AMD Radeon R9 380X Nitro Launch Review

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.


I love the same old retarded arguments by people saying AMD power consumption is a fail.....If you really want to look at power consumption Nvida doesn't fair any better. Under full load 980 draws just as much power is 390.....

Even if we don't look at it from the prior perspective, I'm pretty sure people have spent more money on the hardware side of their gaming build than really to care about few bucks more on the electric bill.....

Its a shite argument to say AMD power fail.

And good job spamming 3 posts TZN
 


390 uses 50-100 watts more than a 980 during gaming loads, 390x uses more
390 is 970 level performance
390x is 980 level performance
http://www.tomshardware.com/reviews/sapphire-nitro-r9-390-8g-d5,4245.html
http://www.tomshardware.com/reviews/amd-radeon-r9-390x-r9-380-r7-370,4178-10.html

Though, the power consumption is not really an issue for most anyways nor is it an issue for me.
 
390 consumes quite a bit more than a 980. Power draw is overrated though and most Nvidia fans who complain about it happen to be kids who don't even pay their electricity bill. but if they do, I happen to have that all calculated here handily: http://www.tomshardware.com/forum/id-2547993/psu-tier-list.html

Also, the higher power draw cards get better coolers so heat between a 390 and 970 GPU is generally the same. Dissipation is greater on the 390.
 
Be careful of how you're quantifying heat. While the two cards may stay at the same temperature internally, as you said the dissipation on the 390 is higher. All the heat generated has to be moved somewhere, and usually that means it gets dumped into your case. Depending on the rest of your setup, this can be a legitimate concern.
 


Yes it would be more heat exhausted but that doesn't always play a large role. Usually cooling systems on cards are made pretty well to exhaust the heat out of the computer, and the backplate prevents heat from reaching the CPU in general. CPUs seem to be more of a concern with this than graphics cards IMO.
 
Are you really trying to say heat exhausted from a high-end GPU rarely plays a large role in case thermals? Centrifugal fan coolers do a decent job of exhausting heat outside the case, but almost no axial fan cooler does. Those things dump heat everywhere. Air going down from the card hits the mboard and more or less recirculates back into the GPU again. Air exiting out the front gets sent into your HDDs and your case intake fan, only to get blown back into the case. Air exiting up hits the side of the case and often bounces up and over the backplate into the CPU area.

A strong airflow through the case itself can limit the recirculation, particularly if you've got vented card slot covers. If you have your case and cooling well planned out, it's not always a problem. But to say it's rarely a concern is simply not true.
 


Turkey is correct, having an internal exhaust gpu only increases internal case temps by a few degrees C, depending on how well your case airflow is. Id rather have slightly higher internal case temp than a leaf blower on my gpu that still runs into the 80c and higher before overclocks.
 

Emphasis added. Not everyone runs a massive Enthoo or CM HAF case where cooling and airflow are off the charts. I've got a Raven 3. I've put an axial 290X in my case and it made zero difference. That doesn't apply to everyone. A lot of people have smaller towers, or cases that are primarily about running quiet instead of running cool. ITX is also a popular format right now.

I'm basically saying the same thing, I'm just focusing on the "it depends on your setup" part whereas it seems you are on the "it's not really a problem" side. I think it's better for someone to learn more about a problem that may never happen to them rather than get blindsided by it by not investigating further because someone said "it's not a problem."
 


Even in the smaller itx cases, the difference between internal exhaust and blower styles coolers is around 5c. If you have a case with no airflow, your temps are going to be terrible regardless of the card you selected. A large majority of users have internal exhaust cards aswell, even in itx builds. As long as you have atleast 1 exhaust case fan the internal exhaust vs external exhaust temperature is minuscule in comparison to the actual temp of the gpu, with blower style coolers just having terrible cooling and acoustics. Im emphasizing thats its not a big issue because it isnt simply an issue for 97% of users.
 


This would make a good Toms article.
 

