The GTX 590 Flashing & Overclocking Resource Thread
Here we talk about the overclocking the GTX 590, notoriously known for not overclocking... or at least not easy for those who know a little something.
This thread is not to replace the Overclock.net Owner's Thread (HERE) but to be supplemental to it. We do have new 590 users coming it every day, and it seems that there are many questions (which have been addressed) in some form or another. Since most of the answers are scattered throughout a hundred pages of posts & benchmarks, I thought I would sum up most of it as best as possible and add my own humble experiences.
Quick Links
Part 1: GTX 590 - Q&A - You Are Here. Continue Reading.
Part 2: GTX 590 - Power Draw Limitation Investigation
Part 3: GTX 590 - Overclock Prediction
Part 4: GTX 590 - BIOS Grabbing & Flashing
Part 5: GTX 590 - BIOS Info & Archive
Part 6: GTX 590 - Troubleshooting
Word of Caution
We, the 590 owners, are well aware of the issues brought up with launch. Any user who flashes their card, and lights their cards up Hiroshima-style, will know that they, and they alone are at fault. I bear no responsibilities for your actions.
I will however tell you that if you are cautious though, the risk is small. Only you will be able to determine if the risk is worth the reward.
~RC
Official Response From nVidia Regarding GTX 590 Overclocking:
Source: http://forums.nvidia.com/index.php?showtopic=196244
Questions and Answers
Using driver set Forceware Betas 280.91 and newer, or 267.85 and older. Any drivers inbetween you have to use a BIOS flash.
Well there are risks in anything, if you lose power in Windows when it happens, you may delete your ROM and have nothing to boot from. There is always the chance that you have a very weak card that won't take too kindly to more voltage. In general though, its safe to flash your GPU if you know what you are doing.
It does one of three things depending on which one you are downloading.
1.) It either takes you back to stock (original factory.)
2.) It gives you a new minimum voltage.
3.) It gives you a new minimum voltage and a stock overclock.
It does not disable PDL or OCP, or anything else. The modified EVGA BIOSes also unlock the fan to 100%.
So far so good. We are not oblivious to the cards that died on release, we are aware a lot of people also didn't like to see nVidia "skimping" on the VRMs and just think they cheapened the power delivery system. Tie that in with Sweclockers and W1zzard blowing up their cards very publicly and loud. Neither of them really admitted any responsibility and blamed the cards. W1zzard even scored the 590 a seven despite being extremely competitive to a 6990 which received a solid 9 out of 10. Listen I totally respect W1z, he seems like a great guy, and really analyzes the cards both hardware, software, and benchmarks, but I was disappointed by how he handled the 590s review.
I think everyone should use caution when overclocking this card, or any new card that comes to town. You have to know the limits of the cards, try and follow in the steps that other users have successfully overclocked (with cards that are still running obviously) and you shouldn't (emphasis on shouldn't) have any hardware failure. Do cards die? Yes, all GPUs of every generation can be killed with overclocking & overvolting. The cards seem to be very resolute. The most damage I have ever seen is a BSOD, and I had the highest % overclock (on water keep in mind) until I was absolutely crushed by TiN. Excluding TiN and his evil creation, I achieved 30+% overclock for benchmarking on HWBot, on a card that every tells me is crap.
That can be found here:
HWBOT RangingCain Submission for Heaven Extreme
We hear this a lot. Daily, even. It is usually everyones argument to not buy the GTX 590 when someone asks if they should. Anything from the card is "fail" to "its a weak death trap waiting to blow up, kill you, set you on fire, and then kill your dog." To any outside observer, that sounds really negative, but since I have had a pair of these for a few months, as well as a pair of 580s for six months before that. I can tell you that this cards is not fail, or a danger to your pets.
If I recall correctly, your atypical 570 died with overclocking in a range of voltages that should not have even been allowed for in the first place. That is bad and definitely some responsibility lies with the manufacturers / nVidia. Similar to the issues with ASUS and its 590, or further more by nVidia themselves (who according to their own admission) had failed to turn on OCP / PDL (more on these two later.)
Analyzing the VRMS:
The GTX 580 has a 6 + 2 VRM design, 6 for the Core and 2 for the Memory.
The GTX 570 has a 4 + 1 VRM design, 4 for the Core and 1 for the Memory.
The GTX 590 has a 5 + 1 VRM design, 5 for the Core and 1 for the Memory Controller, for each GPU.
To separate it further from the GTX 570, it used a stronger type of VRMs like in the 580. I don't have the source so I will have to edit it in later, but its the same more robust VRMs of the 580 design. So it uses stronger VRMs in addition to having an extra one.
