NVIDIA Announces The New Titan X

I suspect he may have loud high rpm fans blowing at full blast for those temps, not the noise conscious ones people usually use in custom loops.

**That reminded me of an old post years back someone ran 3-4 blocks off a single GTX120mm rad lol. Yea you could do it, at a big delta with some ear bleeding Delta fans.

Well that chart does say push only :p.

In my experience when it looks too good, add 10c delta at least lol.
 
yup! Mine shipped this morning. Should be here by Monday!
Same here, delivery expected by Monday! I also pre-ordered the EK backplate, but separately as it looked like that item wasn't shipping until late August and I didn't want to chance delaying the waterblock... and I just found out that I'm traveling for work all next week. :( Well, at least next weekend will be fun!

Using a 360 and and 480 rad, both in a push/pull config with 1850 Typhoons, so cooling is adequately covered. :D

Plan to OC the hell out of it and sell my 980TI's. Should hold up well until Volta arrives in a couple years.
 
Last edited:
Same here, delivery expected by Monday! I also pre-ordered the EK backplate, but separately as it looked like that item wasn't shipping until late August and I didn't want to chance delaying the waterblock... and I just found out that I'm traveling for work all next week. :( Well, at least next weekend will be fun!

Using a 360 and and 480 rad, both in a push/pull config with 1850 Typhoons, so cooling is adequately covered. :D

Plan to OC the hell out of it and sell my 980TI's. Should hold up well until Volta arrives in a couple years.

Yeah, I have a 3x140mmx45mm thick rad in push up top and a 2x140mmx80mm rad in push-pull up front.

All fans are Noctual iPPC Industrial 140mm 2000rpm. They can wake the dead when at full blast, but have a minimum rotational speed of 480rpm and will even turn off if the PWM signal goes low enough. At slow speeds they are damned near silent.

I plan on controlling my fan speed with a third party PWM fan controller using a water temp sensor. I'm thinking of setting the target temp for the loop to 30C, as I'd rarely be using the computer with my office hotter than 75-80F (24C-27C). I usually set my AC to 68F (20C) so that gives me a 10C delta before max fan speed during normal use which ought to be more than enough, but I'd be open to suggestions.

Anyone know what kind of GPU max temp I should see, with a loop water temp of 30C? What about 35C?
 
Coolabratory Liquid Ultra resistor shunt mod, which netted me about 10-15% on the power limit:

12325925

That's the resistor? lol that makes it too easy. Thanks for the pic. How many are you going for this time Vega? Just two? Can you only do just two?

Oh I was thinking, you can use 1/4" thick insulation/sealant foam that is sold in short rolls at a hardware store to box the resistors in as a safety measure. Its kind of like the tape method of CLU application. CLU will stick to the tape so it doesn't go where you don't want it. And the insulation in this case will keep the CLU from getting on other components. I use the insulation on rads to create a seal between rad to fan, and fan to fan on the radiator.

https://smile.amazon.com/M-D-Building-Products-6577-Sponge/dp/B000HE5Z9U/
 
Last edited:
...
As for the shunt mod, you paint all three R50 resistors with a thin layer of CoolLabritory Liquid Ultra. This lowers the resistance and lowers power target reading giving you more amperage headroom.

So you don't wanna completely short them then?

I thought that would completely remove the limit altogether. I mean if you're going this far anyway right?

I didn't bother on my 980 since it never throttled due to power limited, though that was due to modded bios which we can't do right now....
 
Personally I have plenty of confidence in AIO solutions. I've never used the EVGA branded ones, but everyone in the industry uses the same two suppliers so they should be of equal quality to the Corsair ones I've been using for 6 or 7 years thus far without an issue.

That cooler looks awfully large. Does it allow you to install the two cards in SLI properly?



Ahh, so all it does is make the card think that it is using less power than it actually is, so the 120% power limit doesn't kick in?

Yes to both.

That's the resistor? lol that makes it too easy. Thanks for the pic. How many are you going for this time Vega? Just two? Can you only do just two?

Oh I was thinking, you can use 1/4" thick insulation/sealant foam that is sold in short rolls at a hardware store to box the resistors in as a safety measure. Its kind of like the tape method of CLU application. CLU will stick to the tape so it doesn't go where you don't want it. And the insulation in this case will keep the CLU from getting on other components. I use the insulation on rads to create a seal between rad to fan, and fan to fan on the radiator.

https://smile.amazon.com/M-D-Building-Products-6577-Sponge/dp/B000HE5Z9U/

Just two cards. 3 and 4 way are basically just use for benchmarks these days.

So you don't wanna completely short them then?

I thought that would completely remove the limit altogether. I mean if you're going this far anyway right?

I didn't bother on my 980 since it never throttled due to power limited, though that was due to modded bios which we can't do right now....

You can completely short them with wire. This thin layer of CLU gave me about a 15% power target lowering, which ended up well.
 
Found someone to take my old 970's off my hands for $500, so that'd make it $700 for one of these...

