sirmonkey1985
[H]ard|DCer of the Month - July 2010
- Joined
- Sep 13, 2008
- Messages
- 22,414
like dan said further up, ask 3DFX why you use 3rd party vendors to make your cards. they learned the hard way when they tried to produce their own shit.
![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
[Bit-Tech] Nvidia accused of crippling board partners' designs
- Thread starter hammerforged
- Start date
like dan said further up, ask 3DFX why you use 3rd party vendors to make your cards. they learned the hard way when they tried to produce their own shit.
Cool Vibrations
Gawd
- Joined
- Oct 29, 2008
- Messages
- 660
Sorry, but if consumers are going to receive a vanilla board that looks absolutely bland and has no visibly-pleasing enhancements, then they should just end up producing the cards themselves. Why would I put this in my system over this? If their GTX 700 series is going to look like shit, then I guess I'll thank them in advance for letting me know what generation to skip.
Eh, 3DFX bought a shitty OEM. STB was going down in flames anyways, 3DFX just rode it down.
Nvidia uses Foxconn for their branded cards sold at Best Buy. Not the best OEM, but worlds better then STB was. Nvidia realizes they can make ALOT of money if they sell direct to customers, and by leveling the playing field them make it easier for themselves to compete.
For the third time, you aren't paying more just for the overclock - you are paying more for the better cooler, which gives better temps and noise (and maybe better VRMs and power circuitry, depending on the card). Get over the 1.175V limit already. It's not new. The limits on the Classified and Lightning card is new, but that doesn't change anything about the other 99% of Kepler cards that have been sold.
Why by a better cooler if there's nothing to do with it (i.e. your overclocks are restricted due to the clamp on voltage)? It's not like the reference Kepler coolers are terrible in the first place.
Like the Keplers do not overclock without voltage tweaks. My 670 went from 980 mhz stock core to 1215 mhz. That's about 300mhz more oomph with few simple tweaks. I know, for some people getting about 30% more is no OC at all, but for me it's quite decent result.
Custom cooler works like charm - no noise, 28 degrees idle, 58 degrees game load. I believe the stock cooler would not work that well.
Actually your 235MHz clock increase is just shy of 24%, quite the over-estimation, and that level of overclock would be easily attainable on a Kepler reference cooler. The temperatures are irrelevant. I'm not sure how different the noise signature would be, but I've only heard one GTX 670 reference and it wasn't terribly loud. Again, there's less of an argument for aftermarket coolers if the hardware is voltage-restricted.
Nearly all GPU overclocks are limited by voltage and stability, not heat. Very rarely will you get a card to throttle due to overclocks, particularly the one's with custom beefed-up PCBs and coolers.
1215-980=235 You can't even round that up to 300.
I agree that this really lowers the viability of custom designed boards. Regardless of what anyone says, slight voltage bumps will give you higher clockspeeds and that has been the tone that custom boards have set for years. To use an analogy, you wouldn't over volt a CPU with a stock cooler - but once you get some good liquid cooling or a great air cooler you can over volt it with very, very little risk. GPUs are similar. If you have adequate components / cooling there is very, very little risk of any problems.
Basically, nvidia is hurting the market for custom boards with this change. While you can overclock slightly without a voltage bump, it will never allow as much headroom as a custom board with an improved PWM / VRM / cooler.
Basically, nvidia is hurting the market for custom boards with this change. While you can overclock slightly without a voltage bump, it will never allow as much headroom as a custom board with an improved PWM / VRM / cooler.
Commander Suzdal
2[H]4U
- Joined
- Jul 29, 2004
- Messages
- 2,134
How is this even news? People were talking about the issue of nV not supporting voltage tweaks from the first days of the Kepler release. The speculated reason at the time (which makes sense IMHO) is that nV has been cherry-picking and pushing GK104's to the wall to make 680's, both because the results were great and because of problems with GK100/110. If the design is already at or close to max safe voltage in order to push a midrange part to high-end performance, why roll the dice and go any farther, with the resulting failures and bad PR?
