Rumored (now Denied). nVidia to Recall 600 series cards? (xbitlabs via pnosker.com)

[BD231]

BS, one 670 has died so far on our forums and no 680's. Within the first two weeks of the 7970 release we had three deaths.

 

[cmdrdredd]

Heh...times are crazy. Also there is a quote that has incorrect information in the OP. EVGA never issued an official recall. They have publicly stated this on their forums. Not every one of their cards was bad and it's unrelated to this rumor.

BS, one 670 has died so far on our forums and no 680's. Within the first two weeks of the 7970 release we had three deaths.

Two...mine died a week before the thread popped up. I had already gotten it replaced and everything by the time the thread appeared.

 

[KompuKare]

Wow, if true NV would be in a whole lot of trouble. After bumpgate, this would be a PR disaster, not to mention with shortages of GTX670/680 as it is, that means refunds would be the only viable alternative for customers.

I wonder how performance of a chip can degrade since it's not like they are overvolting these 28nm chips. Perhaps TSMC's 28nm high-performance process for these chips is flaky.

Not commenting on the credibility of this rumour but a badly designed chip could degrade. That's essentially what happened with Intel and Cougar Point. And if it can happen to Intel on a mature process (not sure what P67 and H67 were made on but their chipsets generally get to use capacity on older nodes - so definitely not 32nm like the SB CPUs themselves), it can happen to anyone never mind Nvidia who have made basic manufacturing mistakes before.

[Quoting Nvidia spokeperson:]
"Before making it to the market, all the new GPUs are vigorously tested in various applications; they also pass Microsoft HCL tests, which not only stress graphics chips, but take hours to complete. Therefore, chances that chips may degrade in performance over time are pretty low."

Hm, if a chip has those kind of flaws a few hours of Microsoft HCL is totally worthless. I hope that NV spokesperson is just kidding: the minimum I would expect from anyone working for a semiconductor company is the knowledge that a few drivers tests cannot in any way be considered the equivalent of hardware validation.

 

[Ferzerp]

In an effort to be first, people (including some of these "tech sites") post anything, regardless of truth once one person starts a rumor.

Here is a tip. Ignore anything until it comes from a reputable source. (hint: none of these sites willing to repost that are a reputable source. They obviously will repost anything without any sort of fact cehcking).

 

[cmdrdredd]

I ran OCCT for 12 hours, EVGA's OC ScannerX fur and tessellation test for 6 hours each, and 12 hours maxed out Heaven benchmark. his was with TDP set to +145 and my GTX 670 overclocked. The boost clock was running at 1250Mhz and the memory at 6402Mhz.

So if that type of testing didn't force some degredation (my FPS remained the same in all benchmarks), then I don't think there's truth to this.

 

[tigersty1e]

why would you run occt for 12 hours and the other tests for 6-12 hours?

that proves nothing related to stability of overclock.

its not like prime95 that needs to run through different tests. occt runs the same test over and over and over.

 

[amenx]

I think Nvidia would be the first to know if their products suffered from "serious performance degradation over long periods of heavy load" since thats the type of testing both Nvidia and AMD subject their cards to before release. It would seem highly unlikely that some card owners out there had put their cards under greater stress for longer periods than Nvidia would already have done through months of testing.

But as I have seen in other forums (newly registered) people claiming their 680s burned out and showing photo-shopped pics of cards with burnt vrms and which in the end just made them look pathetic and desperate to rain on Nvidias parade at any cost. While I can not rule out the possibility of problems with the cards, I similarly wont rule out the sort of pathetic bullshit that some would pull just to tarnish the other side.

 

[BD231]

Two...mine died a week before the thread popped up.

Noted.

I had already gotten it replaced and everything by the time the thread appeared.

Yeah EVGA is a real pain in the ass aren't they? lol

 

[tigersty1e]

I think Nvidia would be the first to know if their products suffered from "serious performance degradation over long periods of heavy load" since thats the type of testing both Nvidia and AMD subject their cards to before release. It would seem highly unlikely that some card owners out there had put their cards under greater stress for longer periods than Nvidia would already have done through months of testing.

But as I have seen in other forums (newly registered) people claiming their 680s burned out and showing photo-shopped pics of cards with burnt vrms and which in the end just made them look pathetic and desperate to rain on Nvidias parade at any cost. While I can not rule out the possibility of problems with the cards, I similarly wont rule out the sort of pathetic bullshit that some would pull just to tarnish the other side.

not really. most of these recalls happen because consumers bring to light the issues.

these happen because nvidia/intel/other companies can't test thousands of these products. they just test a handful and use the statistical p test to determine a larger scale issue.

 

[OCGuy]

Rumour started by upset AMD fan?

 

[tigersty1e]

rumor started because of evga's mini-recall.

 

[cmdrdredd]

why would you run occt for 12 hours and the other tests for 6-12 hours?

that proves nothing related to stability of overclock.

its not like prime95 that needs to run through different tests. occt runs the same test over and over and over.

Because if you can't force an error (OCCT has artifact checking as does OCScannerX) by putting a heavy load and letting it heat up, then it's a good baseline indicator. If it gives even one artifact it's not stable for me. If it crashes after 2 minutes that's obvious. Most times, an overclock that is too high will give some error or issue when put under a heavy load. After that I run 3dmark and crysis 2, battlefield 3, dirt 3 etc. Let those run through for a while. Most of the time using OCCT I have been able to get close to a good overclock.

I ran it then went to bed and after breakfast, I checked on it. Shut it off at that time.

 

[Atreidin]

Amd was hoping this to be true so they could increase prices

It's important to the community to have people like you to help us all by inserting off-topic posts into threads to veer discussions into other directions and remind us how awful AMD is. They are constantly stalking us, lurking in the shadows, and waiting to screw us all over at every opportunity. From all of us, thank you for your pointless vigilance.

 

[Ferzerp]

It's important to the community to have people like you to help us all by inserting off-topic posts into threads to veer discussions into other directions and remind us how awful AMD is. They are constantly stalking us, lurking in the shadows, and waiting to screw us all over at every opportunity. From all of us, thank you for your pointless vigilance.

This whole thread is just unfounded alarmism anyway. His post really isn't out of place because the whole thread is totally unfounded rumor.

 

[nyker96]

I say this is work of amd's advertising department, if you can't beat them ... ;]

 

[RussianSensation]

Amd was hoping this to be true so they could increase prices

One of the new Ghz edition 7970s has been on sale on Newegg for a couple of weeks now. Not a single review yet. I think even despite lack of availability of GTX670/680 series, it looks like people are choosing to wait / use Auto Notify than buy readily available 7970s. That doesn't bode well for AMD increasing prices on 7970 series as it is.

I just tried adding 100 units of MSI GTX670 on Newegg and it worked. Looks like new stock keeps coming in and they are just selling out really fast. AMD is asking nearly $100 more for its Ghz edition reference card with a loud cooler over the 670. No wonder there is barely any demand for it.

Amazon has EVGA and PNY too.

 

[Rvenger]

One of the new Ghz edition 7970s has been on sale on Newegg for a couple of weeks now. Not a single review yet. I think even despite lack of availability of GTX670/680 series, it looks like people are choosing to wait / use Auto Notify than buy readily available 7970s. That doesn't bode well for AMD increasing prices on 7970 series as it is.

I just tried adding 100 units of MSI GTX670 on Newegg and it worked. Looks like new stock keeps coming in and they are just selling out really fast. AMD is asking nearly $100 more for its Ghz edition reference card with a loud cooler over the 670. No wonder there is barely any demand for it.

Amazon has EVGA and PNY too.

TD told me yesterday that they had 100 units of the MSI 670 as well.

 

[cmdrdredd]

TD told me yesterday that they had 100 units of the MSI 670 as well.

I just looked and there's an EVGA ref., MSI OC, and PNY ref. in stock at the egg. I notice the PNY is priced $30 higher than every other reference card. I guess you are paying for the lifetime warranty??

 

[3DVagabond]

Wow, if true NV would be in a whole lot of trouble. After bumpgate, this would be a PR disaster, not to mention with shortages of GTX670/680 as it is, that means refunds would be the only viable alternative for customers.

I wonder how performance of a chip can degrade since it's not like they are overvolting these 28nm chips. Perhaps TSMC's 28nm high-performance process for these chips is flaky.

I've never heard of Pnosker.com though. I am going to wait for a more credible source.

This will be a PR disaster if they handle it the same way they handled bumpgate and force people to bring a class action lawsuit against them to settle it. If true (that's a big if) and it's handled well, it will cost them money, but little else. Nobody's too concerned about Intel chipsets ATM, for example.


Edit: I love how some posters have to immediately post something defamatory about AMD. Somehow they have to post something to try and make nVidia look good and AMD look bad.

Instant spin mode: "What's this? Trouble at nVidia? Quick, badmouth AMD, and do it a lot!"

 

[Rvenger]

I just looked and there's an EVGA ref., MSI OC, and PNY ref. in stock at the egg. I notice the PNY is priced $30 higher than every other reference card. I guess you are paying for the lifetime warranty??

You are paying for the HDMI cable lol.

 

[blackened23]

You are paying for the HDMI cable lol.

More like the lifetime warranty. PNY is the only brand that defaults to a lifetime warranty, that is the reason for the cost.

That said, I wouldn't buy it, I don't see myself holding onto a video card more than 2 years.