Next time it's in stock, I'm thinking the credit card is coming out finally. :)

This ought to really smooth things out in some games, though I should really work at boosting the OC on my CPU too to make the most of it probably.
 
You can completely short them with wire. This thin layer of CLU gave me about a 15% power target lowering, which ended up well.

I wonder if one could short them with something slightly less likely to run onto other components. Maybe a conductive trace pen? Not sure how removable they are if you change your mind though.
 
I wonder if one could short them with something slightly less likely to run onto other components. Maybe a conductive trace pen? Not sure how removable they are if you change your mind though.

Like I wrote above you can encase the resistor in foam, like the weatherstripping. Get a wide single sided foam, cut a rectangle out and stick over the resistor. If it happens to get reflowed and move it will be caught in the foam. One could even tape over the foam casing. hmm thats even better.

Also, I read that ya don't want to use copper wire as that will affect the resistance and drop your power target limiting you to 135w. Too bad about the wire, because the good ole gum mod would be effective here.
 
I wonder if one could short them with something slightly less likely to run onto other components. Maybe a conductive trace pen? Not sure how removable they are if you change your mind though.

Well, if you're only gunna go halfway why just bust out the ole pencil...
 
I just solder over the resistors... which isn't for everybody.

Also, I read that ya don't want to use copper wire as that will affect the resistance and drop your power target limiting you to 135w. Too bad about the wire, because the good ole gum mod would be effective here.

All these methods drop the resistance of the resistors. Including copper wire, in parallel with the resistor, will drop resistance and increase power limit.
 
The Arctic Accelero III is a 3-slot cooler. You can still SLI cards using them if you have an E-ATX motherboard like Vega does and still have enough room for adequate airflow.
Yep, its a HUGE cooler.
I had a 290x with an AC Accelero Xtreme III, it "just" fitted in my Antec 900.
I sold it to a friend and we had to chop the drive bay up on his Coolermaster CM90 case to get the length of the card in.
I had to drill new holes to make more securing points to get it rock solid gain.
For dual card you definitely need a motherboard that places the cards far apart.

The AC Accelero Xtreme IV is just as long and a bit wider!
The rear card heatsink pokes out quite a distance and needs fans to get its full effect.
The advantage is that nothing is glued, everything is cooled by the 2 huge heatsinks and the original cooler can be replaced in the event of a warranty issue.
The hardest part of fitting it is cutting a plastic shield to shape to prevent unwanted components touching the main heatsink.
This cooler also fitted in my Antec 900 by the skin of its teeth. There is a little notch in the drive bays the cooler drops into, otherwise I would have had to cut my case.
(attached to an EVGA ACX2.0+ 980ti)
 
I just solder over the resistors... which isn't for everybody.



All these methods drop the resistance of the resistors. Including copper wire, in parallel with the resistor, will drop resistance and increase power limit.

Yeah, I can't imagine copper wire (which is the conductor in just about every wire you buy) having higher resistance than a resistor :p
 
All these methods drop the resistance of the resistors. Including copper wire, in parallel with the resistor, will drop resistance and increase power limit.

Yeah, I can't imagine copper wire (which is the conductor in just about every wire you buy) having higher resistance than a resistor :p

That contradicts what I read here.

Cant do copper wire, resistance will be too low and put the card into fault mode and lock you at 135MHz.
The reason more people are willing to try this over replacing the whole resistors is that it is far less permanent and can be wiped away to undo the mod. If you de-solder the current resistors and replace them with 3mohm ones then guaranteed you cannot get a warranty approval.

Tutorial: Power Target limit hardware mod (Shunt mod) for Titan X and many other Nvidia GPUs
 
No it doesn't. They are saying the resistance is too LOW, not too HIGH. That makes sense. :p It also suggests that just using solder might also be a risk, as you can get very low resistivity that way too.

I didn't say too low or too high lol. Using wire will drop tour PT and limit you to 135w. It's right there in the post, not sure what you're still on about.
 
Samsung Elec gets order for Nvidia's next-gen GPUs: Chosun Biz

The paper said Samsung would start making the next-generation GPUs using its 14-nanometre production technology before year-end, based on the U.S. company's Pascal architecture. It did not specify the value of the order.

A Samsung Electronics spokeswoman declined to comment, while Nvidia could not be immediately reached for comment.

This doesn't make much sense.
 
I didn't say too low or too high lol. Using wire will drop tour PT and limit you to 135w. It's right there in the post, not sure what you're still on about.

That's interesting because I have multiple cards with giant globs of solder. I wonder what that guy's source is. Maybe they accidentally disconnected a resistor causing the fault? With the amount of solder I use the resistance is less than any copper wire I'd ever pick and I have never had an issue.

I'll have to keep this in mind as a possible cause if I ever get locked at 135Mhz. :).
 
What I'm saying is it doesn't make any sense to shrink to 14llp from 16ff while there is no competition in sight.
 
What I'm saying is it doesn't make any sense to shrink to 14llp from 16ff while there is no competition in sight.

I'm saying it's not about performance... it's about cost. If you can get the same performance out of both then the lowest cost wins. A.K.A. "lowest cost technically acceptable".

