The Cause Of And Fix For Radeon R9 290X And 290 Inconsistency

Page 9 - 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.

slomo4sho

Distinguished


Oh your posts reek of ignorance. 1GB of vram is a bottleneck for modded Skyrim. Having two 1GB cards still only provide you 1GB of vram for the system as the vram is shared and the lowest ram card is the baseline for the other cards. Also I can guarantee that, the GTX 660 or the HD 7870 would provide equal or better gaming performance at a lower cost at the time of the purchase. PEBKAC is typically projected on the hardware :)
 

anthony8989

Distinguished


I can attest to the GTX 660 being better than the 7790 crossfire, on account of the 7790 crossfire was completely flippin' broken. And FYI, I don't need a crash coarse in graphics cards tyvm. The problem was not VRM at 1080p, the bottleneck fell on the fact that the second 7790 was completely and utterly useless in crossfire due to incredible amounts of micro stutter.

http://us.hardware.info/reviews/4301/15/amd-radeon-hd-7790-2gb-review-does-another-1gb-make-a-difference-the-elder-scrolls-v-skyrim---1920x1080-+-frametimes

OOooooooOOoOoo One whole frame per second at 1080p from 1 gb to 2 gb. You sure know what you're talking about.
 




The 1GB VRAM wasn't the only problem.
6 months ago, Xfire was broken till AMD provided the frame pacing driver in order to fix crossfire problem. So even if he had more VRAM it wouldn't help a lot... That wasn't solved until 2 months ago:

Dual-GPU Battle: Does Frame Pacing In Catalyst 13.8 Turn The Tide? By Don Woligroski September 3, 2013 10:00 PM
 
I disagree with the findings of the article in regards to blame, and jimmysmitty's assertion that Tom's had anything to do with helping AMD to implement their change in the way the driver software handles the fan speed. AMD didn't like the results they saw from board partners, and informed Tom's about the fix after the fact to help dispel the negative press Tom's was spreading, not because Tom's was working with AMD. No company likes bad press, and as Tom's was the squeaky wheel this time, they are a site which was addressed about the fix. There is no current heat issue with the R9 290 and 290x cards and there was no heat issue previously. The software and hardware were working according to how they were designed to work. That board partners chose to implement fan solutions that were of differing specifications to the parts on AMD's press models, is not AMD's fault, but AMD chose to implement a solution to address the situation anyway. Tom's focused on AMD, rather than the suppliers of the supposedly underperforming boards, which is misleading, as AMD is not responsible for the products of other companies.
 
AMD's solution to the alleged problem brings up a question I haven't heard asked yet. What happens when aftermarket coolers start appearing on R9 290 and 290x cards, that have vastly different airflow rates for the RPM targets that AMD is setting in their driver software? Hopefully the manufacturers of cards with those coolers have some control at their disposal to work with the target AMD has set in their driver software, otherwise we're going to see unintentional behavior all over again. The fix only seems suitable for the reference cooler, but how is the driver supposed to know if you're running a reference cooler or not? Hopefully that's already been addressed behind the scenes at AMD.
 

technoholic

Distinguished
Feb 27, 2008
800
0
19,160
I wonder if AMD implemented some hardware level solution on the new R9s to fix the frame pacing issue. Or are they only working on the software side? I hope Tom's investigate frame pacing on multi card setups soon
 

slomo4sho

Distinguished


I never said 1GB being a bottleneck at stock settings did I? However, ram does become an issue when Skyrim is modded. The comparison of performance difference would actually be similar to BioShock in that review since it is one of the few games that actually surpasses 1GB of ram usage at 1080P at stock settings for which there was a 5.6% performance increase going from 1GB to 2GB. You keep bringing up crossfire even though my initial argument was that your decision to crossfire two low end cards was not smart to begin since reviews already illustrated that the card was outperformed at 1080P with single card solutions at the same price as the two 7790s. I guess you couldn't be bothered to read crossfire reviews that were available on the release date of the 7790 and make an informed decision. Again, having a clue about what you are talking about goes a long way.

Stay on topic? This thread is nothing more than fanboy rhetoric from both sides as the article is useless to basically everyone who would actually be in the market for a 290 or 290X. The entire issue revolves around the cooler which has been declared a poor match for the GPU in every review to the point that cometary of the issues revolving around the cooler is moot. Wait for non-reference designs, go water, or add an aftermarket cooler if you are in the market for a 290 or 290X


Both the 290 and 290X have hardware based solutions for frame pacing and it seems have resolved the issues of crossfire.
 

