[Kitguru] Nvidia continue to come under fire for poor GTX590 design

[blastingcap]

Don't know why people feel the need to thread crap, it's a legitimate thread and it has a good amount of information for everyone.

I think some people think it's old news and others, well, I suspect that many pro-green folks were in denial around the time of bumpgate, too (which cost NVIDIA hundreds of millions or even more than a billion dollars, IIRC).

That said, who the heck buys these cards anyway, when SLI GTX 570s cost so much less? And dual GPUs kind of suck anyway, due to software problems, microstutter, far higher heat/noise/power than the marginal graphics power they add, etc.

I can sort-of understand the people who bought GTX 480s when they first came out, as it was the single-GPU king even if it was grossly overpriced, but this card is a failure for the reasons above.

Bring on single-GPU 28nm. Nothing 40nm excites me. *prays that NVIDIA enables Surround on a single GPU for its 28nm lineup*

 

[Cookie Monster]

Those 4 capacitors is most likely to be used for filtering purposes since the type of VRM used for a video card I suppose is a multiphase synchronous buck converters, same as motherboards.

The capacitors are there so that the output voltage to the GPU has minimal ripple. Not sure what the 8 capacitors at the back is for but its also there to smooth out the DC output.

Anywho, Ive built DC to DC converters before and yes they always blow up (mostly the MOSFET or whatever transistor is used is the one thats often the one to bite the dust) when going over spec e.g raising the input voltage outside of spec or putting a load outside of spec. However most DC to DC converters tend to have some sort of voltage/current or even a power limiter incase they go over the spec.

These limiters often use some sort of micro controllers, and Im guessing that with regards to the GTX590, nVIDIA's drivers sets the values/limits via software for the micro controller. So instead of having such power limiters programmed outside of the driver influence, its incorporated within the driver so that these "limits" set by the limiter can be controlled easily via software without having to re-program those micro controllers individually.

From what I can see, the GTX590 has no issues at stock (remains to be seen) and thats where it counts the most. This product is not a failure because it's set out to do it what it was built to do without failing. Within that list of what it was suppose to do, overclocking and overvolting is not part of that list.

In engineering there are always trade offs. nVIDIA chose to build this card to be physically smaller, quiet card than its competition while providing decent performance (ref - xbitlabs). Where as the HD6990 was aimed at pure performance. Because of the physical limitations, and from what I think is a budget constraint since two low voltage GF110 chips are bound to be more expensive, corners were indeed cut and all engineering projects are like this.

People need to understand that overvolting and overclocking is never something that's guaranteed hence YMMV. Some video cards (referring to stock cards from either IHVs and not the AIB partners) are over engineered in some respect but this isn't so that it allows overclocking headroom which is often unintentional. That's rarely the case. Arguments like "its an enthusiast product so it must overclock(old trend) and overvolt(newer trend)" doesn't make much sense unless the IHV or the AIB partners clearly advertise that it does.
In this case, Asus is clearly wrong for falsely advertising and hence their move to clamp the voltages via new BIOs.

So unless a legitimate hardware site or the IHV/AIBs state that there are issues, clearly explaining that this card is unfit to run at stock settings, the product itself is not an engineering failure but rather the execution of launching this card was a failure in the sense that the CDs that come with the package caused some cards to blow up when driven to the extreme.

Its quite entertaining in the fact that users nowadays expect overclocking and overvolting a given in video cards and if the card somehow dies, its the IHVs fault for having "weak" VRMs and etc. Compare this to , users about 6 years ago, most thought gaining about 25~50MHz from stock cards was like winning the lotto.

 

[Aristotelian]

Yes an OCd 6990 "can run at 580 SLI speeds".

And how does an overclocked 6990 compare to overclocked 580s? Are we really going to have to go down this road again, of debates regarding 'apples to apples' comparisons? Don't overclock your left hand and claim that it's comparable to your stock right hand when your stock right hand overclocks, too. I'd be happy to see some overclocked 580 SLI benchmarks versus an overclocked 6990 to see support for this claim that 6990 runs at 580 SLI speeds.

 

[Silverforce11]

I think some people think it's old news and others, well, I suspect that many pro-green folks were in denial around the time of bumpgate, too (which cost NVIDIA hundreds of millions or even more than a billion dollars, IIRC).

That said, who the heck buys these cards anyway, when SLI GTX 570s cost so much less? And dual GPUs kind of suck anyway, due to software problems, microstutter, far higher heat/noise/power than the marginal graphics power they add, etc.

I can sort-of understand the people who bought GTX 480s when they first came out, as it was the single-GPU king even if it was grossly overpriced, but this card is a failure for the reasons above.

Bring on single-GPU 28nm. Nothing 40nm excites me. *prays that NVIDIA enables Surround on a single GPU for its 28nm lineup*

Look at the power use for CF/SLI setups, then look at the dual GPU setup that offer equivalent performance. It's always in the favor of the dual GPU card. Also, perf/watt of single or cf/sli setups are a lot better, add a CF/SLI and get on average 80% increase in performance for ~100% increase in power. It's not inefficient. Compare that to a dual-GPU which are made to be more efficient in perf/watt, it's an even better deal.

There's a disadvantage to be had with CF/SLI of mid-range cards to exceed top end single GPUs, if CF/SLI is broken, your game will run like its on a mid-range card. But these dual GPU using top end cores do not have that problem.

Whenever 28nm GPUs come, i will grab two of the top end GPUs and CF/SLI them, without hesitation. The improvements in compatibility and performance improvement has been amazing.

 

[Silverforce11]

And how does an overclocked 6990 compare to overclocked 580s? Are we really going to have to go down this road again, of debates regarding 'apples to apples' comparisons? Don't overclock your left hand and claim that it's comparable to your stock right hand when your stock right hand overclocks, too. I'd be happy to see some overclocked 580 SLI benchmarks versus an overclocked 6990 to see support for this claim that 6990 runs at 580 SLI speeds.

Not a fair comparison, price delta is huge.

My statement which some ppl have expanded on is that the current 590 is a poor enthusiast product because it has terrible OC headroom (with recent drivers), especially for users who like to go with water cooling and up the vcore to get a big OC.

Look at the 6990. It can certainly flex and exceed 6970 CF.

Which is why i've said, whenever the non-reference model ship with better components, the 590 will truly shine as it will easily flex to 580s speed or beyond. AIBs can charge a big premium just due to that.

 

[Arzachel]

Its quite entertaining in the fact that users nowadays expect overclocking and overvolting a given in video cards and if the card somehow dies, its the IHVs fault for having "weak" VRMs and etc. Compare this to , users about 6 years ago, most thought gaining about 25~50MHz from stock cards was like winning the lotto.

Its not that people expect a huge overclocking headroom, they expect that the safety features will kick in and prevent hardware damage if they run it too far out of spec. Refusing to boot is infinetly better than burning up.

 

[badb0y]

Its not that people expect a huge overclocking headroom, they expect that the safety features will kick in and prevent hardware damage if they run it too far out of spec. Refusing to boot is infinetly better than burning up.

Not only that but the direct competition is able to OC pretty well so it looks poor when compared, especially when the price is so close.

 

[wahdangun]

Those 4 capacitors is most likely to be used for filtering purposes since the type of VRM used for a video card I suppose is a multiphase synchronous buck converters, same as motherboards.

The capacitors are there so that the output voltage to the GPU has minimal ripple. Not sure what the 8 capacitors at the back is for but its also there to smooth out the DC output.

Anywho, Ive built DC to DC converters before and yes they always blow up (mostly the MOSFET or whatever transistor is used is the one thats often the one to bite the dust) when going over spec e.g raising the input voltage outside of spec or putting a load outside of spec. However most DC to DC converters tend to have some sort of voltage/current or even a power limiter incase they go over the spec.

These limiters often use some sort of micro controllers, and Im guessing that with regards to the GTX590, nVIDIA's drivers sets the values/limits via software for the micro controller. So instead of having such power limiters programmed outside of the driver influence, its incorporated within the driver so that these "limits" set by the limiter can be controlled easily via software without having to re-program those micro controllers individually.

From what I can see, the GTX590 has no issues at stock (remains to be seen) and thats where it counts the most. This product is not a failure because it's set out to do it what it was built to do without failing. Within that list of what it was suppose to do, overclocking and overvolting is not part of that list.

In engineering there are always trade offs. nVIDIA chose to build this card to be physically smaller, quiet card than its competition while providing decent performance (ref - xbitlabs). Where as the HD6990 was aimed at pure performance. Because of the physical limitations, and from what I think is a budget constraint since two low voltage GF110 chips are bound to be more expensive, corners were indeed cut and all engineering projects are like this.

People need to understand that overvolting and overclocking is never something that's guaranteed hence YMMV. Some video cards (referring to stock cards from either IHVs and not the AIB partners) are over engineered in some respect but this isn't so that it allows overclocking headroom which is often unintentional. That's rarely the case. Arguments like "its an enthusiast product so it must overclock(old trend) and overvolt(newer trend)" doesn't make much sense unless the IHV or the AIB partners clearly advertise that it does.
In this case, Asus is clearly wrong for falsely advertising and hence their move to clamp the voltages via new BIOs.

So unless a legitimate hardware site or the IHV/AIBs state that there are issues, clearly explaining that this card is unfit to run at stock settings, the product itself is not an engineering failure but rather the execution of launching this card was a failure in the sense that the CDs that come with the package caused some cards to blow up when driven to the extreme.

Its quite entertaining in the fact that users nowadays expect overclocking and overvolting a given in video cards and if the card somehow dies, its the IHVs fault for having "weak" VRMs and etc. Compare this to , users about 6 years ago, most thought gaining about 25~50MHz from stock cards was like winning the lotto.

but the fact is almost every AIB since 8800GT always advertise overclocking on this market segment like BFG(now defunct), EVGA, XFX, asus etc. and its the most catastrophic launch in the graphic card world since FX 5800 and HD 2900XT because so many card dyeing in the hand of reviewer so it maybe shocked some user.

 

[SirPauly]

I think some people think it's old news and others, well, I suspect that many pro-green folks were in denial around the time of bumpgate, too (which cost NVIDIA hundreds of millions or even more than a billion dollars, IIRC).

That said, who the heck buys these cards anyway, when SLI GTX 570s cost so much less? And dual GPUs kind of suck anyway, due to software problems, microstutter, far higher heat/noise/power than the marginal graphics power they add, etc.

I can sort-of understand the people who bought GTX 480s when they first came out, as it was the single-GPU king even if it was grossly overpriced, but this card is a failure for the reasons above.

Bring on single-GPU 28nm. Nothing 40nm excites me. *prays that NVIDIA enables Surround on a single GPU for its 28nm lineup*

I don't think it's old news and worthy of discussion but how many different threads does one need? How about a sticky with just one thread discussing this?

 

[bryanW1995]

Yuk Yuk Yuk :biggrin:

I guess people still see the 590 failure as important news. Although its the same ole vrm bashing.Would rather hear bashing based on its price to performance ratio. And I guess for not being able to take the performance crown from AMD. But even then the 590 bashing is getting old. I'm sure nvidia will fix the design issues and/or help their partners to fix it.

Don't hold your breath. EVGA already has the water block edition for $879 on their website, imho that's the best fix but the premium is just too high. Nvidia has already moved on to other things, like breathing down TSMC's neck and stabbing their intel voodoo dolls.

Just FYI folks and a reference to Kitguru as of late. Not exactly exuding the best track record over the last week or so when it comes to reporting. So grain of salt.
http://forums.anandtech.com/showthread.php?t=2154162&highlight=kitguru

I'm not pleased with their reporting, but didn't they get it at least somewhat right? Predicting a new bios is usually a safe bet, but with several new ones on the horizon they weren't completely out of the ballpark. They just pointed the finger at NV instead of the vendor partners.

 

[bryanW1995]

Ouch, you could toast a marshmallow on that thing.

I wonder if cards like this will eventually start coming with OEM water coolers similar to the corsair products.

You mean like this?

