AMD Radeon R9 290X Retail Performance Variance Review @ [H]

FrgMstr

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,601
AMD Radeon R9 290X Retail Performance Variance Review - The AMD Radeon R9 290X arrived recently with a high level of performance, and a high level of controversy. There have been reports of performance variance between Radeon R9 290X video cards. We have two purchased retail cards today with stock cooling that we will test and see if performance variances exist.
 
First of all - I love these Sunday posts - something new to read after a long weekend.
It seems to me that these variances with the retail card #2 are pretty small and would fall into just a range that you would expect to see across different venders. Would we see the same variance with a retail Nvidia 780. The review that I am really waiting for is for a non-reference cooler. I think it will be really revealing and will give us a lot of insight into the true capability of the 290 series.
 
Last edited:
TLDR Summary:

Yes there is some variance, but no its nothing big.

Perhaps AMD just needs to take a marketing lesson from this.
 
Ultima99 said:
Perhaps AMD just needs to take a marketing lesson from this.
Click to expand...

Yep. AMD shot itself in the foot with the "up to" marketing. The Quiet and Uber mode thing hasn't really helped them either. Quiet and Uber mode should just been driver toggles and not a switch soldered to the PCB. The card running at where a lot of people consider extremely hot has not been great for it's image either. I still see people talk about how it's "too hot" and all that jazz. At least custom cards handle a lot of the complaints.
 
My two msi 290Xs run exactly the same.
I have them set to msi afterburner 1125/1350 and they perform identically.
Nice article.
 
Would it be worth removing the heatsink and double checking to see whether or not giving the card(s) a high quality aftermarket TIM makes a difference in either operating temperatures or reduction of throttling vs out of the box? As hot as these cards run, I can imagine that every little bit helps as far as thermal conductivity goes.
 
Slash3 said:
Would it be worth removing the heatsink and double checking to see whether or not giving the card(s) a high quality aftermarket TIM makes a difference in either operating temperatures or reduction of throttling vs out of the box? As hot as these cards run, I can imagine that every little bit helps as far as thermal conductivity goes.
Click to expand...

Unless they're using cheese for the TIM I doubt it would be a huge difference. The cause of the heat "issue" is the stock cooler. AMD used a cheap cooling solution and designed the card to withstand the heat their cooler produced. The custom and 3rd party solutions out already have significantly lower temps.
 
Meh, pretty negligible difference, but still an interesting find. Of course, you know how easily these things get blown out of proportion.
 
sedgiemon said:
Brent - long time reader thanks for the article, I've been holding off on a 290x until this had been confirmed, or at least the aftermarket cards had taken care of the heat/noise issues.

The way I see it however, is that the problems may be be more prominent for people who:
a.) do not use an open test bench
b.) possibly don't have A/C and cannot keep the ambient temperature at 20 celsius the whole time.

We know that these cards potentially throttle with excessive heat, and have seen that across various review sites. To me, a more interesting comparison would have used a higher ambient temperature and maybe a couple of different cases with various airflow. What may not be an issue on the open bench, could prove to have more of an affect in a case (which, let's face it, is what most of us use).

furthermore - throw a couple together in crossfire in a case, with say 28 degree ambient temperature, and then compare the performance.

This is honestly what has stopped me buying 1 or 2. I really do believe that a gtx 780 may show these up in those sort of 'everyday' operating conditions.
Click to expand...

A case with proper air flow should perform better then an open bench since cool air is being directed over the components. Proper air flow will also help "negate" higher ambient temps, if you don't have fresh air coming into your room you have bigger problems then a $500 video card.
 
sedgiemon said:
Brent - long time reader thanks for the article, I've been holding off on a 290x until this had been confirmed, or at least the aftermarket cards had taken care of the heat/noise issues.

The way I see it however, is that the problems may be be more prominent for people who:
a.) do not use an open test bench
b.) possibly don't have A/C and cannot keep the ambient temperature at 20 celsius the whole time.

We know that these cards potentially throttle with excessive heat, and have seen that across various review sites. To me, a more interesting comparison would have used a higher ambient temperature and maybe a couple of different cases with various airflow. What may not be an issue on the open bench, could prove to have more of an affect in a case (which, let's face it, is what most of us use).

furthermore - throw a couple together in crossfire in a case, with say 28 degree ambient temperature, and then compare the performance.

This is honestly what has stopped me buying 1 or 2. I really do believe that a gtx 780 may show these up in those sort of 'everyday' operating conditions.
Click to expand...

Here's my take on people that buy cheap cases that have terrible air flow. You get what you pay for. Don't blame the processor or the video card for overheating in your $19.99 computer case. Blame yourself for cutting corners in your system design. Same thing for the guys that build a $3,000 rig and hook it to a 15ms or worse monitor and cry that something is wrong with their components. No, you bought a cheap monitor with high latency and that's why you're getting shot and killed before you can see your opponent. Don't blame the game. Blame yourself. :)
 
sedgiemon said:
Brent - long time reader thanks for the article, I've been holding off on a 290x until this had been confirmed, or at least the aftermarket cards had taken care of the heat/noise issues.

The way I see it however, is that the problems may be be more prominent for people who:
a.) do not use an open test bench
b.) possibly don't have A/C and cannot keep the ambient temperature at 20 celsius the whole time.

We know that these cards potentially throttle with excessive heat, and have seen that across various review sites. To me, a more interesting comparison would have used a higher ambient temperature and maybe a couple of different cases with various airflow. What may not be an issue on the open bench, could prove to have more of an affect in a case (which, let's face it, is what most of us use).

furthermore - throw a couple together in crossfire in a case, with say 28 degree ambient temperature, and then compare the performance.

This is honestly what has stopped me buying 1 or 2. I really do believe that a gtx 780 may show these up in those sort of 'everyday' operating conditions.
Click to expand...

My personal 290X CrossFire experience article hit at the heart of that.

http://hardocp.com/article/2013/12/13/4_weeks_radeon_r9_290x_crossfire/
 
Hmmm I thought I read somewhere that Gigabyte 290's and 290x AIB cards were recalled over heatsink failures.

You guys could of got one of those....let me see if I can find it.

edit: Here it is, take it with a grain of salt?

http://videocardz.com/48592/gigabyte-halts-radeon-r9-290x-windforce-production-due-design-flaw

Gigabyte:

There is a problem with a heat sink. The production sample is different from the sample provided by our vendor. The sink for early media samples are not optimized to meet the design requirement. We will improve it immediately and we stop the mass production today.
 
DASHlT said:
Hmmm I thought I read somewhere that Gigabyte 290's and 290x AIB cards were recalled over heatsink failures.

You guys could of got one of those....let me see if I can find it.

edit: Here it is, take it with a grain of salt?

http://videocardz.com/48592/gigabyte-halts-radeon-r9-290x-windforce-production-due-design-flaw

Gigabyte:

There is a problem with a heat sink. The production sample is different from the sample provided by our vendor. The sink for early media samples are not optimized to meet the design requirement. We will improve it immediately and we stop the mass production today.
Click to expand...

The problem is with the Windforce coolers; [H] used reference coolers.
 
From Brent's Thoughts in the conclusion:
Brent said:
With NVIDIA GTX 600 and 700 series the video cards are "running faster than advertised" and with AMD R9 290X the video card is running "slower than advertised."
Click to expand...
This is why I pay more for Nvidia cards.
 
evilsofa said:
From Brent's Thoughts in the conclusion:

This is why I pay more for Nvidia cards.
Click to expand...

You do realize though with Nvidia cards, that you are limited to how much power you can send to it, and it downclocks (just like AMD) when you use too much power/heat?

It's the same damn shit.