You all may take it as a personal affront, but out of GPU buyers, how many OC? Out of those, how many would even think about overvolting? As Morgoth mentions (math issues aside) the OC headroom is still there for those who want to squeeze out extra value from their purchase, and again, this isn't news--it's been this way for 7-8 months. If a few exceptions slipped past and nV is now cracking down on them, it's not because of some secret new anti-voltage conspiracy.
You all may take it as a personal affront, but out of GPU buyers, how many OC? Out of those, how many would even think about overvolting? As Morgoth mentions (math issues aside) the OC headroom is still there for those who want to squeeze out extra value from their purchase, and again, this isn't news--it's been this way for 7-8 months. If a few exceptions slipped past and nV is now cracking down on them, it's not because of some secret new anti-voltage conspiracy.
Last edited:
because you could do some pretty serious voltage changes with the evbot. why not stop them from day one instead of months after release complain about it.
i planned on getting a 3rd 680 classified but now i doubt ti because i dont want to gamble and not get a evbot port and i dont buy used graphics cards.
mikeohara
Limp Gawd
- Joined
- Dec 24, 2002
- Messages
- 368
Speaking as a customer who owns one of the higher-end GTX 680s (and I'm looking to actually move to SLI'd 680s in January), I am not really surprised at what nVidia is trying to do here. They're looking to stifle competition from companies like Asus and EVGA, who sell factory-overclocked products while increasing their overall bottom line. I wish I could go back to AMD and the Radeons, but driver issues have haunted me in the past with the 5870 cards.
Why would Nvidia try to stifle competition from Asus and EVGA? Nvidia doesn't sell their own branded cards, so why would they care if third-party makers sell more expensive cards? They get paid for the die either way. I have no trouble believing this is just what Nvidia says it is - they are tired of footing the bill for RMA dies when EVGA/Asus let end-users pour too high volts into them.
I think they are trying to lock the models into a narrow performance slot so if you want more performance you spend more. Not up the voltage and keep up with the next higher model.
That, or maybe there's something to the rumor that the chips are susceptible to degradation from electromigration?
That, or maybe there's something to the rumor that the chips are susceptible to degradation from electromigration?
TheGardenTool
2[H]4U
- Joined
- Mar 12, 2006
- Messages
- 3,820
Yes the company that supplies the professional line of cards is the amateur partner...sigh.
Hmm
Could they not have an over-voltage detection system? Perhaps some separate read-only memory to store both the hardware-monitored and user-inputted voltage-history of the GPU?
Then set a reference-cooling max voltage, with instructions not to exceed or the warranty will be voided.
And perhaps a slightly higher max-voltage for cards equipped with better cooling (hopefully).
I suppose that setup could be abused, but I'm not sure it'd be any worse than what they already have to deal with. Hopefully enough people will consider it fair enough that there won't be many who try to cheat in some fashion.
Then finally - perhaps a Zero-Warranty card as well, hopefully at reduced cost.
Could they not have an over-voltage detection system? Perhaps some separate read-only memory to store both the hardware-monitored and user-inputted voltage-history of the GPU?
Then set a reference-cooling max voltage, with instructions not to exceed or the warranty will be voided.
And perhaps a slightly higher max-voltage for cards equipped with better cooling (hopefully).
I suppose that setup could be abused, but I'm not sure it'd be any worse than what they already have to deal with. Hopefully enough people will consider it fair enough that there won't be many who try to cheat in some fashion.
Then finally - perhaps a Zero-Warranty card as well, hopefully at reduced cost.
i dont believe theyre trying to cut down on RMA costs. the AIB partner is the one building, branding, and warrantying the card. if the user breaks it, the AIB partner replaces it, not NV. i dont buy that excuse for a second until someone confirms that NV is giving out free chips to AIBs.
i totally buy that NV is tired of people buying 670s and OCing them to higher than a stock 680. if someone wants 680 performance they should have to pay 680 prices dammit!
i totally buy that NV is tired of people buying 670s and OCing them to higher than a stock 680. if someone wants 680 performance they should have to pay 680 prices dammit!
I seriously doubt that the AIB partner eats the cost of the die if the die fails. If something else on the card fails, sure, but the cost of a die failure probably goes back to Nvidia. This statement from the Nvidia guy pretty much confirms that Nvidia warranties the die itself: