NVIDIA Admits to Data Corruption Issues with 790i Chipset

[DaMulta]

I like the overclockers on this board.

techPowerUp! FTW!!!!!


It's cool here, because you get a mix of ocing for games, and for records. We may not have the sponsorship that the people on XS have for world records, but we do good with what we have IMO. I have seen some really cool stuff here over the years.

++++ you know that those XS guys get the cool techPowerUp! soft!!!!



DDR3 well hmmmm, I have had it for a couple of weeks now. I havent been able to crank it up to the high high speeds as of yet. I'm still working in the DDR2 speeds.

Low in DDR3 is just as good, or not as good as high end DDR2 is at the moment. This will change here soon, when DDR2 starts to slow down, and they start making a crap load of DDR3.

 

[NU(GFX)T]

I had my problems with NF4 and RAID corruption, and of course I didn't run stock clocks back then so it might just have been an isuue with OC+RAID+SLI.

Non of my windows installations lasted more than a month on RAID configs, then I just gave up on RAID and switched to ATi xpress3200 chipset for A64 CPUs with ULi south bridge the moment I could get my hands on one.(RAID was on ULi chipset)

Man was it a super stable board. I had two ATi cards in CrossFire and OCed CPU and RAID0 config.I never faced data corruption. Not even once. One of the best configs I ever had with that DFI UT CFX 3200 mobo.

That makes me wonder what on earth did nVIDIA do to ULi? We have seen two new generations of nForce chips after nVIDIA's acquisition of ULi (three if you count the NF5) yet they still have data corruption issues. At least I know that ULi can design chipsets wich operate stable when the system is overclocked. Did nV just flush ULi down the drain after the buyout instead of using their expertise in designing chipsets?

 

[GSG-9]

I had my problems with NF4 and RAID corruption, and of course I didn't run stock clocks back then so it might just have been an isuue with OC+RAID+SLI.

Non of my windows installations lasted more than a month on RAID configs, then I just gave up on RAID and switched to ATi xpress3200 chipset for A64 CPUs with ULi south bridge the moment I could get my hands on one.(RAID was on ULi chipset)

Man was it a super stable board. I had two ATi cards in CrossFire and OCed CPU and RAID0 config.I never faced data corruption. Not even once. One of the best configs I ever had with that DFI UT CFX 3200 mobo.

That makes me wonder what on earth did nVIDIA do to ULi? We have seen two new generations of nForce chips after nVIDIA's acquisition of ULi (three if you count the NF5) yet they still have the data corruption issues. At least I know that ULi can design chipsets wich operate stable when the system is overclocked. Did nV just flush the ULi down the drain?

I had issues with my NF4 as well, The intel Chipset on my Foxconn has had corruption twice. But it managed to recover the raid and go back to the green 'healthy' status.

 

[imperialreign]

$500 DDR3 ram has a Cas Latency of 8. I call that immature, I dont expect it to be 2 like DDR, but I expect it lower than 8. Newer technology that runs faster should not be an excuse to stop aspiring to tighter timings.

I can't say CAS-8 is immature for DDR3, considering that DDR3 (IIRC) supports higher bandwidth over DDR2, runs at a lower supply voltage, and is capable of extreme speeds; but based upon what I've seen, read, and experienced with my setup - at the speeds that DDR3 runs best at, you can't have such low timings as found with DDR2 - not unless you have a highly effective DRAM cooling solution to deal with the heat of low timings and an insane clock speed, and even still, because of the bandwidth of the modules, it's hard to break any tighter than a CAS of 5.

I wouldn't say the technology is immature, as it's been around since 04-05. We didn't really start seeing implimentation until last year, though. The price of DDR3, IMO, is immature still - it's too costly for most to justify upgrading a system to, and there's too much stigma, IMO, as to DDR3's performance.

 

[btarunr]

$500 DDR3 ram has a Cas Latency of 8. I call that immature, I dont expect it to be 2 like DDR, but I expect it lower than 8. Newer technology that runs faster should not be an excuse to stop aspiring to tighter timings.

Oh....so since DDR2 1066 operate at CL 5, they're immature too, and so are DDR2 800 at CL 4. I want them all to operate at CL 2 ~ CL 2.5, the timings at which PC-133 SDRAM operated

DRAM latency is high for a reason, being that at that high clock-rate you can't maintin such tight timings. Besides, tight timings don't alone mean better performance, DDR3 scales to higher bandwidth. It's bandwidth that counts.

 

[eidairaman1]

hence why i got the PC 4000 DDR back in the day, Designed for Clock Pushing, now lets get back on Topic which is Nvidia, Not RAM.

 

[AddSub]

Interesting. I see many of you are reporting NF4 RAID issues. I have three nForce4 based systems and I'm running RAID0 on two of them. I have been running RAID0 on them for long time now and I've never experienced any corruption issues or anything similar.

Could it be because all my systems are AMD+NF4? Maybe Intel+NF4 = higher chance of RAID problems/corruption?

 

[OzzmanFloyd120]

NVIDIA has admitted that its new nForce 790i chipset may cause data corruption when overclocking. A support article on the company’s website says:

Unfortunately NVIDIA hasn’t revealed any workarounds or specific details on what causes the data corruption, so it may be best for 790i owners to run at stock speeds until a solution is found.

Source: NVIDIA

Sure am glad I went with the 500i series.