So you are paying more for something both companies do.
 
sedgiemon said:
I don't know if i agree... I have a $500 Lian-Li case and have fitted all 4 x 120mm fans with Noctua case fans and have a dirty big noctua cpu cooler (can't remember the exact model). My cards still run SIGNIFICANTLY cooler with the side of the case OFF. At the end of the day, the more cool air that the card can intake the better, and i believe that with SLI/Xfire setups this is particularly true.

Thanks Kyle i'll have another read. I personally have a good case, and A/C etc. But sometimes i feel like playing with the fan on and the windows open. I'm in Australia so it gets pretty damn hot. I just feel safer going with Nvidia cards and that they'll perform regardless? Maybe it's just me.
Click to expand...

I'm in Beerwah in the Sunshine Coast hinterland - on Saturday I recorded 42.8C at 3pm with about 14% humidity so I had the aircon running at a 26C target but it was a constant 28.8C in my office (infrared thermometer gun).

I run a stock 290 non x with MSI afterburner in an Antec 900 case with an overclocked FX8320 @ 4100 on air. I have profiles set for different uses and I game with headphones, my case is on the desk only 40 - 50cm away from my left ear.

Normal use and browsing I run stock clocks and 45% fan, as to me it's no louder than the rest of the case fans but I like to keep stuff as cool as possible. Older, less intensive games like X3-TC I use the same profile and see max temps on GPU of around 55-68C depending on ambient. If I'm playing Crysis 3 I ramp fan to 75% and game at 1100 / 1325 and never break 80C, usually 75 or so on GPU.

In a month or so when prices start to settle I'll go a full water loop on this card with an XSPC block, but for now it's ok. Noise is subjective and to me being a motor mechanic by trade temps are a bigger concern than noise. To me high performance usually is noisy haha.

Your preferences may be different but at the end of the day all the FUD that was spread over these cards with regards to throttling and noise was pretty clever, but IMO there isn't an issue at all - it's an enthusiasts card and isn't ever going to be silent (unless you go full water - but even then you have a pump and radiator fans).
 
evilsofa said:
From Brent's Thoughts in the conclusion:

This is why I pay more for Nvidia cards.
Click to expand...

So if AMD said "Base clock of 600Mhz, boost clock up to 1Ghz!" you would buy AMD? I guess Nvidias marketing really does work.
 
Pretty sure the power limit is too marginal to account for silicon variance. Some GPU chips draw more power than others. I have 10 7970s at identical settings and you see up to 20 watts difference between them. And that's undervolted, the differences can be bigger at stock volts. Same with CPUs anyway. Some simply draw more current. The higher current chips don't necessarily run hotter or clock differently, but with the default power limit so close to what the chip actually needs this is what you end up with.
 
darkstarcow said:
So if AMD said "Base clock of 600Mhz, boost clock up to 1Ghz!" you would buy AMD? I guess Nvidias marketing really does work.
Click to expand...

It comes down to certain people perception if you checked out the whole GTX 680 to GTX 670 reviews then you would see improvement that scales up to the card in the higher bracket. Back then it was a funny read , why would Nvidia release cheaper cards that have almost the same performance as their more expensive ones. And that would happen to the 660TI as well it is weird that people call that marketing. Of course the cards were slower but that was hardly noticeable at 1080p.

Reality is that AMD has "problems" with the default cooler if anything it teaches them with this series of cards that it is just under performing instead of being decent. How much of variance is the cause of the heat and how much of it is the chip?

It seems that so far the AIB version does not suffer the same problems...
 
Much ado about nothing...

Marketing ... Marketing ... Marketing ...
 
This marketing discussion is interesting, but I think it is also distracting. As an end user, what I want is the card that gives me the best experience for the best price with my current monitor. I don't really care if it's throttling up, down, running at 600Mhz or 1000Mhz - as long as I can turn up the settings and resolution and get smooth gameplay. In this review, even the 290x #2 that performed lower had the same "experience" for the end user. I would bet that a 290x running in uber mode would be quieter than my 2 680s in SLI and would probably give me the same experience. It would probably take a 780ti for an extra $200 to get me there on the Nvidia side.
 