 

[OILFIELDTRASH]

Look at the power use for CF/SLI setups, then look at the dual GPU setup that offer equivalent performance. It's always in the favor of the dual GPU card. Also, perf/watt of single or cf/sli setups are a lot better, add a CF/SLI and get on average 80% increase in performance for ~100% increase in power. It's not inefficient. Compare that to a dual-GPU which are made to be more efficient in perf/watt, it's an even better deal.

There's a disadvantage to be had with CF/SLI of mid-range cards to exceed top end single GPUs, if CF/SLI is broken, your game will run like its on a mid-range card. But these dual GPU using top end cores do not have that problem.

Whenever 28nm GPUs come, i will grab two of the top end GPUs and CF/SLI them, without hesitation. The improvements in compatibility and performance improvement has been amazing.

So you're going to buy two of the highest end 28nm gpu's at launch? Looks like you're rocking a last gen middle tier gpu and a two generation old CPU. So now you're gonna spend 3 times what your computer is worth on next gen cards? :scratches head:

 

[Chiropteran]

Just FYI folks and a reference to Kitguru as of late. Not exactly exuding the best track record over the last week or so when it comes to reporting. So grain of salt.
http://forums.anandtech.com/showthread.php?t=2154162&highlight=kitguru

I don't get it. Asus has a new bios for the GTX590. The GTX590 is an nvidia card. Therefore, there is a new nvidia bios for the GTX590. Does nvidia ever offer new BIOS directly from their website? I thought it has always been standard practice for the board manufacturers to offer the BIOS updates.

I mean, I get the complaint- "oh no, kitguru said nvidia was doing it but it was really an nvidia card manufacturer, so they lied!" But ultimately what does that change? The original BIOS clearly had a problem, and a new one was made to fix or improve it. Whether it was actually offered from nvidia or Asus changes nothing.

 

[notty22]

Look at the power use for CF/SLI setups, then look at the dual GPU setup that offer equivalent performance. It's always in the favor of the dual GPU card. Also, perf/watt of single or cf/sli setups are a lot better, add a CF/SLI and get on average 80% increase in performance for ~100% increase in power. It's not inefficient. Compare that to a dual-GPU which are made to be more efficient in perf/watt, it's an even better deal.

There's a disadvantage to be had with CF/SLI of mid-range cards to exceed top end single GPUs, if CF/SLI is broken, your game will run like its on a mid-range card. But these dual GPU using top end cores do not have that problem.

Whenever 28nm GPUs come, i will grab two of the top end GPUs and CF/SLI them, without hesitation. The improvements in compatibility and performance improvement has been amazing.

That is not true. They rely on sli and crossfire to function.
In rare situations they react differently.
One uses a on board bridge chip* to split communication to the gpu's VS the chipset. But disable crossfire or sli , in driver, the dual gpu's don't work. No profile the game does not use dual gpu's correctly.
There are people with gtx 295's that can't play NFS shift 2, right now because there is no SLI profile in Nvidia's drivers, thats the newest example I can give you.

*which is why the m/b need not support sli/crossfire

 

[bryanW1995]

Those 4 capacitors is most likely to be used for filtering purposes since the type of VRM used for a video card I suppose is a multiphase synchronous buck converters, same as motherboards.

The capacitors are there so that the output voltage to the GPU has minimal ripple. Not sure what the 8 capacitors at the back is for but its also there to smooth out the DC output.

Anywho, Ive built DC to DC converters before and yes they always blow up (mostly the MOSFET or whatever transistor is used is the one thats often the one to bite the dust) when going over spec e.g raising the input voltage outside of spec or putting a load outside of spec. However most DC to DC converters tend to have some sort of voltage/current or even a power limiter incase they go over the spec.

These limiters often use some sort of micro controllers, and Im guessing that with regards to the GTX590, nVIDIA's drivers sets the values/limits via software for the micro controller. So instead of having such power limiters programmed outside of the driver influence, its incorporated within the driver so that these "limits" set by the limiter can be controlled easily via software without having to re-program those micro controllers individually.

From what I can see, the GTX590 has no issues at stock (remains to be seen) and thats where it counts the most. This product is not a failure because it's set out to do it what it was built to do without failing. Within that list of what it was suppose to do, overclocking and overvolting is not part of that list.

In engineering there are always trade offs. nVIDIA chose to build this card to be physically smaller, quiet card than its competition while providing decent performance (ref - xbitlabs). Where as the HD6990 was aimed at pure performance. Because of the physical limitations, and from what I think is a budget constraint since two low voltage GF110 chips are bound to be more expensive, corners were indeed cut and all engineering projects are like this.

People need to understand that overvolting and overclocking is never something that's guaranteed hence YMMV. Some video cards (referring to stock cards from either IHVs and not the AIB partners) are over engineered in some respect but this isn't so that it allows overclocking headroom which is often unintentional. That's rarely the case. Arguments like "its an enthusiast product so it must overclock(old trend) and overvolt(newer trend)" doesn't make much sense unless the IHV or the AIB partners clearly advertise that it does.
In this case, Asus is clearly wrong for falsely advertising and hence their move to clamp the voltages via new BIOs.

So unless a legitimate hardware site or the IHV/AIBs state that there are issues, clearly explaining that this card is unfit to run at stock settings, the product itself is not an engineering failure but rather the execution of launching this card was a failure in the sense that the CDs that come with the package caused some cards to blow up when driven to the extreme.

Its quite entertaining in the fact that users nowadays expect overclocking and overvolting a given in video cards and if the card somehow dies, its the IHVs fault for having "weak" VRMs and etc. Compare this to , users about 6 years ago, most thought gaining about 25~50MHz from stock cards was like winning the lotto.

Nvidia marketed themselves as the overclocking elite when they convinced reviewers to compare a gtx 460 ftw against a 6870. Heck, even before that, for years and years we have consistently seen overclocking as a very strong nvidia advantage. In a typical price bracket, how common is it for AMD to overclock more as a % than nvidia? I'd hazard to guess that over the past years there have only been a handful of AMD cards that had a clear overclocking advantage at a specific price point. How many overclock editions does nvidia have on newegg and/or other well known retailers? How much overclock do those cards have? Compare that to AMD, which only rarely has vendor partners put out a card with a really strong OC. And it has been like this for years. So I think that it is entirely relevant that Nvidia's "bestest card evar" can't overclock worth a shit without turning into a fire hazard. If nothing else this shows how nvidia has learned from AMD (noise/power/heat are important) and AMD has learned from nvidia (highest performance no matter what!!!). Definitely interesting how the wheels turn.

 

[SirPauly]

Imho,

And I agree with that -- it's a limitation and con of the GTX 590. One may of expected more OC headroom and some flexibility with over-volting but simply can't. It's a decision a potential buyer has to decide for themselves.

Considering, AIB's and IHV's are trying to offer flexibility for over-clockers, it's fair game to criticize the irony of this and probably was a tough decision and maybe an over-reaction on nVidia's part to eliminate all over-volting. Sadly, that choice was made.

 

[Aristotelian]

Not a fair comparison, price delta is huge.

So try 3x 580s and 2x 6990s and see what results you get. The prices should be roughly the same.

 

[Tempered81]

It has five per GPU.

IMC is part of the gpu

 

[Madcatatlas]

And how does an overclocked 6990 compare to overclocked 580s? Are we really going to have to go down this road again, of debates regarding 'apples to apples' comparisons? Don't overclock your left hand and claim that it's comparable to your stock right hand when your stock right hand overclocks, too. I'd be happy to see some overclocked 580 SLI benchmarks versus an overclocked 6990 to see support for this claim that 6990 runs at 580 SLI speeds.

You may claim you are only replying to someone else making claims, but why do you insist on derailing this topic?

So try 3x 580s and 2x 6990s and see what results you get. The prices should be roughly the same.

really, take that in another topic, make one if you so please.

The issue is 590s having trouble and that trouble being production errors, design errors which should be made known for potential buyers.

Already we are seeing Asus modified bioses to keep the cards in check. Nvidia is likely to have fixed the issue for the second "batch", but theres still a problem with the first "batch".

