originally Posted by kladner

As far as the gtx560 goes, I was guessing until I looked it up on James' chart. Maybe you've said; is your 560ti a 448 core? That certainly ups the power draw a lot. I was basing my estimate on what my GTX 460 draws, and figuring that efficiency has usually gone up from one generation to the next.

No, it's a 384 core. GeForce definitely shows it as being 170W, the source I was looking at probably had its facts wrong (or I was misinterpreting the setup).

Quote:

Originally Posted by kladner

The other question is, how does your 560ti arrange its cooling? Is it the single-blower-exhausting-out-the-back type, or is it multiple fans which dump mostly inside? If the former, you could be in very good shape with both cards running hard. The latter might not be terrible, either, especially if you aim for positive pressure in your case. This encourages more hot air to go straight out through the rear grills. However, there will still be more heat spilled inside, which will reduce the effectiveness of the liquid cooling, as well as that of your CPU.

It's the latter, multiple fans on the GTX. Then there's a 120 mm fan on the back grill, one side of the case is grated so there's lots of air flow. There's a 200 mm fan on the top of the case, and there's a 200 mm fan in the front grill. And of course a fan on the CPU. I really have no idea how that influences the pressure, I'd have to research it. The case was big enough to fit everything I wanted and it had good reviews so I went with it. For reference, it's a Cooler Master HAF 922. I have no doubt that the dual card by itself should be fine, and given all the fans and airflow improvements, I think it can handle the 560 Ti running concurrently. Then again, I might not want to run it just for power costs >.>

kladner · 2014-04-11, 11:12

Quote:

Then again, I might not want to run it just for power costs >.>

I hear you on that!
EDIT: I do wish I'd gone with a case like that. I also tend to agree that if the price difference isn't that great, that the larger PSU, at the highest efficiency rating you can get would be the way to go. Having some overhead is a good thing, and the efficiency really counts at those power levels.

kracker · 2014-04-11, 20:24

Quote:

Originally Posted by kladner

I hear you on that!
EDIT: I do wish I'd gone with a case like that. I also tend to agree that if the price difference isn't that great, that the larger PSU, at the highest efficiency rating you can get would be the way to go. Having some overhead is a good thing, and the efficiency really counts at those power levels.

Efficiency is higher on GCN than Fermi, though power consumption is higher.

GTX 560 ti= 1.3 Ghz/W
GTX 580= 1.7 Ghz/W
R9 295X2= 2.4 GHz/W

kladner · 2014-04-12, 16:29

Quote:

Originally Posted by kracker

Efficiency is higher on GCN than Fermi, though power consumption is higher.

GTX 560 ti= 1.3 Ghz/W
GTX 580= 1.7 Ghz/W
R9 295X2= 2.4 GHz/W

Correct. I was, in fact referring to PSU efficiency. I saw a measurable reduction in line draw (w/ same components installed) when I switched from a 750 W Bronze supply to a 1000 W Gold.

Of course, at least two factors were in play. Not only was the KW supply higher efficiency, but it was running in a more comfortable power range. The 750 W was running "pedal to the metal."

tapion64 · 2014-04-18, 14:55

A quick question. I don't think this plays that big of a part until I want to crossfire cards, but I'm having trouble finding a concrete answer. It seems like for computing applications that don't have much of a CPU bottleneck like trial factoring, the PCIe slot type doesn't matter much for performance. Is this true?

I have a motherboard that has 2 PCIe 2.0 slots, one running at x16 and the other at x4. I figure the answer is absolutely need to upgrade mobo if I go crossfire and get atleast 2 PCIe 3.0 x16 slots. But is there a performance increase in that upgrade for a single card setup?

kracker · 2014-04-18, 15:22

Quote:

Originally Posted by tapion64

A quick question. I don't think this plays that big of a part until I want to crossfire cards, but I'm having trouble finding a concrete answer. It seems like for computing applications that don't have much of a CPU bottleneck like trial factoring, the PCIe slot type doesn't matter much for performance. Is this true?

I have a motherboard that has 2 PCIe 2.0 slots, one running at x16 and the other at x4. I figure the answer is absolutely need to upgrade mobo if I go crossfire and get atleast 2 PCIe 3.0 x16 slots. But is there a performance increase in that upgrade for a single card setup?

mfaktc/o to the best of my knowledge, does not care for a lot of PCI bandwidth so you technically should be fine.

kracker · 2014-04-18, 21:29

Quote:

Originally Posted by tapion64

A quick question. I don't think this plays that big of a part until I want to crossfire cards, but I'm having trouble finding a concrete answer. It seems like for computing applications that don't have much of a CPU bottleneck like trial factoring, the PCIe slot type doesn't matter much for performance. Is this true?

I have a motherboard that has 2 PCIe 2.0 slots, one running at x16 and the other at x4. I figure the answer is absolutely need to upgrade mobo if I go crossfire and get atleast 2 PCIe 3.0 x16 slots. But is there a performance increase in that upgrade for a single card setup?