As someone who owns a pair of R9 290's in CF and under water I do not see this variance. My cards are happy at 1150 x 1350mhz 24/7 and run at these frequencies all the time (when gaming) as the temps are around 60C on each card.

Perhaps I've been lucky, it appears to me that AMD should have been a little smarter with their marketing campaign and set a minimum expected clock and then a suggested boost clock, like Nvidia.

Also the variance is linked to the temps and the throttling that the card undergoes as it approaches the 94C mark. Again AMD should not have cheaped out on the stock cooler...

When a decent cooler is used the variance tends to disappear and can easily be explained as just regular manufacturing discrepancy.

Just my 2 cents worth.

I am extremely happy with my R9 290's and look forward to my first 28inch 4K monitor :)

As always a nice read on Monday morning.
 
I believe there are two different types of memory installed on the 290x. I would be interested to see if one of the two gigabyte cards has different memory than the other.
 
RedLava said:
I believe there are two different types of memory installed on the 290x. I would be interested to see if one of the two gigabyte cards has different memory than the other.
Click to expand...

The cards are 1 serial number apart and I've already put them in my daily driver... I'll check for you when it is time to upgrade :D:D:eek:
 
kyle, brent

could you give us the asic sore from gpuz on all 4 of the cards?
 
trick0502 said:
kyle, brent

could you give us the asic sore from gpuz on all 4 of the cards?
Click to expand...

Brent has press sample #1 right now, here's the other 3.

Press Sample #2 - 77.3%
Gigabyte #1 - 73.5%
Gigabyte #2 - 80.1%
 
AMD should have just marketed the cards at 850MHz, then people will be like WOW it runs at near 1GHz a lot of the time!! What great value... gold award, platinum award, editor's choice award, etc etc etc ... :p

Either way, the non-reference coolers are definitely the way to go, as their performance is "how the card should have behaved from the get-go".
 
darkstarcow said:
So if AMD said "Base clock of 600Mhz, boost clock up to 1Ghz!" you would buy AMD? I guess Nvidias marketing really does work.
Click to expand...

No, the difference is that nvidia advertised x, and instead people got x+1.

AMD advertised x+1, and instead people are getting x.

It's the same as getting a DSL line that runs up to 10mbps, and only getting 8mbps, and getting an 8mbps cable line you actually get 10mbps out of. If I buy purely based off of the marketing, I know that I'm going to get at least 8mbps from the cable line. From the DSL line I might get 8, maybe 7, could be 6, who knows? People have an expectation, and they want their expectation to be met. If the company can exceed that expectation, great. If they cannot meet that expectation... people get pissed.
 
I have a GTX 780 Ti, I can't stand Boost and I couldn't stand powerplay when I had a AMD. I just want to say this issue is on both Nvidia and Amd.. I have disabled boost on my 780ti with a custom bios so I get what I paid for.

Now I will say I have a Galaxy reference gtx 780 ti and on my stock bios my boost clock was 1020mhz instead of 928 and by no means complaining but is it typical for Nvidia cards to come with different boost clocks depending on ASIC quality?
 
ApeXiTT said:
I have a GTX 780 Ti, I can't stand Boost and I couldn't stand powerplay when I had a AMD. I just want to say this issue is on both Nvidia and Amd.. I have disabled boost on my 780ti with a custom bios so I get what I paid for.

Now I will say I have a Galaxy reference gtx 780 ti and on my stock bios my boost clock was 1020mhz instead of 928 and by no means complaining but is it typical for Nvidia cards to come with different boost clocks depending on ASIC quality?
Click to expand...

All cards are different, NVIDIA boost clocks can vary widely.

On the AMD R9 290X the card will run as fast as it can keeping at 95C at 50% fan.

Again each card will be different.
 
