The R9 295 thread - Reviews are in - Quiet, cooler, fast, $1500

[Rvenger]

Lol what are they using to get those numbers? Over 300 watts for a single 290x is just wrong. Furmark?

Look at the standard 7970 and the 7990 power consumption. Something is not correct on this chart.

 

[Kenmitch]

I find it hard to believe AMD designed this card as it seems to be beyond their capabilities....Being a sarcastic Debbie Downer

 

[dguy6789]

So you telling me it draws 275W+ on cables and PSU specced for 150W? That sounds like a really good idea

Also I see both the HD7990 and GTX690 in the charts. Both dual GPU solutions.

Where were you when all of the motherboards and power supplies blew up when the GTX 480 drew waaaaaaaaaaay more power than spec allowed for? Oh wait nothing blew up nevermind.

 

[96Firebird]

edit: plus, very convenient to pick a chart where it's being compared against single gpu configs. Never saw you complain about top CF or SLI power consumption. Also, it's a card for enthusiasts, people who overclock and buy only high end stuff.

Did you miss the 7990 and 690 there?

With Metro (what TPU used for their peak power consumption numbers), the 295x2 was at 500W. I know it won't matter to most, but going over the PCI-E spec for the PEG power connectors at stock is a little concerning to me. A third 8-pin connector would be best, but I can see why they didn't add it (as explained in the AT article).

 

[Gikaseixas]

Did you miss the 7990 and 690 there?

With Metro (what TPU used for their peak power consumption numbers), the 295x2 was at 500W. I know it won't matter to most, but going over the PCI-E spec for the PEG power connectors at stock is a little concerning to me. A third 8-pin connector would be best, but I can see why they didn't add it (as explained in the AT article).

Oops yeah i missed, thanks

Why the concern now if we have cards with a least one 8pin connection inb the market for a couple of generations? I found no review so far sharing your views and concerns. Odd no?

 

[Face2Face]

Obviously it's possible to pull more than spec from a PEG connector, but is it okay for long term? I mean I have pulled 500w from two 8-pin PEG + 16X PCI-E on my GTX 780, but just for benching.. I would like to know how much a 6 and 8 pin are really capable of...?

 

[Leadbox]

PCper review, surprisingly good frametimes in there

 

[parvadomus]

pcper about 4K: "At 4K the SLI configuration suffers a bit more from a lack of consistent frame rates - surprising based on how dominant NVIDIA had been in this area for the past two years. Even though the average frame rate of the Radeon R9 295X2 and two GTX 780 Ti cards are evenly matched, the R9 295X2 only has the frame time variance of a single GTX 780 Ti or a single R9 290X. That is impressive!"
Nvidia getting owned at their own game )

 

[guskline]

Where were you when all of the motherboards and power supplies blew up when the GTX 480 drew waaaaaaaaaaay more power than spec allowed for? Oh wait nothing blew up nevermind.

I agree that this card is not for a budget mb or psu. However, I really doubt that buyers of a $1500 card would slap it in a low end rig.

I'm impressed that AMD could pull this off especially at this price.

 

[DarkKnightDude]

Its crazy how the frametime variance is equal to one card.

 

[96Firebird]

Why the concern now if we have cards with a least one 8pin connection inb the market for a couple of generations? I found no review so far sharing your views and concerns. Odd no?

I have no problem with 8-pin connectors, they are rated for 150W. I have a problem with stock cards going over this spec. I understand if you want to overclock and go over the spec, that is on the end user. Like I said, most people won't care. But I think the spec is in place for a reason...

AT mentions the 2x 8-pin connectors and TDP in their article.

 

[VulgarDisplay]

Why are we still talking about power consumption on a chart that shows all the amd cards using over nearly 100 watts more load that generally accepted numbers from other reviews?

A 7950 draws 171 watts and a 7970 draws 301watts. Seems legit.

These are likely furmark numbers and completely unrealistic.

 

[DarkKnightDude]

Well, the problem with that bench is that its Furmark lol

 

[Creig]

I understand that and corrected my previous post. However, when the GPU temp reported was 70-71C what was the max core? @935. No throttling????:whiste:

Go back and look at the video again. The GPU clocks bounce all over the place from start to end regardless of temps. The GPU hits a solid 70.0c at about 5:45 and I see plenty of speeds above 935 MHz.

I don't know what they're using to keep the GPU at 100% load, but it appears to stressing various parts of the GPU differently since the clocks keep fluctuating. Or maybe it's the fact that they're using a passively-cooled Core "Haswell" processor for this test and it can't keep up. It's hard to say. But the "throttling" we've seen with the 290x has dropped clocks down to the 700 MHz range.

If the clocks had stayed at a solid 1018 MHz until the card hit 70.0c and then started to drop, I would say that would definitely show that thermal throttling was taking place. But what they showed in this video may or may not be.

From the Anandtech review:
"With the exception of our strategy games, both of which throttle at certain points due to power restrictions on both the R9 295X2 and R9 290X, AMD’s latest dual-GPU card is able to maintain 1018MHz on all of our games.

In fact it does so well that we’re unable to determine the card’s base clockspeeds. FurMark bottoms out at 860MHz from power throttling, but try as we might we can’t get it to level out at any other lower clockspeed when introducing throttling scenarios, completely unlike the 290X. AMD’s refusal to publish their base clockspeeds – or even realistic average boost clockspeeds (ala NVIDIA’s boost clock) – continues to disappoint us. But in the case of the R9 295X2 in particular we’re happy to report that the 1018MHz boost clock AMD claims is the clockspeed the card should be able to sustain across all games."

 

[Gikaseixas]

Its crazy how the frametime variance is equal to one card.