Sorry, didn't answer last question. For single card, it should be fine with 2.0 x16(at most a single % digit slowdown I think)
..This is assuming for programs that use the bandwidth, games etc.

Gordon · 2014-04-18, 23:44

Quote:

Originally Posted by tapion64

A quick question. I don't think this plays that big of a part until I want to crossfire cards, but I'm having trouble finding a concrete answer. It seems like for computing applications that don't have much of a CPU bottleneck like trial factoring, the PCIe slot type doesn't matter much for performance. Is this true?

I have a motherboard that has 2 PCIe 2.0 slots, one running at x16 and the other at x4. I figure the answer is absolutely need to upgrade mobo if I go crossfire and get atleast 2 PCIe 3.0 x16 slots. But is there a performance increase in that upgrade for a single card setup?

It doesn't make a scrap of difference, I have a GTX-660 in the "fast" slot, no overclocking it gets ~172 days/day. Another GTX-660 in the "slow" slot - gets exactly the same throughput.

kladner · 2014-04-19, 01:22

It could, conceivably have made a difference with CPU sieving. But with that happening on the GPU, it seems that traffic would be greatly reduced.

kracker · 2014-04-19, 01:39

Quote:

Originally Posted by kladner

It could, conceivably have made a difference with CPU sieving. But with that happening on the GPU, it seems that traffic would be greatly reduced.

It is. Well... not many people sieve on the CPU now I think

2014-04-11, 03:33	#12
kladner "Kieren" Jul 2011 In My Own Galaxy! 10158₁₀ Posts	As far as the gtx560 goes, I was guessing until I looked it up on James' chart. Maybe you've said; is your 560ti a 448 core? That certainly ups the power draw a lot. I was basing my estimate on what my GTX 460 draws, and figuring that efficiency has usually gone up from one generation to the next. I would also guess that the fan on the AMD is dissipating all the heat of the 500 W infrastructure of the card, while the closed loop is doing the chips themselves. With the middle fan design, some hot air goes out the rear, but a lot spills in the case. While I'm sure the VRMs, etc. get quite hot delivering that kind of power, the chips' dissipation has to be the greater heat output; and that will presumably be going straight out, taking some of the ambient heat with it. This can bode well for the amount of load added to your case. The other question is, how does your 560ti arrange its cooling? Is it the single-blower-exhausting-out-the-back type, or is it multiple fans which dump mostly inside? If the former, you could be in very good shape with both cards running hard. The latter might not be terrible, either, especially if you aim for positive pressure in your case. This encourages more hot air to go straight out through the rear grills. However, there will still be more heat spilled inside, which will reduce the effectiveness of the liquid cooling, as well as that of your CPU. Your thought to try a new dual card alone is probably a very good one. It would probably also be good to wipe the nVidia driver to the best of your ability, get the AMD running right, and then take on adding the nVidia back in. Besides taking it slow on the driver front, this would also give you a good read on the effect the 295X2 has on your system and its cooling. EDIT: On the other hand, a single 120mm radiator for 500 W might still run pretty hot on the GPU itself. I tried a similar liquid cooler on my fx8350 (125-140 W), and promptly went back to the huge-ass air cooler I had been using. The direct exhaust did not compensate for the reduced size of the heat exchanger. I did not keep the liquid cooler long enough to look into the other cooling issues which LaurV mentions. Last fiddled with by kladner on 2014-04-11 at 03:51

2014-04-18, 14:55	#17
tapion64 Apr 2014 5×17 Posts	A quick question. I don't think this plays that big of a part until I want to crossfire cards, but I'm having trouble finding a concrete answer. It seems like for computing applications that don't have much of a CPU bottleneck like trial factoring, the PCIe slot type doesn't matter much for performance. Is this true? I have a motherboard that has 2 PCIe 2.0 slots, one running at x16 and the other at x4. I figure the answer is absolutely need to upgrade mobo if I go crossfire and get atleast 2 PCIe 3.0 x16 slots. But is there a performance increase in that upgrade for a single card setup? Last fiddled with by tapion64 on 2014-04-18 at 14:56

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
What's the best project to run on a Radeon RX 480?	jasong	GPU Computing	0	2016-11-09 04:32
Radeon Pro Duo	0PolarBearsHere	GPU Computing	0	2016-03-15 01:32
Radeon R9 290X Review, new GPU	firejuggler	Hardware	50	2014-01-02 02:25
AMD Radeon R9 290X Hawaii	kracker	GPU Computing	0	2013-09-23 00:14
Radeon hd 77xx benchmarks?	kracker	GPU Computing	2	2012-06-12 15:04

2014-04-19, 01:22	#21
kladner "Kieren" Jul 2011 In My Own Galaxy! 2×3×1,693 Posts	It could, conceivably have made a difference with CPU sieving. But with that happening on the GPU, it seems that traffic would be greatly reduced.