GTX 590 reviews

[SKYMTL]

PCI-E slot power is usually only linked to a set component on the card which will keep it at or under PCI-E spec which is 75 watts for the slot itself. The power regulation circuity of the card should not have to worry about load balancing between the power connectors and the pci-e slot.

I agree but that doesn't always happen. PCI-E slot power for the 2.0 spec is rated "up to" 150W, hence why some mobo makers have used auxiliary power connectors on their boards for the slot while retaining the PCI-E certification.

However, as I have said in the past: the 75W spec is usually adhered to in order to guarantee backwards compatibility.

[~CS~]

I agree but that doesn't always happen. PCI-E slot power for the 2.0 spec is rated "up to" 150W, hence why some mobo makers have used auxiliary power connectors on their boards for the slot while retaining the PCI-E certification.

However, as I have said in the past: the 75W spec is usually adhered to in order to guarantee backwards compatibility.

Motherboard makers add an auxiliary power connectors,because if they didn't the PCB would need to be more complex,with lots of wide traces to accommodate for high current,adding an auxiliary connector is nothing but a cost cutting feature,in some cases it's done on boards that are meant for benching,because everything is being pushed out of spec.

What you saying makes zero sense to me:"75W spec is usually adhered to in order to guarantee backwards compatibility",PCI-E 2.0 is backwards compatible with PCI-E 1.0 all you need to adhere to is 2.0 standard and it will be compatible. All PCI-E 1.0 devices will work in PCI-E 2.0 slots.

Abandon ship ?

[Sherman Tank]

Instead of being just app detection, it is now hardware based as well in order to limit current before the card exceeds a certain threshold.

I could be wrong, but doesn't this just show that their hardware based solution is ineffective/redundant, considering further power reductions were needed with multiple driver revisions and a rumored bios update?

Overclocking is a value-added feature but is never guaranteed. I know people love to do it but supporting it past a certain extent isn't in NVIDIA's best interest.

It's in Nvidia's best interest if it's cards are robust enough to enable as much OC headroom as is safe for the GPU die, as it allows them to charge either higher prices than they otherwise could, or gain market share.

Then again, you forget that the card CAN be overclocked but is highly limited by its default voltage; a voltage which was put in place to limit power consumption.

While the OCP and locked down low voltage, is to limit power consumption, the reason they are limiting power consumption so vigorously is simply to prevent untimely ends to 590's which they wouldn't have to do if the cards had been sufficiently kitted out.

[SKYMTL]

What you saying makes zero sense to me:"75W spec is usually adhered to in order to guarantee backwards compatibility",PCI-E 2.0 is backwards compatible with PCI-E 1.0 all you need to adhere to is 2.0 standard and it will be compatible. All PCI-E 1.0 devices will work in PCI-E 2.0 slots.

You seem to be missing my point. I am talking about power compatibility rather than the slot compatibility you are talking about. One of the reasons for the compatibility of current 2.0 spec GPUs with the 1.0 spec is the fact that they do not draw in excess of 75W from the slot.

[SKYMTL]

I could be wrong, but doesn't this just show that their hardware based solution is ineffective/redundant.

Bingo. Which lines up perfectly with what I have been saying: I believe any blame in this situation rests with the implementation of their power capping methods rather than onboard PWM hardware being under-designed.

It's in Nvidia's best interest if it's cards are robust enough to enable as much OC headroom as is safe for the GPU die, as it allows them to charge either higher prices than they otherwise could, or gain market share.

Again, I agree.

But then again, I am sure you will admit that the situations where we have seen these cards dying are far above and beyond what normal or even most enthusiast users would do.

While the OCP and locked down low voltage, is to limit power consumption, the reason they are limiting power consumption so vigorously is simply to prevent untimely ends to 590's which they wouldn't have to do if the cards had been sufficiently kitted out.

That's an assumption again. But if we stay in this vein, there is some truth in it. Sometimes the core will fry when pushed too far. This time it was something else but in both situations, the design was pushed far beyond its reference spec. As they say: overclock at your own risk.

Here is my main issue with this situation: AMD markets their card as having great overclocking headroom and they give end users the tools to achieve their claims. Because of this, it is only right to hold them to their claims. On the other hand, NVIDIA makes NO marketing claims of overclocking potential on the GTX 590 but people seem to have unrealistic expectations based on the highly overclockable nature of other refreshed Fermi products.

[InCredible]

I dont think the expectations are unrealistic...if the same core on Single setup can overclock well I think its dual counterpart could as well however id expect cooling to be am issue..but with water then id expect the same clocks..maybe I'm use to amd but the voltage doesn't seem that much..on my 6950 I went up to 1.35v on air water id try 1.4 ..thats close to the same difference in percentage..idk maybe that is unrealistic?

[bamtan2]

maybe I'm use to amd but the voltage doesn't seem that much..on my 6950 I went up to 1.35v on air water id try 1.4 ..thats close to the same difference in percentage..idk maybe that is unrealistic?

you can't compare volts like that. you're comparing apples to oranges. the "why" is an electrical engineering degree or at least high school physics, but suffice it to say that even though the 590 uses less volts than 6990, it still uses more power.

[Sherman Tank]

you can't compare volts like that. you're comparing apples to oranges. the "why" is an electrical engineering degree or at least high school physics, but suffice it to say that even though the 590 uses less volts than 6990, it still uses more power.

It isn't really apples and oranges, they both use the same process, so both have pretty similar silicon, especially since Nvidia fixed their process issues.
The main reason Nvidia's chip uses more power is because they are allot bigger, and when your only running them at slow clocks (590) you basically kill performance/mm2, yet still have much more inefficient baggage like leakage from the extra die area dragging your performance/watt ratio down.

The thermal benefit of larger dies is that because the die has a larger surface area, it's easier to transfer the heat from the chip to the heatsink, thus combating issues with temps, due to higher TDP's.
Also if your not limited by TDP and you have a good architecture, then you can get more performance with a large chip Vs a smaller rivals chip.

The disadvantage of using larger dies, is they will generally be more prone to failure, similar to how yields are affected by larger dies, simply because the is more points of failure, and engineering the chip becomes harder, as the challenges of chip manufacture become amplified.

Personally I think the 580 cores are wasted on the 590, a couple 570 chips would have worked just as well, especially if Nvidia didn't fuse off the memory bus, so that it could use the full 1.5gb of Vram, or instead just doubled up the ram, they could have probably saved some cash using cheaper cores, and then spending the rest of the cards budget elsewhere.

[STEvil]

I agree but that doesn't always happen. PCI-E slot power for the 2.0 spec is rated "up to" 150W, hence why some mobo makers have used auxiliary power connectors on their boards for the slot while retaining the PCI-E certification.

However, as I have said in the past: the 75W spec is usually adhered to in order to guarantee backwards compatibility.

The 150w spec includes the use of a 6-pin power plug.

No power plug, no 150w.

[Micutzu]

@SKYMTL: the power limiting circuits are there to protect the pwm as it is undersized, these kind of failures have happened on GTX 570 as well. It is not a proper engineering solution but a shortcut to cheaper boards that does not take into consideration all situations and enough safety magins, just look at the TDA21211 specs and at real-life power consumption of the 590.

Sent from my HTC Wildfire using Tapatalk

[Sherman Tank]

@SKYMTL: the power limiting circuits are there to protect the pwm as it is undersized, these kind of failures have happened on GTX 570 as well. It is not a proper engineering solution but a shortcut to cheaper boards that does not take into consideration all situations and enough safety magins, just look at the TDA21211 specs and at real-life power consumption of the 590.

Sent from my HTC Wildfire using Tapatalk

This

[Chumbucket843]

It isn't really apples and oranges, they both use the same process, so both have pretty similar silicon, especially since Nvidia fixed their process issues.
The main reason Nvidia's chip uses more power is because they are allot bigger, and when your only running them at slow clocks (590) you basically kill performance/mm2, yet still have much more inefficient baggage like leakage from the extra die area dragging your performance/watt ratio down.

The thermal benefit of larger dies is that because the die has a larger surface area, it's easier to transfer the heat from the chip to the heatsink, thus combating issues with temps, due to higher TDP's.
Also if your not limited by TDP and you have a good architecture, then you can get more performance with a large chip Vs a smaller rivals chip.

The disadvantage of using larger dies, is they will generally be more prone to failure, similar to how yields are affected by larger dies, simply because the is more points of failure, and engineering the chip becomes harder, as the challenges of chip manufacture become amplified.

Personally I think the 580 cores are wasted on the 590, a couple 570 chips would have worked just as well, especially if Nvidia didn't fuse off the memory bus, so that it could use the full 1.5gb of Vram, or instead just doubled up the ram, they could have probably saved some cash using cheaper cores, and then spending the rest of the cards budget elsewhere.

almost all of what you said is incorrect. where are you getting this information from?

[kromosto]

just look at the TDA21211 specs and at real-life power consumption of the 590.

Sent from my HTC Wildfire using Tapatalk

[Sherman Tank]

almost all of what you said is incorrect. where are you getting this information from?

How is it incorrect exactly?

[SKYMTL]

The 150w spec includes the use of a 6-pin power plug.

No power plug, no 150w.

Check out the 2.1 base spec. It doesn't.

Edit: whoops. Members only. I'll try and find the PDF I downloaded before my membership ran out.

[SKYMTL]

@SKYMTL: the power limiting circuits are there to protect the pwm as it is undersized, these kind of failures have happened on GTX 570 as well. It is not a proper engineering solution but a shortcut to cheaper boards that does not take into consideration all situations and enough safety magins, just look at the TDA21211 specs and at real-life power consumption of the 590.

Sent from my HTC Wildfire using Tapatalk

You may have a point but unless I am reading this wrong, the spec sheet's listed power loss / efficiency versus temperature / output current seems about right considering the number used on the 590. The delineation curves only go out of wack when ambient temps reach above the 110C mark.

Maybe I missed something?

[kazukun]

http://www.youtube.com/watch?v=Zeey1...e_gdata_player

[haschioz]

nice setup, what the is that noise^^

[Postmodum]

Nice setup ... but man ... the video is only to say " i have this this and that " !?!?! You should do more testing and less e-peen showing ... Like... games? Oh... and my 2 cents... just disable the physix from gpu and set it to cpu ... i wish there was a consumer cpu that did 90k scores :P

And you are using the first drivers correct? The newer won't let you test at those speeds without OCP kicking in...

[W1zzard]

You should do more testing and less e-peen showing ... Like... games?

console ports on this rig will make you cry for the money spent

[Postmodum]

Yeah ... i know ... but ... i mean ... i don't play Vantage or 3DMark11 ...

Also, missing Heaven and other ....

But ok ... people buy 980x and 2 590's for that ... eheheh ... and he can't update the driver... because if he does ... there goes the 750Mhz performance :P

Sorry... nothing against the user.... just i hate these videos showing all the components over and over again just to see how cool the rig is ... If you had like 80% testing and 20% showing off ... ok ... but no ... you have 80% showing off and 20% testing ... :P

Sorry ... it's just me ...

[W1zzard]

good way to show off this rig and get a ton of youtube traffic would be to install a smoke detector in the case, bump the voltage, run some stability tests and film with 2 or 3 hd high-speed cameras from multiple angles till *boom*

[Postmodum]

ahahahha ... i tested 2 590's ... a Asus and a POV TGT Ultra Charged ... i just couldn't see the benefits of such cards... they never felt solid performers on my hands ... with a 580 you can give the voltage you want ... i mean ... the last card i tested was 580 Lightning ... and you can just give it 1.3v if you want ... even if that doesn't make the card overclock more ... but ... 580 is solid ... 590 ... nope ... far from that ... it's more being held on strings And i had lots of crashes and freezes ...

but ... well... i still prefer single gpu's

Btw... any news on TiN's LN2 benching of the 590's? He had to rebuild the entire VRM circuitry :P

[STEvil]

You may have a point but unless I am reading this wrong, the spec sheet's listed power loss / efficiency versus temperature / output current seems about right considering the number used on the 590. The delineation curves only go out of wack when ambient temps reach above the 110C mark.

Maybe I missed something?

Did they list ambient as TCase or something???