Just read PCper review, what a turnaround for AMD

 

[el etro]

About the clockspeeds:




Computerbase "tells yes, the card throttles". Card is designed with a power and temperature limit, but sure you can override it on Powertune settings on Catalyst.

The biggest problem will be achieving overclocks far from 1.1Ghz, since the 2x290x power config(4+1+1) will probably not allow the card go so much high....


Are any AIB here to custom this card for "us"?? :biggrin:

646W draw on cables/mobo designed for 375W. Jeebus...

375w is a recommendation. GTX590 already did past this numbers too. No GTX590 card fried any system up to date.

 

[DiogoDX]

Lol what are they using to get those numbers? Over 300 watts for a single 290x is just wrong. Furmark?

Maximum = Furmark

Average = metro last light

 

[guskline]

Creig, I rewatched the video and agree with your assessment that the core clocks jump all over when reaching 70C but for the most part they hoover around the 935 to 965 mark. I would love to see that test rerun with the radiator equipped with two fans for push/pull. My experience with AIOs is that a push/ pull combo usually drops the temp at least 2 degrees and up to 4 degrees. On this rad, with this card that might affect the core clock. Again, I'm amazed at how well this cooling system does for a 120mm rad.

 

[rtsurfer]

646W draw on cables/mobo designed for 375W. Jeebus...

If you are going to criticize something at least know what you are doing.

The Spec is 375W from 2 x 8 - pin + 150W from the PCI-e slot for a total of
525W.

Edit: This is wrong.
As 96Firebird pointed out.
ShintaiDk ids right.

 

[Creig]

Computerbase "tells yes, the card throttles". Card is designed with a power and temperature limit, but sure you can override it on Powertune settings on Catalyst.

Well, both AT and [H] are saying that there is no thermal throttling taking place.

From [H]:

During our testing we paid close attention to each video card configuration's clock speeds while gaming. We needed to find out if the water cooling on the AMD Radeon R9 295X2 was keeping the clock speed consistent, and how that compared to the two AMD Radeon R9 290X video cards.

Let's start with the two reference cooled AMD Radeon R9 290X video cards running in performance mode. In performance mode the fans run at 55%. The GPUs have a maximum clock speed of 1000MHz. We experienced clock throttling, to different degrees, in every game. Remember, we run our tests after the video cards have warmed up for at least 15 minutes of gaming before doing our run-throughs. The worst offender was Far Cry 3, where the frequency would dip down in the 200MHz+ range; that's right, 800MHz, and sometimes under. In other games it would throttle, but not as severe. The average downclocking we saw was 100MHz below 1000MHz.

With the new AMD Radeon R9 295X2 AMD has set the clock speed to 1018MHz. We reached out to AMD to ask why this number, which sounds like an odd number choice. AMD answered: "Between the Asetek cooling solutions operating temperatures, and striking balance between maintaining top clockspeeds, this is what we came to." In our testing we found that this clock speed was consistent in every game, it never did fluctuate below 1018MHz. This means you can be assured that the specified clock speed is actually the clock speed you will get now. The water cooling keeps the GPUs under 70c, and thus no throttling. We had a very consistent performance from this video card.

 

[Creig]

Creig, I rewatched the video and agree with your assessment that the core clocks jump all over when reaching 70C but for the most part they hoover around the 935 to 965 mark. I would love to see that test rerun with the radiator equipped with two fans for push/pull. My experience with AIOs is that a push/ pull combo usually drops the temp at least 2 degrees and up to 4 degrees. On this rad, with this card that might affect the core clock. Again, I'm amazed at how well this cooling system does for a 120mm rad.

It is impressive how a small radiator can dissipate so much heat if you're willing to allow the loop temp to rise a bit.

 

[el etro]

PCI-e spec can be passed.

You can make a card draw 150w on a PCI-e slot and make a card draw 300W on each 8-pin slot.
You can even make power draw past the 150w/300w/300w on PCI-e connectors, but connectors will warm so much on these conditions.

There are many guys here with GTX780s/Tis/290s/290x with big overclockings. Ask them about their power usage in GPU full stress and system full stress without GPU to make your conclusions.


AMD is preparing a table with PSUs certified to use with 295x2, and this will be launched on their site on the day 295x2 goes Retail.

 

[Attic]

Well beyond my expectations. The cooling solution is really something that goes way beyond what I thought was acheivable on dual 290x.


Of course system design will need to be considered for a card of this performance, but this thing can go into a lot of cases and builds where I thought 2x 290x performance would be impossible. Given that most of the heat from the 295 is being dumped outside of the case it means this thing can go into mid towers and a lot of other options. For dual 290x you either had reference cards that are LOUD, or open air non reference that dump tons of heat in the case, or you have to watercooling. Best options was open air 290x in full tower case with a lot of airflow and low rpm's. The 295 offers a significant option here for getting that 290x performance with a lot more convient options of case choice. I'm more picky on noise than most. Perhaps adding a fan to the rad in a push pull config can get rpms of the default rad fan down to >1500rpm during load.

The first dual GPU card that i'm really impressed with, would be perfect at $1200.

 

[Creig]

Creig, I rewatched the video and agree with your assessment that the core clocks jump all over when reaching 70C but for the most part they hoover around the 935 to 965 mark.

The thing that makes me suspect their test is invalid is that if they were truly maxing out the GPU, the clock should have stayed solid at 1018 MHz. Both AnandTech and HardForum reported 1018 MHz on every game they tested (except for a couple that were power throttled). In the video you linked, the GPU frequency bounces around right from the start. Something doesn't seem right.

 

[el etro]

Well, both AT and [H] are saying that there is no thermal throttling taking place.

From [H]:

Between throttle or not trottle, the truth is: 295x2 trottles, yes. But not as often as 290/290x-deafult/290x-uber throttles.