 

[GSG-9]

DDR2 is standard and implemented by all hardware developers currently (save those already using DDR3). All the market for ram, we can generalize and say there 2% on something else, and company's have (EXPENSIVE) boards out for early adopters with DDR3. That makes it mature, its being implemented, improved and used theres is and HAS BEEN a market and incentive for companies to push the technology. DDR3 has NOT been implemented for an extended period and until quite recently has not been implemented by motherboard makers/chipsets.
Wow, good job its been around for three/four years. GDDR5 was announced 4/5 months ago, it must be being used and improved in this time even though its not on the market. Boxx built computers that had 12 cores starting around 2004 therefore 12 core computers should be a mature viable solution by now.

I said I did not expect a latency of 2. High end DDR2 still outperforms DDR3 as far as I know. By the definition of the word, it is a more effective and mature technology.

Could it be because all my systems are AMD+NF4? Maybe Intel+NF4 = higher chance of RAID problems/corruption?

I had a Athlon 64 2800+/3000+/3200 M/Opitron 148 when I had corruption, but I believe mine was due to overclocks.

 

[btarunr]

Two things, price isn't indicative of a product getting mature, secondly timings of DRAM are proportional to their speeds, and it's not possible for manufacturers to make mainstream DDR3 products that operate at the timings of DDR2, just as DDR2 couldn't operate at the timings of DDR..... Find me a DDR2 800 module that is rated to operate at CL 2T (OCZ made one that is rated to run at CL 3T, but that's as far as it went), I'll find you a DDR 400 module that does, similarly, DDR3 doesn't operate at such tight timings because of the high frequency at which it operates.

Finally, maturity of a product doesn't mean the market receives it better (unless maturity means price-cuts). You still get a i975X based motherboard for $175 made by ASUS....that's pretty 'mature'.

 

[GSG-9]

Two things, price isn't indicative of a product getting mature, secondly timings of DRAM are proportional to their speeds, and it's not possible for manufacturers to make mainstream DDR3 products that operate at the timings of DDR2, just as DDR2 couldn't operate at the timings of DDR..... Find me a DDR2 800 module that is rated to operate at CL 2T (OCZ made one that is rated to run at CL 3T, but that's as far as it went), I'll find you a DDR 400 module that does, similarly, DDR3 doesn't operate at such tight timings because of the high frequency at which it operates.

Finally, maturity of a product doesn't mean the market receives it better (unless maturity means price-cuts). You still get a i975X based motherboard for $175 made by ASUS....that's pretty 'mature'.

I did not mention price once.
Then get me a DDR3 module that is at CL4. When 2 came out you could buy CL8-13. going from 8 to 3 is a significant difference. When DDR3 goes to 5 it will be a hell of a lot closer. But it probably wont because companies wont push hard enough for it.

 

[btarunr]

That's my point. As speeds go up, you can't maintain timings that low. Each generation of DRAM has its own standards of timings. Something like CL 6 / 7 for 1600 MHz DDR3 can be considered 'tight'. You can't use the timing standards of a previous generation to judge DDR3. End of it you know which is faster.

 

[DaMulta]

I want 2000Mhz at 3-3-3-10 what's wrong with that?

 

[btarunr]

Nothing wrong, just don't expect them to be priced mainstream, and definitely not in the near future. With Cell-Shock, Crucial, OCZ and Corsair our there, it doesn't look impossible.

We still don't have DDR2 1200 kits that run stable at 3-3-3-10 right now, talk about DDR3 2000.

 

[imperialreign]

I said I did not expect a latency of 2. High end DDR2 still outperforms DDR3 as far as I know. By the definition of the word, it is a more effective and mature technology.

Maybe as far as speed and timings are concerned, but when compared to the bandwidth of DDR3 . . . no, it's just not possible. DDR3 can transfer more blocks of information per clock cycle than DDR2 can.

All I'm saying, is that one can't compare performance of DDR2 and DDR3 based upon latency timings - as has been pointed out a few times already, with higher memory clock speeds come loose timings. There's really no way around it as far as system stability is concerned.

One can push DDR3 quite low from white I understand, but the tighter the timings at higher clock rates requires better cooling solutions for the DRAM modules beyond what a stock or aftermarket heatspreader is capable of. For most setups, you'd more than likely need a good, active cooling solution.

Just to point out: http://forums.techpowerup.com/showthread.php?p=754108#post754108 - although those sticks might only be rated at CAS-8, with an active liquid cooling solution in the hands of skilled OCers, I'd fathom to bet those sticks capable of hitting at least CAS-4 @ 2000MHz. Give the extreme reviewers some time on that one . . .

 

[eidairaman1]

this article is totally lost, it goes from nvidia botching something up to people changing the subject, unbelievable.

 

[Mussels]

this article is totally lost, it goes from nvidia botching something up to people changing the subject, unbelievable.

its cause the fanboys havent come in to defend NV yet, we're just wasting time.

 

[btarunr]

Discussing DDR3 and comparing 780i SLI to 790i SLI and its USP, the DDR3 memory is the ongoing topic here, it has nothing to do at all with the topic going off-track.

Yes, NVidia screwed up with 790i SLI, but we tend to forget, the 790i is NVidia's first chipset with a DDR3 memory controller. They shouldn't have tagged it 'for extreme overclocking', and that's the screw up here, that when you overclock, there's a data corruption issue. Lesson to learn: when coming up with new technology, don't launch it in a big way, don't promise things you can't live up to. They could have done away with the 'fastest motherboard' ad campaign and stuck to 'SLI now with DDR3 support'. But that's no excuse.

 

[eidairaman1]

this just proves nvidia isn't perfect like everyone thinks they are.