I've read 14llp and 16ff are approximately equal? I dont' look into it much. But I've also read 10nm TSMC is equal to 14nm Intel. It's all marketing...
 
TSMC 16nm scales clock very well. I dunno, it sort seems like no point when it'll be a year before there's anything from AMD at the midrange to high end rebuttal.
 
I'm saying it's not about performance... it's about cost. If you can get the same performance out of both then the lowest cost wins. A.K.A. "lowest cost technically acceptable".

I've read 14llp and 16ff are approximately equal? I dont' look into it much. But I've also read 10nm TSMC is equal to 14nm Intel. It's all marketing...
It will be interesting to see how Samsung's process turns out. Apparently the current 14nm process is not as good as TSMC's 16nm.

Apple A9 - Wikipedia, the free encyclopedia
 
Neither my wife nor I have any issues with the Samsung-sourced A9 in our iPhone 6s Plus phones. People will always find something to complain about.
 
I haven't paid attention to it honestly, but my friend with a 6S who is a huge tech fan like I am (so I trust his research) told me that some people saw decreased battery life, likely from higher power consumption. Just thought it might be relevant here.
 
If one of us had a TSMC-sourced one and the other had a Samsung-sourced one I'd still bet it would be hard to tell a difference. Her and my usage patterns are so different it would negate any sourcing 'penalty'.

The gist of the articles I did read about the 'controversy' was essentially where one would have an inherent battery life advantage at certain functions the fact that phones are generally used for more than a single operation would negate the advantage TSMC had in power usage over Samsung.
 
I'm debating if I should install my block right away or wait for the backplate. I don't really feel like pulling the loop apart again to install the back plate later on..
 
I'm debating if I should install my block right away or wait for the backplate. I don't really feel like pulling the loop apart again to install the back plate later on..

The loop doesn't ahve to come apart does it?

I've never installed an EK block before, but I'd imagine that the worst case you just need to disconnect the block from the GPU without breaking into the loop (just be careful so you don't pull any tubes off, but they are usually stuck on there pretty well.

Best case, it might be able to be installed without removing the block from the GPU at all, by just doing it screw by screw?
 
The loop doesn't ahve to come apart does it?

I've never installed an EK block before, but I'd imagine that the worst case you just need to disconnect the block from the GPU without breaking into the loop (just be careful so you don't pull any tubes off, but they are usually stuck on there pretty well.

Best case, it might be able to be installed without removing the block from the GPU at all, by just doing it screw by screw?

Actually, according to the install instructions, you are actually SUPPOSED TO install the block first, and then install the backplate, so you won't have to take anything apart. (At least if it is the same as the old Titan X block) I might just try to put my backplate on without even removing the GPU from the PCIe slot, in order to minimize the stress on the tubes.
 
Actually, according to the install instructions, you are actually SUPPOSED TO install the block first, and then install the backplate, so you won't have to take anything apart. (At least if it is the same as the old Titan X block) I might just try to put my backplate on without even removing the GPU from the PCIe slot, in order to minimize the stress on the tubes.


Further confirmation from the EK rep over on the OCF:

akira749 said:
The backplate and the block share a certain amount of screws. The block is secured by 15 screws and out of those 15, 6 are used to secure the backplate. So yes you will be able to install the backplate without completely removing the block. :thumb:
 
Actually, according to the install instructions, you are actually SUPPOSED TO install the block first, and then install the backplate, so you won't have to take anything apart. (At least if it is the same as the old Titan X block) I might just try to put my backplate on without even removing the GPU from the PCIe slot, in order to minimize the stress on the tubes.

I assume Rizen means that he'd potentially have to pull the loop apart because he wouldn't otherwise have room to install the backplate with the card up and running in the loop.

Man, I am getting really tempted to buy one of these...
 
I assume Rizen means that he'd potentially have to pull the loop apart because he wouldn't otherwise have room to install the backplate with the card up and running in the loop.

Man, I am getting really tempted to buy one of these...


Possibly.

I feel like unless you are doing multiple cards, the GPU is usually the top card.

Unless you have a very space constricted case, it should be possible to remove the six screws that hold the back plate on, slip the backplate in, and then tighten them, down using a short screwdriver.

If your GPU isnt your top card, it's not that difficult to remove a card not attached to the loop.

If you have SLI you may be SOL though, as you'd almost definitely need to remove the video card in order to be able to tighten the screws. It could be possible to pull the gpu VERY CAREFULLY and do it, but I'd be uncomfortable about putting too much force on the tubes.

This is a rationale for spending the extra money and putting quick disconnects on each side of every block. (I didn't)
 
I'm debating if I should install my block right away or wait for the backplate. I don't really feel like pulling the loop apart again to install the back plate later on..

I always put QDCs on my gpu for this reason. It makes working on them easy, well easy enough. You're more likely to pull the gpus than the cpu in general use. Also, I'm not big on assembly with the card in the slot and you're more likely to fubar something with such little room to work with.

3Q9e3Lfm.jpg
 
Back
Top