The use of the number of VRMs was the first thing that everyone picked up on and called it design flaw, even before a single person could buy one. I personally did not think it was a design flaw, although I agreed it definitely hampered its overclocking potential.
The GTX 580 was able to achieve 900~1000 MHz on 6 VRMs depending on the binning of the GPU, and without any other modifications was able to hit as high as 1100~1200 under LN2. The reduction in VRMs saved space on the PCB and saved money. Two things nVidia is probably always looking to do, or any company really. For two watered down 580 GPUs, there simply was no need for such a robust system, all things considered. I even anticipate if they really wanted to limit the potential of the card, they could have gotten it running on 4 VRMs per GPU.
I am just a little more observant and I do assume the billions of dollars we have given nVidia, they spent a little on research, and that it is being at conducted by people with PhDs in the related field. So I analyze what they did, as if they designed it with purpose.
This is just one way to look at the design:
I am not saying this is how it works, but lets assume you needed 6 VRMs as a minimum for the speed of 772 MHz (stock) on a GTX 580. We all know that the 590 is two 580 GPUs sandwiched together and weakened. Physically, most of the parts are identical.
So lets make comparisons to the GTX 580. We can begin by observing the performance levels of 772 MHz by a factor of 6 to determine a VRM for each level of power.
772 MHz / 6 VRMs = 128.667 MHz / VRM, or for every increase of 130 MHz of GPU Core Frequency, you need an additional VRM to deliver the current necessary for that frequency.
We know the basic ratio, lets see how many VRMs a 607 MHz GTX 580 GPU would need:
607 / x VRMs = 130 MHz / VRM = 607 MHz / 130 MHz = x VRMs
x VRMs = 4.669 VRMs
So, since we can't have 0.669 VRMs, much like we can't have 0.669 people, we round up to an even 5 VRMs. Which is enough for 607 MHz clocked GTX 580, and then it has some wiggle room.
That's assuming you needed absolutely a minimum of 6 VRMs for the stock 772 MHz frequency on a GTX 580, and we know that the wiggle room on the GTX 580 was +230-500 MHz depending on the temperature.
I know the math I did doesn't actually mean anything, and thats nowhere near how you calculate how many VRMs you should have on your card, but a lot of people saw 5 VRMs and just started talking trash without actually thinking about it. I just found one way that using 5 VRMs actually kind of makes sense for the clock frequencies using their associated voltages we saw on the stock GTX 590.
A lot of the GTX 590 "fail" and distaste stems from a few cards that were pushed beyond their limits, and pure opinion. I am not a circuit design expert. A lot of people are though apparently, but I still am not seeing any mass graves filled with these cards.
THAT does not mean you should overclock willy-nilly its safe. We know it has wiggle room but how much? That is for you to figure out.
WHQL 275.33 or BETA 275.50. Both WHQL and BETA 275.50 employ very stringent software overclocking protection that amplifies (does not replace) the hardware's over current protection (OCP) and the power draw limiter (PDL). These are two distinct things, but most users are only familiar with having an OCP on their cards. They also have the added a "voltage lock down" and THEN they limited max voltage (it even over-rides BIOS settings!)
Over current protection is hardware on our GPUs designed to prevent damage to the circuity by controlling spikes or increases of current beyond the hardware specification.
Power Draw Limiter is just the phrase we have coined for the GTX 590s special brand of card protection. This feature acts like a strict governor on a car engine. Despite how powerful your engine is, there is a device that "governs" over the power output of that engine.
PDL is a combination of hardware and software in order to limit the work your GPU can do
There is the standard performance limitation on the GTX 590, that once you go past this line, you enter the throttling realm. As you may have seen, and its even been documented to some degree by [H]OCP in their article on the GTX 590 overclocking. They didn't do much research beyond finding where the upper limit is.
The software limitation seems sneaky and underhanded. IF there was a mistake in the design of the card, it was with PDL. Contrary to what everyone says about the amount of VRMs, this would be the only mistake they made. I feel that when they had released these cards, the GTX 590 only had the standard performance limiting PDL. However, it was here through a combination of reviewers and some consumers going overboard with voltage, that they were able to fall through the cracks as the PDL wasn't aggressive enough to prevent damage in some applications such as the notoriously rough Furmark. Furthermore, its apparent that PDL may have even not been activated at the time. So what they have done is created a series of profiles (similar to SLI Profiles) for each title, and some titles are more aggressively power limited than others now. There is a night and day difference in 3DMark11 and Unigine Heaven 2.5 Benchmark. These cards really shine in Heaven, where on the other hand they look very weak in 3DMark 11. To what degree does it affect games? No one knows for sure, and we will probably never see 100% performance unlocked on these cards.
