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

[Grooveriding]

Some more comments on kitguru about the GTX 590 VRMs and related issues are up.

Their earlier posting related to the 590s receiving updated BIOS', http://www.kitguru.net/components/g...-gtx590-bios-rushed-out-to-repair-reputation/, which some felt was false, while not 100% dead on, did turn out to have the right of the situation with an updated BIOS being releaed shortly after:
http://support.asus.com/Download.aspx?SLanguage=en&m=ENGTX590+Series&p=9&s=2
http://techreport.com/discussions.x/20711

Here is Kitguru's latest blurb on GTX 590 VRMs and their characteristics:

http://www.kitguru.net/components/graphic-cards/carl/nvidia-come-under-fire-for-poor-gtx590-design/


'...KitGuru reported before that respected website Hardware.fr analysed the GTX590 with a thermograph.

Their GTX590 testing showed a reading of around 112c under load. Since we reported on this, our readers have been filling our inbox with comments and views on the matter. Some partners have been updating card bioses to supposedly help with power delivery however the main issue is apparently down to a poor VRM design. The very high temperatures are due to the VRM’s being underspecified. Each of the GPU/IMC phases has 35A each, and only four phases per GPU, with one phase for the IMC.

With the DrMOS chips being operated at close to peak forward current rating with a high duty cycle, this is part of the reason of the stress related failures. At very high load levels the DrMOS chips can become very inefficient as they are optimised for efficiency at low loads. Some claim they shouldn’t be used for VRMs for high power electronics in the first place. This inefficiency is adding more wasted heat into the scenario. Nvidia have placed them right under the fan’s dead spot too which doesn’t help with cooling.

Why wasn’t this a problem with the GTX580? Well this card was complimented with 6+2 power phases and the design is dramatically different. It looks as if the bios changes might not be the cure, and many people will be best waiting on third party custom designs from companies such as ASUS which will rectify the poor design aspects of the latest Nvidia flagship card....'


In the lab501.ro review where they had a 590 fail, they also comment on the quality of the 590's VRMs here:

http://translate.google.ca/translat...dia-geforce-gtx-590-studiu-de-overclocking/12


While I cannot find a detailed layout of the 590 PCB, are these below examples of 590s that have failed, all with the same component burning up; giving an example of VRM failure ? The component that has caused the 590 to fail in these three separate examples is not a VRM. The 590s VRMs are in this first picture, they are the two strips of five larger cube type components and below that are three examples of GTX 590s that failed with different components on the PCB burning up.

Could the failure of VRMS cause these other smaller components to exhibit failure and ignite, or are these smaller components something independent to the VRMs. And can anyone identify what these smaller components that burned are ? Resistors ?


The 590 VRMs, the large cubelike components in a series of five on the top and bottom

Three examples of the same components failing on three different gtx 590s, note these are not the VRMs:

http://translate.google.com/transla...ulletin/showpost.php?p=3960849&postcount=352'

http://forums.overclockers.co.uk/showpost.php?p=18781867&postcount=140

http://img231.imageshack.us/img231/2055/sdsdu.jpg

Oddly with the gtx 590 that failed at Techpowerup http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_590/26.html he identifies a resistor

and then a MOSFET that burned out

These are neither the VRMS or the smaller components that burned out in these other photos.

Unfortunately the other review sites that had 590s burn out did not post pictures of where the failure occured on the PCB, ie: the sweclockers youtube video, you just see the spark and then the system crash: http://www.youtube.com/watch?v=sRo-1VFMcbc

For those that understand this better, is it possible for VRMs pushed beyond spec due to overclocking to cause these separate components to have failed or would it be isolated to the VRMs ?

Per this hardocp review http://hardocp.com/article/2011/04/03/asus_geforce_gtx_590_overclocking_followup power management functions are now effective on the 590 via drivers, 267.71 and this prevents the 590 from burning out and shows examples of the card throttling when voltage is adjusted.

 

[Genx87]

ok

 

[SirPauly]

Thanks for sharing!

