ATI Radeon HD 4890 1 GB GDDR5 in Pretty Pixels

[DarkMatter]

You just proved our point there. The 4850/70 cards are known to idle high, but their chip design can handle sustained temperatures much higher than that - almost 30 degrees centigrade higher, to be precise, which is a large difference. I understand that might *seem* unsettling at first glance, but really it has no need to be.


That is a very good point, and naturally some people will be unlucky and get cards that are less able to tolerate the heat - but that's an unfortunate fact of life, and is the same for all areas of the consumer electronics industry. The RMA figures show that there's not an unreasonable number of problem cards.

My friends cards started artifacting at over 110ºC, a temperature that was reached in an eye blink.

The RMA figures show nothing as they do not represent a big enough base.

If the card tests as fine at their centre, then it might indicate a problem at your end causing artifacts - not necessarily one that's your fault, either, there are a lot of variables to deal with. It could be a problem peculiar to your own configuration. Temperature being high but still within specifications is not a good enough reason for an RMA, though - if you don't like the card idling high, then you can always change the cooler. And any skew to the RMA figures resulting from consumer reluctance will apply to the competition, too, so it still proves nothing about the scale of this problem.

Of course the "problem" is in our end. Namely ambient temperature higher than 15-20ºC, and the fact that we don't have the closed antiseptic test environments as they have. Nor we use testbeds, but cases that will make airflow worse no matter how good it is. It's our fault? Of course NOT. They have to make cards to work everywhere and from what I've seen the REFERENCE HD48xx cards can't be used everywhere without problems. That's all that I'm saying all the time. Non-reference cards have no problems, but why should average joe buy an aftermarket cooler if he bought a reference card?

Haha, thanks for the heads-up
I've said my piece now though, I think I can probably leave it at that.

Never!!

 

[Spunjji]

You're right, I couldn't resist replying just this once - but let's keep this short and not hijack the thread, eh?

Non-reference cards have no problems, but why should average joe buy an aftermarket cooler if he bought a reference card?

As I understand it, the 'average joe' isn't really the target market of the 4850 and above cards But you're right, nobody should *have* to change the cooling if they don't want to.

However, regarding your comment about lab testing, a well-designed case (nothing extravagant, just decent airflow design) will often keep a card cooler than it will be in free-air on a test bench.

It sounds like your card had a genuine problem if it was hitting 110 degrees, possibly shipping damage or a badly fitted cooler, but if they refused an RMA it sounds like you're the unlucky recipient of everyone's old friend - bad customer service. In which case I'm sorry to hear that, it sucks to get shafted that way.

Back closer to topic, though, I honestly don't think it's indicative of a larger problem. I can counter your friend with two of my own who have cards that haven't failed, it doesn't really *mean* anything though, which is why I didn't use it as the crux of my argument.

 

[EastCoasthandle]

What's all that crap? You can think whatever you want. Even that I'm making all this up. But what about TheMailMan, in this same thread?? He states that his card idles at 81ºC and he is making it up too? To support my point, that is, we are best friends after all... Eh MailMan??

On the other hand he is not having artifacts. Well temperature here might be a little bit higherand we do like a little bit higher temperatures than what seems to be common in the US and other countries. When I went there in vacations I almost got freezed when I entered any public building and I was told it was the same temperature that people uses at home. I got countless headaches in CAL, becaue of the temperature difference between outdoors and indoors.

Also not all chips can handle the same temperatures before artifacting, so even if only 10% of the cards overheat and only half of those artifact when overheating, that's still a lot of cards having issues. If it was a competition of course you would find more evidence of working cards with no problems at all, because more than 90% of the cards runs well, but there's still a lot of problematic cards noentheless. As I said my friends had artifacting and in order to solve the problem we had to put the fan @80-100% load, which was quite noisy. No need to go through RMA (RMA wouldn't help anyway), so the link about failure rates means nothing. We are reluctant to RMA a card because of temperatures and artifacting (I already did it once with no luck). Even if it is a design problem, they will very likely demostrate that it runs well, under their lab conditions, of course, and will return you the same card. Result for you: 20 euros less in delivery expenses. Wohoooo!!

With this post you contradicted everything you tried to claim about heating and noise
Edit:
Mailman posted his opinion(s) and left it at that. He did not try to convince others of his personal opinion by repeated replies to others who posted differently.

 

[DarkMatter]

You're right, I couldn't resist replying just this once - but let's keep this short and not hijack the thread, eh?


As I understand it, the 'average joe' isn't really the target market of the 4850 and above cards But you're right, nobody should *have* to change the cooling if they don't want to.

However, regarding your comment about lab testing, a well-designed case (nothing extravagant, just decent airflow design) will often keep a card cooler than it will be in free-air on a test bench.

It sounds like your card had a genuine problem if it was hitting 110 degrees, possibly shipping damage or a badly fitted cooler, but if they refused an RMA it sounds like you're the unlucky recipient of everyone's old friend - bad customer service. In which case I'm sorry to hear that, it sucks to get shafted that way.

Back closer to topic, though, I honestly don't think it's indicative of a larger problem. I can counter your friend with two of my own who have cards that haven't failed, it doesn't really *mean* anything though, which is why I didn't use it as the crux of my argument.

First of all, they are 2 cards and are not my cards, but of 2 different friends and were not the one I RMA'd. From what I've been seing in the last months 100-110 under load seems to be the norm for reference cards in this latitudes in hot days. The cases and the cabling are perfect. And so is the airflow, but the ambient... We won't change our ambient temperature just to make some cards work.

They use fans on testbeds.

And you would say that 150-200 euros/dollars is not what average joe gamer would spend? He can even spend $400. Average joe in the context we are speaking is that guy who buys a card for playing without wanting to deal with anything. In fact, he is pissed he has to install the drivers.

All in all I dont like where you are going in general. As in "good for most, bad for a few, who cares" attitude. That's a little bit Machiavellian. By making the cards run so hot they are just taking an unnecessary risk that will affect a lot of people. Does it matters if the group is statistically small, when the answer to the problem is so easy?? Aftermarket coolers and non-reference cards show that the cards can run cool, so is not just better to improve the reference coolers?? That was my original point about this anyway. Probably the new chips will run hotter, higher clocks, higher temepratures, is almost mathematical. Architecture improvements can help, but they will always find the same wall, which it is the original problem in RV770, the ratio between the power it consumes/dissipates and the area the chip occupies, which is the area through which the heat is going to be transferred. Unless they make a significant redesign, contrary to what rumors say, I don't see the temperature being any better, I see it higher if the same cooler is used.

 

[DarkMatter]

With this post you contradicted everything you tried to claim about heating and noise ;
Edit:
Mailman posted his opinion(s) and left it at that. He did not try to convince others of his personal opinion by repeated replies to someone post.

How I am contradicting anything there?? Heat is going to be higher the higher the ambient temperature is, but Nvidia cards and non-reference Ati ones run cool here. The ambient adds to the cards temperature, but it's only a problem if the card runs so close to it's maximum temperature threshold.

We are not going to spend a grand per month to power our AC just to fix the screw up of a company.

 

[EastCoasthandle]

How I am contradicting anything there?? Heat is going to be higher the higher the ambient temperature is, but Nvidia cards and non-reference Ati ones run cool here. The ambient adds to the cards temperature, but it's only a problem if the card runs so close to it's maximum temperature threshold.

We are not going to spend a grand per month to power our AC just to fix the screw up of a company.

I've already answered your repeated question back on page 2 or was it page 3 . But in all what you posted has without a doubt in my mind completely debunked your heat/noise issues. Thanks for all the dialogue.

 

[DarkMatter]

I've already answered your repeated question back on page 2 or was it page 3 . But in all what you posted is now without a doubt in my mind completely debunked. Thanks all the dialogue.

That makes absolutely no sense. Explain.

Maybe is this? When I say WE won't change our ambient, I don't mean me and friends, I mean the whole country. 44 million people. I'm talking about Spain, but I could probably speak for Portugal, Italy, Greece and Mexico, between others. Because they are probably in the same situation.

 

[Binge]

Naahhh. Just the model numbers got an overclock, LOL

That's totally incorrect. Why not take a look at the GTX285 FTW that starts at 710core/1620shader/1300memory. Nice joke but rebranding and overclocking is the name of the game with both companies.

 

[TheMailMan78]

What's all that crap? You can think whatever you want. Even that I'm making all this up. But what about TheMailMan, in this same thread?? He states that his card idles at 81ºC and he is making it up too? To support my point, that is, we are best friends after all... Eh MailMan??

Who are you?! Get out of my room!

 

[Spunjji]

All in all I dont like where you are going in general. As in "good for most, bad for a few, who cares" attitude. That's a little bit Machiavellian.

Close, but not quite. I'm trying to take a pragmatic approach. I don't think nobody should care about the problems of a minority of users, but at the same time, companies cannot design for every eventuality. I think the best approach you can take here is to let ATi know about your problem, and hopefully that way later products will take the problems on board and improve the design.

And as mentioned before, we don't know what differences have been made to the GPU / PCB / power delivery / cooler design that aren't obvious in the shots so far, so this model *may not* have the problems you anticipate. It still might, it's just not as clean-cut as you're implying.

EDIT:
I forgot to mention that I agree with this point:

Aftermarket coolers and non-reference cards show that the cards can run cool, so is not just better to improve the reference coolers??

Sadly the cheap nastiness of most stock coolers is a direct result of market demand, particularly OEMs that have slim margins. It sucks, and I hope it changes.

 

[Spunjji]

Who are you?! Get out of my room!

LOL

 

[TheMailMan78]

Close, but not quite. I'm trying to take a pragmatic approach. I don't think nobody should care about the problems of a minority of users, but at the same time, companies cannot design for every eventuality. I think the best approach you can take here is to let ATi know about your problem, and hopefully that way later products will take the problems on board and improve the design.

And as mentioned before, we don't know what differences have been made to the GPU / PCB / power delivery / cooler design that aren't obvious in the shots so far, so this model *may not* have the problems you anticipate. It still might, it's just not as clean-cut as you're implying.

Do you mean (gasp) not all PCs are the same?!?

 

[Spunjji]

Do you mean (gasp) not all PCs are the same?!?

Terrifying idea, isn't it! XD

 

[DarkMatter]

Close, but not quite. I'm trying to take a pragmatic approach. I don't think nobody should care about the problems of a minority of users, but at the same time, companies cannot design for every eventuality. I think the best approach you can take here is to let ATi know about your problem, and hopefully that way later products will take the problems on board and improve the design.

I don't think that the gamer community of several countries with a population 40 million each is a minority. Sure not everyone will have the problem, but IMO it's something that should e fixed. Ati never played so lose with the risk factor before, and there's no need to IMO. what would matter a little bit improved cooling? Adding $1 to the price of the reference board?

But I guess that in the end, it's just that we disagree on the fundamental basics regarding globality/locality (don't know the actual word). I mean you think that it's better to make the card the best posible for the mayority of people, price included even if that supposes a problem for a few, and I think that they should make it better for all the people, even if that supposes a slightly less interesting product in general, like costing $1 more for all the people. I simply don't believe in the inconscious sacrifice of a few for the good of most.

And as mentioned before, we don't know what differences have been made to the GPU / PCB / power delivery / cooler design that aren't obvious in the shots so far, so this model *may not* have the problems you anticipate. It still might, it's just not as clean-cut as you're implying.

Well I am always basing everything in the info that has been said . Like that it will be a tweaked RV770 and that it will use the same board, just with extra phases. That has been said at least. I mean, you are suggesting another RV670, but Ati never marketed RV670 as a tweaked R600, but as a redesigned chip. In Ati's words "some tweaks" to the core, looks like some very small changes, considering that RV670 was a redesign.

 

[Spunjji]

Adding $1 to the price of the reference board?...

...you think that it's better to make the card the best posible for the mayority of people, price included even if that supposes a problem for a few...

...I simply don't believe in the inconscious sacrifice of a few for the good of most...

...I mean, you are suggesting another RV670, but Ati never marketed RV670 as a tweaked R600, but as a redesigned chip. In Ati's words "some tweaks" to the core, looks like some very small changes, considering that RV670 was a redesign.

Right, let's clear this up.

Firstly, AMD/ATi are a company in serious financial trouble right now, so $1 a board is going to be a lot of money to them - they have to justify it all to the shareholders. Personally, I think that it sucks and that is NOT how a technology company should do things, but it's the way capitalist societies work so we have to live with it until we figure out something better.

Secondly, and most importantly, you've misunderstood where I stand here - I think the best graphics cards are naturally those which would never encounter these problems. I haven't bought a 4850 and never would, because I consider the power draw to be too high. I'm waiting for a 40nm refresh of it (which sadly looks like it will be a disappointment regarding power efficiency). A lot of people don't care though, and for the reasons mentioned in my first point, graphics companies are duty-bound to address that market first, because it's larger. So while, I *personally* believe that engineering concerns should transcend financial ones - in other words, this problem should never have occurred - that's an unrealistic expectation in today's economy. Make as big a noise as you can to the right people, and if we're lucky the problems will eventually be resolved.

And finally, I do know that this is no RV670, they don't seem to have made nearly enough changes and there's no die-shrink. I just hope that what they *have* done will be more than pump their chip full of volts, ATi can be much more subtle than that. If they have, it's the marketing people's fault, and sometimes we have to live with products made to their standards. It sucks, but there you go.

 

[DarkMatter]

Right, let's clear this up.

Firstly, AMD/ATi are a company in serious financial trouble right now, so $1 a board is going to be a lot of money to them - they have to justify it all to the shareholders. Personally, I think that it sucks and that is NOT how a technology company should do things, but it's the way capitalist societies work so we have to live with it until we figure out something better.

Secondly, and most importantly, you've misunderstood where I stand here - I think the best graphics cards are naturally those which would never encounter these problems. I haven't bought a 4850 and never would, because I consider the power draw to be too high. I'm waiting for a 40nm refresh of it (which sadly looks like it will be a disappointment regarding power efficiency). A lot of people don't care though, and for the reasons mentioned in my first point, graphics companies are duty-bound to address that market first, because it's larger. So while, I *personally* believe that engineering concerns should transcend financial ones - in other words, this problem should never have occurred - that's an unrealistic expectation in today's economy. Make as big a noise as you can to the right people, and if we're lucky the problems will eventually be resolved.

And finally, I do know that this is no RV670, they don't seem to have made nearly enough changes and there's no die-shrink. I just hope that what they *have* done will be more than pump their chip full of volts, ATi can be much more subtle than that. If they have, it's the marketing people's fault, and sometimes we have to live with products made to their standards. It sucks, but there you go.

We can agree to disagree. Just one thing, I was suggesting $1 more was passed to the customer, not paid by the company. In fact, even though I didn't stated so, I was thinking more about spending $1 more in the card and charging $2 more to the customer. Hell, I would happily pay $5 more for an error free (or safer) card. That's why I like Nvidia cards better lately, they are better built IMHO. Yes they may have a much higher failure rate, but the one that works, it works well everywhere, if you take one bad you can always change it. With HD48xx cards if it overheats in your region/house/case you are already screwed up, because it's a design flaw, RMAing won't work, you will get the same thing, but new.

 

[Spunjji]

We can agree to disagree. Just one thing, I was suggesting $1 more was passed to the customer, not paid by the company. In fact, even though I didn't stated so, I was thinking more about spending $1 more in the card and charging $2 more to the customer. Hell, I would happily pay $5 more for an error free (or safer) card. That's why I like Nvidia cards better lately, they are better built IMHO. Yes they may have a much higher failure rate, but the one that works, it works well everywhere, if you take one bad you can always change it. With HD48xx cards if it overheats in your region/house/case you are already screwed up, because it's a design flaw, RMAing won't work, you will get the same thing, but new.

That's fair enough. It still doesn't work for the OEM customers (numero uno to the graphics card company) as the dollars matter a lot to them too, but I fully understand your point. I think we can leave it at that now, right?

 

[Haytch]

Renaming, re-anything always generates sales from the less aware. Perfect strategy, something not left only for the green camp.

Im a strong believer that many many shit products should exist, for the many many shit people that live in this world.

If by chance this card was ideal for an enduser, then so be it, but if someone ended up with this card via assumption, then, HAHA, gawd im cold today!

 

[Spunjji]

For anyone still having their head inserted firmly into their rectum over this card because it's a 'rename' that will have terrible thermal characteristics:

http://www.xbitlabs.com/articles/video/display/radeon-hd4890_6.html

 

[DarkMatter]

For anyone still having their head inserted firmly into their rectum over this card because it's a 'rename' that will have terrible thermal characteristics:

http://www.xbitlabs.com/articles/video/display/radeon-hd4890_6.html

I changed my mind when I learned that the card would have a different transistor count and specieally a greater die area. If any has readed my comments about RV770 in the past, I always said RV770 was hot because Ati had put all the units too close to each other in order to save space and make it cheap. It backfired and they have fixed it. But RV790 hardly is a tweaked or reworked RV770 IMO, nor it's based on it, except for the fact that it shares the number of untis. It's a completely different animal and all my previous comments were based on the reports that this was going to be a tweaked RV770. Reports made even by AMD reps BTW. Well it's not, end of story.

EDIT: Also the cooler is not the same. It looks the same, but according to one friend I have at Barcelona (400 km away from me) the 4890 weights a little bit more (suggests a little bit thicker block) and the heatpipe/fins layout is a bit different from what he could remember. The fan is a little bit higher rated too (and thus a tiny bit noisier at same %, comparing two saphires). I can't find it out myself as none of my close living friends will buy that card (if they buy any the GX275 is better and cheaper here so far), but I'm confident of his findings.