More information on the [H] article can be found here:
[H]OCP Article - http://www.hardocp.com/article/2011/...king_followup/
Around driver release 267.91, we lost the ability to increase voltage with 3rd party utilities. nVidia even condone overclocking at the start (although small) it was enough to overclock about ~5% going on their recommendations. It seems they went back on their original claim and decided to remove everyones ability to overvolt.
The last driver to have the ability to increase voltage is 267.85. In fact, while normally they keep a back log of beta drivers in the archive. You will have a hard time to finding them anywhere on nVidia's site, its almost like they didn't exist.
While we only discovered this recently, it seems that not only did they remove your ability to increase voltage, they put a maximum voltage in the driver, for those of us who know a work around) which overrides any BIOS you have flashed on the cards. Thats only in the more recent BIOS's, the theoretical max voltage is driver 267.85 is still the reference GTX 580 max of 1.201v.
Yes and no. You will more than likely jump between two drivers depending on your mood for overclocking. That will be whatever the latest driver is, and driver set 267.85.
This is an easy question to answer. Visit LaptopVideo2Go.com. Don't let the name fool you, they are the ultimate resource in nVidia / Quadro drivers. They have every driver ever tweaked, edited, leaked, or officially released by nVidia or any of its OEM partners. You simply browse the forums for the release you want. There is nothing wrong with these drivers, LaptopVideo2Go original mission was to aid Laptops with dedicated and integrated GPUs who were receiving absolutely zero driver support by nVidia (who now also do mobile GPU drivers) or the timely (once every 12 months) release by the OEMs. Now a days, they serve mainly as a massive archive, and while their mission hasn't changed much, it seems like nVidia and many companies woke up seeing how popular and necessary it was.
To get any driver you need head here:
LV2Go!
Well with the 275.33 WHQL and above, the voltage is still lock down, meaning its no higher than minimum but it can be adjusted lower. Furthermore, at some point in the driver releases, no matter how high of a minimum/maximum voltage you have set in BIOS, the drivers limit the voltage max to 0.963v. To get an overclock that is higher, you have to go with the older drivers.
This is only desirable for benchmarking, its always more desirable to use the latest drivers due to patches/bug fixes/stability/driver performance/SLi Profiles.
Other than overclocking, yes there is. There is using the SLI Profile Enhancement from EVGA. That will improve game performance, benchmarking too. Alternatively, you would want to visit the next post regarding more about Experimental BIOSes, and understanding why we can't overclock till it just crashes any higher.
Extreme Overclocking & Modding
Many of you know TiN or K1ngp1n by reputation, and yes, TiN decided to mod, freeze, mod again, freeze the GTX 590. His Frankenstein monstrosity can be found @ KingPinCooling.com. Big props to TiN for his achievement with these cards, as very few probably understand what he had to do, and even fewer of us understand that the scores he are getting with the cards are ridiculous. Perhaps not in comparison to SLI 580s, but I don't think its fair to compare the 590 to SLI 580s. Its just not the same for a variety of reasons, but regardless, he has done an amazing job with the 590!
You can find more on their forums here:
http://kingpincooling.com/forum/showthread.php?t=1279
GTX 590 Links & Information
Edited by RagingCain - 8/5/11 at 8:31pm
Here we talk about the overclocking the GTX 590, notoriously known for not overclocking... or at least not easy for those who know a little something.
This thread is not to replace the Overclock.net Owner's Thread (HERE) but to be supplemental to it. We do have new 590 users coming it every day, and it seems that there are many questions (which have been addressed) in some form or another. Since most of the answers are scattered throughout a hundred pages of posts & benchmarks, I thought I would sum up most of it as best as possible and add my own humble experiences.
Quick Links
Part 1: GTX 590 - Q&A - You Are Here. Continue Reading.
Part 2: GTX 590 - Power Draw Limitation Investigation
Part 3: GTX 590 - Overclock Prediction
Part 4: GTX 590 - BIOS Grabbing & Flashing
Part 5: GTX 590 - BIOS Info & Archive
Part 6: GTX 590 - Troubleshooting
Word of Caution
We, the 590 owners, are well aware of the issues brought up with launch. Any user who flashes their card, and lights their cards up Hiroshima-style, will know that they, and they alone are at fault. I bear no responsibilities for your actions.
I will however tell you that if you are cautious though, the risk is small. Only you will be able to determine if the risk is worth the reward.
~RC
Official Response From nVidia Regarding GTX 590 Overclocking:
Quote:
|
Originally Posted by ManuelG
In the web release driver of GeForce GTX 590, we have added some important enhancements to our overcurrent protection for overclocking. We recommend anyone doing overclocking or running stress apps to always use the latest web driver to get the fullest protection for your hardware. Please note that overcurrent protection does not eliminate the risks of overclocking, and hardware damage is possible, particularly when overvoltaging. We recommend anyone using the GTX 590 board with the reference aircooler stick with the default voltage while overclocking, and avoid working around overcurrent protection mechanisms for stress applications. This will help maintain GTX 590's great combination of acoustics, performance, and reliability. NVIDIA has worked with several watercooling companies to develop waterblocks for GTX 590, and these solutions will help provide additional margin for overclocking, but even in this case we recommend enthusiasts stay within 12.5-25mV of the default voltage in order to minimize risk.
These are guidelines only - any overclocking/overvoltaging can void your manufacturer's product warranty. As of 24th March, 2011 |
Questions and Answers
Quote:
| So how do I overclock my GTX 590? |
Quote:
| Is it safe to flash the BIOS? |
Quote:
| What does the BIOS flash do? |
1.) It either takes you back to stock (original factory.)
2.) It gives you a new minimum voltage.
3.) It gives you a new minimum voltage and a stock overclock.
It does not disable PDL or OCP, or anything else. The modified EVGA BIOSes also unlock the fan to 100%.
Quote:
| Is it safe to overclock a GTX 590 in general? |
I think everyone should use caution when overclocking this card, or any new card that comes to town. You have to know the limits of the cards, try and follow in the steps that other users have successfully overclocked (with cards that are still running obviously) and you shouldn't (emphasis on shouldn't) have any hardware failure. Do cards die? Yes, all GPUs of every generation can be killed with overclocking & overvolting. The cards seem to be very resolute. The most damage I have ever seen is a BSOD, and I had the highest % overclock (on water keep in mind) until I was absolutely crushed by TiN. Excluding TiN and his evil creation, I achieved 30+% overclock for benchmarking on HWBot, on a card that every tells me is crap.
That can be found here:
HWBOT RangingCain Submission for Heaven Extreme
Quote:
| What about those VRMs? |
Quote:
| The VRMs are weak though, right? They have all the issues that the reference GTX 570s have, right? |
Analyzing the VRMS:
The GTX 580 has a 6 + 2 VRM design, 6 for the Core and 2 for the Memory.
The GTX 570 has a 4 + 1 VRM design, 4 for the Core and 1 for the Memory.
The GTX 590 has a 5 + 1 VRM design, 5 for the Core and 1 for the Memory Controller, for each GPU.
To separate it further from the GTX 570, it used a stronger type of VRMs like in the 580. I don't have the source so I will have to edit it in later, but its the same more robust VRMs of the 580 design. So it uses stronger VRMs in addition to having an extra one.
The use of the number of VRMs was the first thing that everyone picked up on and called it design flaw, even before a single person could buy one. I personally did not think it was a design flaw, although I agreed it definitely hampered its overclocking potential.
The GTX 580 was able to achieve 900~1000 MHz on 6 VRMs depending on the binning of the GPU, and without any other modifications was able to hit as high as 1100~1200 under LN2. The reduction in VRMs saved space on the PCB and saved money. Two things nVidia is probably always looking to do, or any company really. For two watered down 580 GPUs, there simply was no need for such a robust system, all things considered. I even anticipate if they really wanted to limit the potential of the card, they could have gotten it running on 4 VRMs per GPU.
Quote:
| You seem pretty sure about the VRMs and their sturdiness, why is that? |
This is just one way to look at the design:
I am not saying this is how it works, but lets assume you needed 6 VRMs as a minimum for the speed of 772 MHz (stock) on a GTX 580. We all know that the 590 is two 580 GPUs sandwiched together and weakened. Physically, most of the parts are identical.
So lets make comparisons to the GTX 580. We can begin by observing the performance levels of 772 MHz by a factor of 6 to determine a VRM for each level of power.
772 MHz / 6 VRMs = 128.667 MHz / VRM, or for every increase of 130 MHz of GPU Core Frequency, you need an additional VRM to deliver the current necessary for that frequency.
We know the basic ratio, lets see how many VRMs a 607 MHz GTX 580 GPU would need:
607 / x VRMs = 130 MHz / VRM = 607 MHz / 130 MHz = x VRMs
x VRMs = 4.669 VRMs
So, since we can't have 0.669 VRMs, much like we can't have 0.669 people, we round up to an even 5 VRMs. Which is enough for 607 MHz clocked GTX 580, and then it has some wiggle room.
That's assuming you needed absolutely a minimum of 6 VRMs for the stock 772 MHz frequency on a GTX 580, and we know that the wiggle room on the GTX 580 was +230-500 MHz depending on the temperature.
I know the math I did doesn't actually mean anything, and thats nowhere near how you calculate how many VRMs you should have on your card, but a lot of people saw 5 VRMs and just started talking trash without actually thinking about it. I just found one way that using 5 VRMs actually kind of makes sense for the clock frequencies using their associated voltages we saw on the stock GTX 590.
A lot of the GTX 590 "fail" and distaste stems from a few cards that were pushed beyond their limits, and pure opinion. I am not a circuit design expert. A lot of people are though apparently, but I still am not seeing any mass graves filled with these cards.
THAT does not mean you should overclock willy-nilly its safe. We know it has wiggle room but how much? That is for you to figure out.
Quote:
| Which Drivers are SAFEST for overclocking and why? |
Quote:
| What is the difference with OCP and PDL? |
Power Draw Limiter is just the phrase we have coined for the GTX 590s special brand of card protection. This feature acts like a strict governor on a car engine. Despite how powerful your engine is, there is a device that "governs" over the power output of that engine.
Quote:
| Okay, I get OCP, but still don't understand what the PDL does? |
There is the standard performance limitation on the GTX 590, that once you go past this line, you enter the throttling realm. As you may have seen, and its even been documented to some degree by [H]OCP in their article on the GTX 590 overclocking. They didn't do much research beyond finding where the upper limit is.
The software limitation seems sneaky and underhanded. IF there was a mistake in the design of the card, it was with PDL. Contrary to what everyone says about the amount of VRMs, this would be the only mistake they made. I feel that when they had released these cards, the GTX 590 only had the standard performance limiting PDL. However, it was here through a combination of reviewers and some consumers going overboard with voltage, that they were able to fall through the cracks as the PDL wasn't aggressive enough to prevent damage in some applications such as the notoriously rough Furmark. Furthermore, its apparent that PDL may have even not been activated at the time. So what they have done is created a series of profiles (similar to SLI Profiles) for each title, and some titles are more aggressively power limited than others now. There is a night and day difference in 3DMark11 and Unigine Heaven 2.5 Benchmark. These cards really shine in Heaven, where on the other hand they look very weak in 3DMark 11. To what degree does it affect games? No one knows for sure, and we will probably never see 100% performance unlocked on these cards.
More information on the [H] article can be found here:
[H]OCP Article - http://www.hardocp.com/article/2011/...king_followup/
Quote:
| What does the voltage lock down and max voltage mean? |
The last driver to have the ability to increase voltage is 267.85. In fact, while normally they keep a back log of beta drivers in the archive. You will have a hard time to finding them anywhere on nVidia's site, its almost like they didn't exist.
While we only discovered this recently, it seems that not only did they remove your ability to increase voltage, they put a maximum voltage in the driver, for those of us who know a work around) which overrides any BIOS you have flashed on the cards. Thats only in the more recent BIOS's, the theoretical max voltage is driver 267.85 is still the reference GTX 580 max of 1.201v.
Quote:
| So, I will most likely have to start juggling drivers depending on what setup I want? |
Quote:
| You have mentioned that driver X Y Z is hard to find, so where can I get the driver? |
To get any driver you need head here:
LV2Go!
Quote:
| Okay, so what is so special about using the latest drivers? I keep seeing the voltage of 0.963v, why is that? Is that the highest we can go safely? |
This is only desirable for benchmarking, its always more desirable to use the latest drivers due to patches/bug fixes/stability/driver performance/SLi Profiles.
Quote:
| Is there anything I can do to improve performance using 275.85? |
Extreme Overclocking & Modding
Many of you know TiN or K1ngp1n by reputation, and yes, TiN decided to mod, freeze, mod again, freeze the GTX 590. His Frankenstein monstrosity can be found @ KingPinCooling.com. Big props to TiN for his achievement with these cards, as very few probably understand what he had to do, and even fewer of us understand that the scores he are getting with the cards are ridiculous. Perhaps not in comparison to SLI 580s, but I don't think its fair to compare the 590 to SLI 580s. Its just not the same for a variety of reasons, but regardless, he has done an amazing job with the 590!
You can find more on their forums here:
http://kingpincooling.com/forum/showthread.php?t=1279
GTX 590 Links & Information
- Overclock.Net - GTX 590 Owner's Club
- nVidia - Response & Cautioning of GTX 590 Overclocking
- KingpinCooling - TiN w/ A Monster GTX 590
- LaptopVideo2Go - Awesome Driver Archive
- [H] OCP - GTX 590 Overclocking Article
Edited by RagingCain - 8/5/11 at 8:31pm











