 

[OILFIELDTRASH]

Cliffs: nvidia creates the most poorly designed video card ever made. Amd fanboys everywhere rejoice.

 

[MangoX]

Cliffs: nvidia creates the most poorly designed video card ever made. Amd fanboys everywhere rejoice.

This :biggrin:

 

[bryanW1995]

This was sort of big news last week, but I think that by now anybody who can read is avoiding gtx 590, anyway. Let the vendetta go already. I'm going to make a bold prediction here: if you keep beating this horse you're going to get baited by a green-teamer into another, potentially much longer, vacation. Let it go, maybe go talk about how 6790 smokes gtx 550 by 6% or something.

 

[Skurge]

Thanks for sharing!

At least his post a bit more level headed and inpartial. Although it could have been posted in the many other 590 threads.

Still, he raises some good questions. I'll like to know the answers as well.

 

[SirPauly]

You're right, it didn't include turd this time!

 

[formulav8]

Cliffs: nvidia creates the most poorly designed video card ever made. Amd fanboys everywhere rejoice.

Yeah I don't know why people hold on to corps who could care less about you. I like hardware and stuff but I know nvidia (or whoever) could care less about me.

I'm not sure why people would rejoice over a bad card design. People paid alot of money for this thing. Only to find out that its not up to the same standards as nvidia's other cards. Its not to funny when you don't feel like you got your moneys worth.

 

[Idontcare]

ok

Thanks for sharing!

Please keep it respectful and productive.

If you have nothing productive to contribute to the discussion then you are expected to simply refrain from posting.

Idontcare
Super Mod

 

[formulav8]

Let it go, maybe go talk about how 6790 smokes gtx 550 by 6% or something.

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.

 

[bryanW1995]

At least his post a bit more level headed and inpartial. Although it could have been posted in the many other 590 threads.

Still, he raises some good questions. I'll like to know the answers as well.

What answers can we get out of this? The only answer I get is that a smaller, low-power gpu makes a better sandwich card than a much larger, high-power gpu. Oh, and AMD pulls the gloves off to compete in the dual gpu-on-a-stick category, while nvidia instead continues to design their cards to fit consumer and pro graphics equally well. And AMD outmaneuvered nvidia this time.

 

[Idontcare]

This was sort of big news last week, but I think that by now anybody who can read is avoiding gtx 590, anyway. Let the vendetta go already. I'm going to make a bold prediction here: if you keep beating this horse you're going to get baited by a green-teamer into another, potentially much longer, vacation. Let it go, maybe go talk about how 6790 smokes gtx 550 by 6% or something.

The topic of this thread is technical and relevant to the sub-forum.

The OP should not be expected to lower their expectations regarding the possibility of engaging in civil discourse of a technical nature relating to the thread's topic.

I would hope self-censoring like this is not the message learned from the moderator events of the past two weeks.

People do need to stop baiting, but people also need to stop taking the bait. It takes two to crap up a forum.

Resigning oneself to simply giving up, to stop trying to have technical and civil discussions altogether, is not the answer.

Idontcare
Super Mod

 

[SirPauly]

Each of the GPU/IMC phases has 35A each, and only four phases per GPU

It has five per GPU.

 

[Keysplayr]

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

 

[Madcatatlas]

For buyers, this is crucial information and reason for concern. I think whats needed is answers and not people waving it off/shrugging it away. And thats seems to be the OPs intent.
Though honestly in this regard i doubt Anandtech forums are the right place, atleast not the video section.

In addition,
As much as I hate thinking about it, i doubt this Fermi card i have will last more than two years. Ill probably have to bake it in a year as a result of the core issues with the architecture, just like the g80.

At this time, im hoping pc grahics/games get revolutionized and that the 28nm cards can provide a good enough oomph and quality to justify upgrading or even continuing gaming on a pc.

 

[SirPauly]

I don't disagree but how many threads can one have discussing the same thing?

What bothers me is Kitguru's info has mistakes. His main point about is the underlying phases and the most important aspect is wrong and then passed on by the OP and in bolds.

 

[Grooveriding]

His main point about is the underlying phases and the most important aspect is wrong and then passed on by the OP and in bolds.

It's not wrong. There are 5 phases, 4 for the GPU and one for the memory.

So it's actually correctly passed on in bolds.

Cheers

Also their previous article per the updated BIOS turned out correct in the fashion that ASUS released an updated BIOS to address bios management for the 590.

The other threads have degenerated into discussions of fans or accusatory flim flam back and forth. I want to hear about the VRMS and the component failure seen, not argue about unrelated issues.

In 590 threads on other forums such as xtremesystems http://www.xtremesystems.org/forums/showthread.php?t=268500&page=16 they are discussing it without nit-picking. Most of the opinions there reflect some of the OP of this thread in as much as the VRM/components/PCB is lacking on the 590.

 

[Jionix]

This is still important news.

Since there are NOT any alternate designs out there yet available, anybody buying one will have a card that is apparently brittle in design when you take into account it's performance and power draw...

I would hope that the non-existent supply in the USA is indicative of Nvidia having vendors pull back on their shipments to protect users, but that would be out of character for Nvidia.

 

[GaiaHunter]

I don't disagree but how many threads can one have discussing the same thing?

What bothers me is Kitguru's info has mistakes. His main point about is the underlying phases and the most important aspect is wrong and then passed on by the OP and in bolds.

And these threads are good ways to point the mistakes and generate good information.

 

[SirPauly]

It's not wrong. There are 5 phases, 4 for the GPU and one for the memory.

So it's actually correctly passed on in bolds.

Cheers

Also their previous article per the updated BIOS turned out correct in the fashion that ASUS released an updated BIOS to address bios management for the 590.

The other threads have degenerated into discussions of fans or accusatory flim flam back and forth. I want to hear about the VRMS and the component failure seen, not argue about unrelated issues.

In 590 threads on other forums such as xtremesystems http://www.xtremesystems.org/forums/showthread.php?t=268500&page=16 they are discussing it without nit-picking. Most of the opinions there reflect some of the OP of this thread in as much as the VRM/components/PCB is lacking on the 590.

GeForce GTX 590 reference specifications

GPU: 2 x GF110 @ 607MHz, 40nm
Shader processors: 1024 (2 x 512)
Memory: 3072MB (1536MB per GPU) GDDR5 @ 853.5MHz real speed or 3414MHz effective speed, 768-bit (384-bit per GPU)
Texture units: 128 (2×64)
ROPs: 96 (2×48)
TDP: 365W
Power connectors: two 8-pin
GPU VRM: 10 phases (5 phases per GPU)
GPU Memory VRM: 4 phases (2 phases per GPU memory)
Price: US $699

http://www.geeks3d.com/20110324/nvi...ications-and-reviews-hd-6990-is-still-faster/

A suppose Tom Peterson doesn't know either.

 

[AtenRa]

Two days before kitguru article

http://forums.anandtech.com/showpost.php?p=31489737&postcount=453

 

[Keysplayr]

It's not wrong. There are 5 phases, 4 for the GPU and one for the memory.

So it's actually correctly passed on in bolds.

If you're talking about the GTX590 then you, or your source, is incorrect.
5 phases for each GPU, not 4. And an ADDITIONAL 4 for the memory.

Check here about 1:05 into the vid:
http://www.youtube.com/watch?v=5XrVAhcOmmk

Ah, SirPauly beat me to it. :thumbsup:

 

[Grooveriding]

Two days before kitguru article

http://forums.anandtech.com/showpost.php?p=31489737&postcount=453

Going on that what would you say accounts for those smaller components failing on the three cards ? And are they related to the VRMs.

As well what would account for the different MOSFET failure seen at techpowerup and could it be a similar failure as seen in those other three examples or a failure of a different type ?

 

[Grooveriding]

If you're talking about the GTX590 then you, or your source, is incorrect.
5 phases for each GPU, not 4. And an ADDITIONAL 4 for the memory.

Check here about 1:05 into the vid:
http://www.youtube.com/watch?v=5XrVAhcOmmk

Ah, SirPauly beat me to it. :thumbsup:

Thank you.