Also take notice of the various claims made by Asus, MSI etc etc about increasing voltage and clocks for this and other products. For this product, that advertisement is in many cases misleading. Especially if the card throttles at higher voltages anyway and therefore gives you the same performance as stock values.

All in all, not a very good product, a good one yes, but not a very good one. In comparison, the 6990 is a very good product with 1 flaw, that flaw being the noise, not that ive tested it myself.

 

[96Firebird]

Nvidia marketed themselves as the overclocking elite when they convinced reviewers to compare a gtx 460 ftw against a 6870.

That was a retail card that could be bought, and supported, by EVGA. I agree though, Nvidia cards have been known to have a little more overclocking headroom. But that doesn't mean because the GTX 590 can't be overvolted it is a failure of a card, like some are claiming.

 

[Madcatatlas]

It cant be overvolted, or shouldnt be overvolted. Why risk the headache of burning a 700+$ product and the following time spent waiting for a replacement product.

It cant be OCed much either as far as i can see from reviews around the web. This is just a natural result of the card not being able to take higher voltage.

 

[SlowSpyder]

And how does an overclocked 6990 compare to overclocked 580s? Are we really going to have to go down this road again, of debates regarding 'apples to apples' comparisons? Don't overclock your left hand and claim that it's comparable to your stock right hand when your stock right hand overclocks, too. I'd be happy to see some overclocked 580 SLI benchmarks versus an overclocked 6990 to see support for this claim that 6990 runs at 580 SLI speeds.

But I think the point was to show that $700 can get you the performance that only $1000 could get you not too long ago. No one is claiming that the GTX580 SLI setup wouldn't have potential to be faster. But if your goal was stock clock GTX580 SLI performance, you don't have to spend GTX580 SLI prices if you are willing to overclock the lower priced part.

And in regards to this thread, I would certainly only attempt to get one of the available current gent dual cards up to that performance level...

 

[Grooveriding]

Nvidia is likely to have fixed the issue for the second "batch", but theres still a problem with the first "batch".

I am looking forward to this as well.

The 590 has been out of stock everywhere since a few days after release, none on newegg, none at evga, none anywhere that I can see. There are a few in stock at some local computer stores here in Toronto, but none online at the major e-tailors.

I am certain it will not be missed by some review sites when GTX 590s become available for purchase again if there is a design modification to the PCB to correct it's current burn-up issues.

 

[formulav8]

I am looking forward to this as well.

The 590 has been out of stock everywhere since a few days after release, none on newegg, none at evga, none anywhere that I can see.

Maybe nvidia did a secret recall? Will probably never know.

For the most part, if their partners like Asus are releasing bios changes for the 590, its basically the same as nvidia doing it themselves. You can be sure nvidia was directly consulted/involved/questioned/blah/blah for the bios fix for the 590. So a brand like Asus releasing a new bios for the 590 has nvidia's fingerprints on it as well. Just my opinion though...

 

[Aristotelian]

You may claim you are only replying to someone else making claims, but why do you insist on derailing this topic?
really, take that in another topic, make one if you so please.

What do you mean by 'claim'? Here, have a look:

Because it costs the same as a competing product that can run at 580 SLI speeds?

(you didn't quote that), or

You think the 6990 competes with 580 Sli?

That, or

Yes an OCd 6990 "can run at 580 SLI speeds".

That. Why not?

Threads tend to cover multiple topics, over time, and not just what is explicitly mentioned in the OP. If I see claims such as the ones I quoted, I believe I'm free to engage them without being told by you (who somehow missed the instigators of this topic) to go start a different thread. I'm well aware of what the issue is in this thread, but part of the problem on these forums is misinformation. I'm tired of seeing people claim "overclock X, it's as good as Y" when it's an utterly unfair conclusion to draw.

Regarding this topic, it has been mentioned repeatedly that there have been problems (user induced, or not) with the 590. It's good that these issues are being brought to light, but the thread shouldn't masquerade as an opportunity to spread misinformation about other cards (i.e., to glorify the 6990 as some ultra card that is equal to 580s in SLI, which it isn't - and I wasn't the one who brought this claim into the thread).