GOM3RPLY3R

Honorable
Mar 16, 2013
658
0
11,010


+1
 

GOM3RPLY3R

Honorable
Mar 16, 2013
658
0
11,010


Actually he was right. If you do the math it comes out to a few numbers near ~15%. Otherwise I'm with him on that one. I'm probably getting another 770 when christmas comes around as well. :wahoo:
 

GOM3RPLY3R

Honorable
Mar 16, 2013
658
0
11,010


http://www.youtube.com/watch?v=bfAzZjzVcXE

Also, can you send us a picture or two with your GPU in your case? I'd love to see it. ^_^ Maybe some benchmarks while you're at it. :lol:

 

anthony8989

Distinguished
I never said 1GB being a bottleneck at stock settings did I? However, ram does become an issue when Skyrim is modded. The comparison of performance difference would actually be similar to BioShock in that review since it is one of the few games that actually surpasses 1GB of ram usage at 1080P at stock settings for which there was a 5.6% performance increase going from 1GB to 2GB. You keep bringing up crossfire even though my initial argument was that your decision to crossfire two low end cards was not smart to begin since reviews already illustrated that the card was outperformed at 1080P with single card solutions at the same price as the two 7790s. I guess you couldn't be bothered to read crossfire reviews that were available on the release date of the 7790 and make an informed decision. Again, having a clue about what you are talking about goes a long way.

That was a cool story bro. I especially liked the part where you presume to know what settings and specifically what mod I'm using in Skyrim. I also liked how you went on to presume that the poor quality of game play I experienced in Skyrim was caused by VRM and ignored the fact that it was in more than just Skyrim, and up until a couple months ago crossfire was broken no matter what cards you paired together or how much VRM you had at your disposal. My favourite part of all though, is where you assumed I didn't troubleshoot and discover the problem. One 7790 ran me average 44 FPS and two 7790's gave me nearly double at 79 FPS according to FRAPS - but you wanna know the funny thing? It sure as hell didn't look like 79 FPS on the monitor. The single 7790 ran smoother. Doesn't sound like a VRM issue to me, but what do I know right?
I know you love jumping to conclusions, but FYI not everyone has the same budget or even the same build process you do. It being a time before FCAT where we all went by FPS apps to judge a GPU, I saw an opportunity to nearly double the graphics power of my pc for an end cost of $90. Giving me comparable performance to a 7870 GE which at the time was A LOT more expensive than $230 in my region. I don't care if AMD "fixed" the problem. I bought the cards and experienced the problem in 90% of the games I played and said F*** it to AMD and their crap drivers.

I'd rather you not respond to me. Your attitude and AMD's driver quality are a perfect match.
 


Actually review sites tend to work with the hardware devs all the time and since THG is one of the largest, they have a lot of connections to the biggest hardware vendors. It is not unlike THG to notice something in their testing, before the NDA lifts or any benchmarks are out, and for them to consult with them. They release their current findings and if they get updated they update the article or release a new article.

THG is slow on news but their reviews are normally some of if not the first and they have been around long enough to know how to handle these situations.

And no one is saying there is a heat issue. Rather that the GPUs run warmer than people are used to (up to 95c yet most HD7970s barely hit 80c bitcoin mining) and due to that extra heat they are not running at the clock speed AMD wants.

Now AMD might have done this on purpose knowing that NVidia was on the verge of the 780Ti and wanted to put the 290/290X out beating the 780 and matching Titan so NVidia releases the 780Ti thinking it has an advantage. next thing we know AMD releases a better cooler or the AiB coolers are more than enough and we see the 290/290X push the 1100MHz+ clock speeds and start to whomp on NVidia.

Its hard to say what will happen but what is not hard to say is that the reference design is not efficient enough to consistently keep the 1GHz clock speed AMD is saying its supposed to run at.
 
I understand about Tom's being a big review site, and having better connections than the average lay person. My comment was solely in regards to this article. Sorry if I wasn't clear in that respect. :) I highly suspect AMD may have made their changes in response to Tom's big clamor, but didn't work with Tom's in the process.

At 40 - 50% fan speed, I would agree with you about the reference cooler being insufficient. I take offense with Tom's not also benching the card with the cooler turned up to adequate speeds. When in the past has anybody else set an arbitrary limit on their fan speed and then complained their chip got too hot? Just because NVIDIA has a more elegant cooler design, doesn't mean AMD's cooler doesn't work. It's just a noisier solution, that's all. People can nit pick this all day long, but the cooler does work. I suspect Tom's hasn't shown any charts with the cooler running at adequate speeds, as there would be no controversial story to report then, other than the earplug ads that may start showing up along with the article. (-:

It only makes sense that the R9 290 and 290x run warmer than people are used to as they have more transistors than people are used to. A proportional increase in the fan speed to deal with the proportional increase in work being done by the chip, over the 7970 seems logical to me, since the fan is pretty much the same thing AMD has been using for a while.