This is true. I have a micro-ATX case, the Bitfenix Prodigy-M, which is actually about as small as most mini-ITX cases that can accept a full size GPU and PSU. I originally had a R9 290 reference in it and it would instantly hit 80c at idle and the blower fan would spin at over 75% all the time which is unacceptable. I had to liquid cool it just to get the core temps under 40c at idle in that case. The same card, in its original reference form, in my Zalman Z9+ ATX mid-tower hummed along at around 35c at idle. It wasn't even hitting 80c under normal gaming loads in that case. The difference is that the Z9+ not only has roughly twice the volume but the components are spaced a lot further apart. It also has a perforated side panel next to the GPU and bunch of extra 120mm fan mounts all of which gives it superior air flow.
 
5 °C may not sound like much, but in a smaller case where every bit matters, that's easily the difference between a sustainable OC and thermal throttling. Really, that can be the difference between simply staying at peak stock clocks and thermal throttling. I promise you, I'm not speaking out of my hind-quarters on this one. I have well documented experience on this. And how are you counting that 97%? 97% of all computer users? Yeah, since the vast majority of computer users aren't interested in a $300+ GPU. However, if you count just the users who want discrete graphics, I wouldn't suggest it's that high. A lot of consumers who are buying the high-draw cards where heat actually matters are the ones looking at ITX, SFF, HTPC, LAN boxes, and other alternate builds to the typical desktop tower.

Cub, if you think the Prodigy M is the average size of an ITX cases that can take a full GPU and PSU, you don't have much experience in that area. The Corsair 250D, Thermaltake Core V1, CM Elite 130 and 120, and numerous Silverstone cases fit those two criteria and are all significantly smaller than the Prodigy M ( and the ITX Prodigy is bigger than all of those except the 250D ). If you allow SFX PSUs instead of ATX units, that list gets even bigger. And you're also talking about the reference 290X, which was maligned as one of the worst cooled cards ever ( though the thing was designed to withstand high temps ). But altogether your anecdote seems to support my case that even if internal temps aren't a problem, they are a concern to keep in mind while building.

The good news is that heat is going down as GPUs get more efficient and uArches shrink. At least we're not dealing with GTX 580s and 6970s right now.

Tell you what, Turkey, as soon as I'm done with the six write-ups that are already in front of me, I'll volunteer to test this empirically ( though really Thomas and Chris would be better equipped for it ).
 
Correct me if I'm wrong, but isn't a non-reference blower cooler that pulls air from both sides of the card a better alternative to the open air coolers than the standard blower coolers, or is that just marketing fud?
 


Honestly, people who complain about AMD power draw, I find it hilariously stupid. If you spend that kind of money on your build and gpu, and you worry about an extra 15-25$ on your electricity bill, I mean dude you got your priorities wrong.
 


yep.

that said I think there is one huge advantage to the lower power draw on the nvidia cards. It makes running a high end gaming system at <30dbs very achievable, because the amount of power needed to be cooled in those low powerdraw cards is much much lower then was reasonable to expect in the past.

Currently my system is semi-passive, with just 1 fan running 24/7, the cpu fan; everything else is temp monitored, and only turns on at certain temps, including the gtx970's fan.

The result is unless i'm gaming, most fans don't turn on at all, and even then most fans run at the bare min, and temps stay nice and cool throughout the system all sub audible (seriously, you could put your ear up to the system and never hear it). This type of build would be impossible with an AMD.
 


I guess, but it also depends on the manufacturer and the design. I have an Asus card and generally the fan is around 40% under OC load. So its not that noisy. But nvidia cards do tend to overclock higher because of better power delivery design.
 


it's an MSI Gaming G4 gtx970.

I chose it due to the size and quiet of the fans. Then using MSI afterburner i changed the fan profile to turn off the fans under 50C. its amazing how rare they turn on. It's a darn near semi-passive gpu~
 
I just found this gem:
http://www.geforce.com/hardware/desktop-gpus/geforce-gtx-960-oem/specifications

GTX 960 OEM, a cut-down GTX 970. It brings the core count down to 1280 and the memory bus down to 3GB with a 192 bit bus (no memory tricks like on the 970's bus), granted the memory frequency is rather low. Its wider memory bus should still help it compete with the Radeon 380X, unlike the GTX 960 which is so memory bandwidth bottlenecked that the 950 can almost keep up despite having a GPU that is almost 30% slower.
 


Scroll up this thread to the post where I mention the "960 desktop" that is found in the Zotac Steam Machine. That is what Nvidia is calling the GTX 960 OEM. It is just a GTX 970m. They didn't even bother to bump up the stock clock speeds. The reason it is being called a 960 is because the performance is so similar to a regular 960 2GB even though the 960 OEM looks so damn good on paper. In fact, in some games, the 970m/960 OEM is actually slower than a regular 960 2GB. Check the specs and then scroll down to a whole bunch of benchmarks here: http://www.notebookcheck.net/NVIDIA-GeForce-GTX-970M.126694.0.html As you can see, the specs across the board are the same and it makes sense since the Zotac Steam Machine is so small it would need a laptop GPU. Then, compare the benchmarks and games to the 960 2GB: http://www.notebookcheck.net/NVIDIA-GeForce-GTX-960.146738.0.html As you will notice, in synthetics, the 960 2GB tops the 960 OEM/970m in almost every test except for one instance where the 970m was tested with a laptop with a Skylake i7 quad core, 32 GB RAM and updated drivers in Firestrike. The 960 2GB was only tested twice on older drivers in the same test. In the majority of games, the two cards are either dead even or some games (like Just Cause 3, for example) the 960 2GB is faster. The only advantage that the 960 2GB has is clock speed. This is why I believe Nvidia didn't bother to cut GM204 down any further on real desktop PCIe GPUs and call it the 960ti. They would have either had to reduce the price of the 960 or charge something like $250 for the 960ti. Either way, the cheaper 960 would have been able to match a 960ti with just a slight overclock, maybe ~200 MHz which is the difference in core clock speeds between the 970m/960 OEM and the 960 2GB.

There is another cut down GM204 in between the 970 and 960 OEM and it is the GTX 980m with 1,536 CUDAs that was also rumored to be released as the 960ti. But, my suspicion is that the difference in performance between that and the 970 desktop is too small as well. A 128 CUDA difference is too close and this card would have certainly cut into 970 sales especially if it was released with the 256-bit 4GB memory config like the 980m. Even with a lower TDP and core clock to keep power use low (~125 MHz difference), the 980m is only 3 to 4 FPS off of the 970 in GTA V ultra 1080p; http://www.notebookcheck.net/NVIDIA-GeForce-GTX-980M.126692.0.html. If you bumped the clock speed up to match a stock 970, it is probably going to be about even in most games. If you OC a desktop 960ti that is built on the 1,536 CUDA chip by 150-200 MHz, it would probably beat a 970 in most games. Nvidia couldn't have that. This scenario is what is happening with the R9 380 and 380x.
 
Alright, so its basically just a 970m. Still, all that needs is higher clocks and it would fit the bill perfectly. Its 192 bit bus is low enough to not compete with the 256 bit bus on the 970 while being stronger than the 128 bit bus of the 960 and the GPU core count is a good enough improvement from the 960 without getting too close to the 970. Besides, the memory bus is what would make most of the difference. Increasing the memory bus' frequency would make all the difference. Increasing the GPU frequency would help too don't get me wrong but that isn't the main bottleneck for the Maxwell cards in most situations. Comparing the GTX 950 to the GTX 960 proves that outright. Dropping the GPU performance by about 30% got a generally less than 10% performance reduction.

As for the scenario you give with the slightly lower card besting the higher card with overclocking granted it would have been to an extreme with the 980m, this is still very common, granted Nvidia did do a lot more to prevent it this generation than they did in the past. This was happening all over with Kepler. Using the 970m as the base for the 960 Ti instead of the 980m would have prevented this, just like Nvidia would want. It would still cut into 970 sales because of people who want more than a 960 but didn't necessarily want a 970, yet went for it anyway because lack of other choices, but that can go both ways because people who wanted more than a 960 but wouldn't pay for the 970 might go for this 960 Ti.

The 960, with overclocking, could not really overtake a desktop-clocked 970m. The 960 just couldn't best a 50% memory bandwidth advantage with the same architecture and a 25% core count advantage with similar clocks. Overclock both cards for fairness and the 960 wouldn't come close. Overclock the desktop 970 in this example and the 960 Ti and it might come somewhat close, but the 970 would still command at least a 20% performance advantage.
 
Status
Not open for further replies.