Pieter3dnow said:
It comes down to certain people perception if you checked out the whole GTX 680 to GTX 670 reviews then you would see improvement that scales up to the card in the higher bracket. Back then it was a funny read , why would Nvidia release cheaper cards that have almost the same performance as their more expensive ones. And that would happen to the 660TI as well it is weird that people call that marketing. Of course the cards were slower but that was hardly noticeable at 1080p.

Reality is that AMD has "problems" with the default cooler if anything it teaches them with this series of cards that it is just under performing instead of being decent. How much of variance is the cause of the heat and how much of it is the chip?

It seems that so far the AIB version does not suffer the same problems...
Click to expand...

What problems are they having with their cooler?

Merc1138 said:
No, the difference is that nvidia advertised x, and instead people got x+1.

AMD advertised x+1, and instead people are getting x.

It's the same as getting a DSL line that runs up to 10mbps, and only getting 8mbps, and getting an 8mbps cable line you actually get 10mbps out of. If I buy purely based off of the marketing, I know that I'm going to get at least 8mbps from the cable line. From the DSL line I might get 8, maybe 7, could be 6, who knows? People have an expectation, and they want their expectation to be met. If the company can exceed that expectation, great. If they cannot meet that expectation... people get pissed.
Click to expand...

I will say it again since you seem to be having a hard time understanding what I said "If AMD had advertised their cards as having a base clock of 650Mhz, and boost clock of 850Mhz, would you buy them instead?" See what I did their? AMD isnt delivering anything less then Nvidia, their marketing approach is just retarded.

Your example is also completely irrelevant since AMD states "UP TO" in the advertising, they aren't guaranteeing that speed, they are saying that is the "max."

In the end it doesn't matter because AMD still has a product that matches Nvidias offerings, for quite a bit less money.
 
darkstarcow said:
What problems are they having with their cooler?



I will say it again since you seem to be having a hard time understanding what I said "If AMD had advertised their cards as having a base clock of 650Mhz, and boost clock of 850Mhz, would you buy them instead?" See what I did their? AMD isnt delivering anything less then Nvidia, their marketing approach is just retarded.

Your example is also completely irrelevant since AMD states "UP TO" in the advertising, they aren't guaranteeing that speed, they are saying that is the "max."

In the end it doesn't matter because AMD still has a product that matches Nvidias offerings, for quite a bit less money.
Click to expand...

It doesn't matter if AMD states "UP TO" in giant 3 inch high text on the box. I the consumer do not want "UP TO", I want to buy a minimum that meets or exceeds my expectation. That does not mean I want to buy "well, it might be running at 650 sometimes, maybe your card will run 750, but someone's card will run at 850!". That is precisely why I pointed out the problem with most DSL marketing. If I call up a company and order an "UP TO 8mbps" DSL line and only get 768k out of it, I'd be pissed. Now that is an obvious extreme difference, but even if I got 7mbps, I didn't want 7mbps, I wanted 8mbps.

You know damned well that if Intel(or AMD for that matter) said their CPU speed was "up to 3.5ghz" and someone never hit 3.5ghz, they'd be pissed.
 
Merc1138 said:
It doesn't matter if AMD states "UP TO" in giant 3 inch high text on the box. I the consumer do not want "UP TO", I want to buy a minimum that meets or exceeds my expectation. That does not mean I want to buy "well, it might be running at 650 sometimes, maybe your card will run 750, but someone's card will run at 850!". That is precisely why I pointed out the problem with most DSL marketing. If I call up a company and order an "UP TO 8mbps" DSL line and only get 768k out of it, I'd be pissed. Now that is an obvious extreme difference, but even if I got 7mbps, I didn't want 7mbps, I wanted 8mbps.

You know damned well that if Intel(or AMD for that matter) said their CPU speed was "up to 3.5ghz" and someone never hit 3.5ghz, they'd be pissed.
Click to expand...

Pissed at themselves for being a complete fucking moron I would hope.
 
You must log in or register to reply here.
Back
Top