Ultimately, the article highlights for me the differences in what you get from one manufacturer to the next. It's a great example, actually, in cost cutting by board partners effecting the final results a consumer will see. I think it's good that AMD allows the board partners to make their own decisions. It allows me as a consumer a wider selection of choices, whether I am more concerned about price or overall performance.
 

Gurg

Distinguished
Mar 13, 2013
515
61
19,070


The AMD fix is only to standardize the Quiet and Uber modes. The fan speed can be set separately and reports say the throttling goes away on the reference card at a fan speed of around 76%.

 

anthony8989

Distinguished


I get your point dragon, and it's a very good one. But man to crank that fan up high enough that the card can get to its performance threshold:
http://www.youtube.com/watch?v=QQhqOKKAq7o

Like slo said, if you want this card (to function at its max potential) you gotta get an after market cooler or be okay with a leaf blower in your pc. That's basically the bottom line. What most people are picking at is the fact that this is a +/- $500 flag ship card and this kind of thing just makes nVidia look that much better.
 
Thanks for the response, Gurg. I realize you can easily set the fan speeds manually, but I'm wondering what the automatic standardization of RPMs is going to do on cards where the user never manually adjusts fan speed? I suspect it could lead to a situation where some aftermarket coolers vicariously cause clock throttling again, if the fan has a lower airflow for the given RPM.

76% sounds like a pretty realistic number to me. Any idea if this is the R9 290 or the 290x? At 100% then, you might even be able to squeeze a little overclock out of even the reference cooler?
 


I would like to see them at those speeds too but honestly, my Vapor-X at 75-100% is pretty loud so I doubt this would be tolerable at those speeds.

I still think the AiBs will be the best for overclocking and getting the stock speeds.

I am honestly surprised though that AMD did not push a better reference cooler to help obtain those stock clock speeds (1000MHz). It seems like they should right now just advertise the actual speeds because people wont be happy when the GPU only this 915MHz and is supposed to be at 1GHz.

I do have hopes though. I think one at 1GHz will be a very good performer if it can stay around the 70-80c range and never throttle as I think throttling probably doesn't help frame pacing.
 
Can't disagree about the noise, Anthony, but I've actually been a bit disillusioned with PC cooling the last few years anyway. There are so many different ways manufacturers could be dealing with the heat that it's no longer very funny. I mean, to force a graphics card manufacturer to burn as many watts as they do, to generate the graphics we expect, and yet keep their silicon cool in such a small footprint, seems a bit absurd.

Take the BTX standard for example, it was an attempt at alleviating some of the heat problems in PCs. If that could be expounded upon in a reasonable fashion, I think leaf blowers would be a non-issue.

In the mean-time, I have mitigated the problem by just removing my computers to another room and running high quality cabling from them to my console. Problem solved, regardless of the noise and heat that may be being generated. It goes somewhere else. I understand not everybody has the luxury of an unoccupied room to move their equipment to, so better ways to deal with heat would be nice.

When is somebody going to design an aftermarket cooler for graphics cards that ducts to the outside of the chassis, so it can actually draw from cooler, outside air, rather than the warmed air inside the computer case? I can see something like this being especially beneficial in dual card configurations, where heat becomes ridiculous in short order.
 

anthony8989

Distinguished


Yeah we do the same thing with high-end audio equipment. A small room the size of a closet is devoted solely to amplifiers, receivers etc., to keep the audience away from the noise and heat. You've got a nice setup from the sounds of it. Well done.

You're right dragon, it is a lot to expect engineers to devise new and innovative ways of distributing heat away from their ultra high powered chips in such tiny confinements. But AMD's chief competitor is doing a far better job at it. I can't help but feel that if AMD didn't have these cooling and noise issues, and the R9 290x had more OCing head room without the reality of spending more $$$ on an after market cooler, or letting board vendors do the heavy lifting of properly cooling their cards, that the GTX 780 TI would be $100 cheaper and within reach. nVidia's flag ship does more with more consistency and less noise and on a stock cooler - or any cooler really. And sadly that translates into it being stupid-expensive. They charge an arm-and-a-leg for it because they can.
 


Your own stupidity for XFiring lower range card before doing your research...

EDIT : No offence. :)
 

anthony8989

Distinguished


Has nothing to do with the lack of graphics power from the dual 7790s. Next time try reading my whole post. Fraps was showing 79 FPS +/- . Microstuttering from poor drivers was reducing perceived smoothness. Kthx
 


Did you know about the Microstuttering beforehand?
Did you know its worse the smaller the memory bus on the card is?
Did you know that generally Xfire has ALWAYS been a bad idea on low to mid range cards?

Pretty sure its yes to at least 2 of the above. In which case, you should have known better.
 


Precisely, blaming the manufacturer because you cut yourself with the razor sort of thing.
 
Status
Not open for further replies.