 

[HeavyHemi]

I'm telling you guys, it's just a matter of time...
http://i17.photobucket.com/albums/b96/Heavyhemi/nvidia_ceo_jen-hsun_huang-1.jpg

Welcome to AT, but we're not 4Chan. Please post with a bit more substance.
-VIRGE

 

[blastingcap]

One of the new Ghz edition 7970s has been on sale on Newegg for a couple of weeks now. Not a single review yet. I think even despite lack of availability of GTX670/680 series, it looks like people are choosing to wait / use Auto Notify than buy readily available 7970s. That doesn't bode well for AMD increasing prices on 7970 series as it is.

I just tried adding 100 units of MSI GTX670 on Newegg and it worked. Looks like new stock keeps coming in and they are just selling out really fast. AMD is asking nearly $100 more for its Ghz edition reference card with a loud cooler over the 670. No wonder there is barely any demand for it.

Amazon has EVGA and PNY too.

GTX 680 supplies would be pretty decent as well, if they were all going into gaming video cards. However, Tesla K10 is much more profitable to NV (their pro cards are about 1/3 of their discrete video card business but generate 2/3 of their profits; the reverse is true for gamer cards), and I'm sure NV has been taking the best binned GTX 680 parts for its GTX 690 and Tesla K10 cards, instead, leaving the leftovers for "regular" GTX 680 cards. As far as I know, no Quadro or Tesla cards use GTX 670 chips.

Due to the above, and due to the nature of yields and die harvesting, why would one be surprised at all that there is a big gap between the availability of GTX 670 vs GTX 680?

 

[3DVagabond]

One of the new Ghz edition 7970s has been on sale on Newegg for a couple of weeks now. Not a single review yet. I think even despite lack of availability of GTX670/680 series, it looks like people are choosing to wait / use Auto Notify than buy readily available 7970s. That doesn't bode well for AMD increasing prices on 7970 series as it is.

I just tried adding 100 units of MSI GTX670 on Newegg and it worked. Looks like new stock keeps coming in and they are just selling out really fast. AMD is asking nearly $100 more for its Ghz edition reference card with a loud cooler over the 670. No wonder there is barely any demand for it.

Amazon has EVGA and PNY too.

New GHz edition? Where does it say that? Are are thinking that because they aren't calling it an O/C model? Could be, I guess. Pricing is $10 above the retail price of the reference 7970. As much as you'd like it not to be, the 7970 is priced to compete with the 680, not the 670. This could prove to be a (another!) mistake for AMD though. We'll have to see if they have to drop 7970 pricing further to compete with the 670, instead.

 

[RussianSensation]

Due to the above, and due to the nature of yields and die harvesting, why would one be surprised at all that there is a big gap between the availability of GTX 670 vs GTX 680?

Not sure where you got that part of your statement as I haven't seen anyone mention GTX670 vs. GTX680 availability. In my post I said that despite GTX670/680 being hard to find, HD7970 cards don't seem to be selling well. It's not surprising of course since GTX670 = HD7970 at stock for $50-80 less.

New GHz edition? Where does it say that? Are are thinking that because they aren't calling it an O/C model? Could be, I guess. Pricing is $10 above the retail price of the reference 7970. As much as you'd like it not to be, the 7970 is priced to compete with the 680, not the 670. This could prove to be a (another!) mistake for AMD though. We'll have to see if they have to drop 7970 pricing further to compete with the 670, instead.

That diamond card has 1025mhz reference clocks but their original reference card had 925mhz clocks. It also launched the same week as Sapphire released the new 950mhz 7950 card. During that week many reports came out that AMD is following up with higher clocks 7950/7970 cards. Just my guess this is the new wave of cards releasing slowly. Hopefully we'll see more of these. I could be wrong though but my point still stands that a 1Ghz+ reference 7970 is pushing $490 before shipping....and still loses in performance, noise, component quality and price to the Asus 670 TOP.

The bolded part, AMD can price it at $5,000 for all the market cares. GTX670 has the same performance as an out of the box 7970. So HD7970 is priced against a 680 but has performance of a 670 which is why it's overpriced. It's only justifable if overclocking is considered. That means out of the box the 7970 competes with 670s and factory pre-overclocked 670s = 680 which are even faster! Reference 7970s have an MSRP of $479 based on AMD's pricing, but are slower than a $399 Gigabyte Windforce 670. How does that make any sense? It doesn't. AMD might be more competitive in other places in the world but 7970's pricing in US / Canada seems high.

Another way to think about it: AMD dropped prices then $500 GTX680 came out. Well now many $400-$450 GTX670 with factory preoverclocks offer GTX680 level of performance and have quiet coolers to boot. AMD's 7970 pricing is even more out of sync now then when GTX680 launched since GTX670 OC is cheaper, quieter and faster. I guess AMD thinks the 3 free games they are throwing in are worth $100.