2990wx watercooling try 2

[Markfw]

no... the thread does not seal... the orings do the seal.

Thanks ! I think I have everything ordered, and I think I know which case it goes in, and I think I know how to plumb and install it. Now if the parts would just show up, I can get this on the road ! But Monday will tell me how hard this may be, to do what I want (32 cores/64 threads@ 4.0 ghz or more from 3 ghz stock)

 

[Fir]

No its not... the TDP on TR2 32c is 250W.
That means even if its overclocked to 4ghz, that is not going to double the TDP on the chip.
If that did, that would be one hell of a poorly efficient chip.

At most it will put out 350W, while drawing probably 500W of power.
If it put out 500W of heat, then it would mean it would easily draw way more then that in power, which would make it the crown chip in inefficiency.



Your butching the laws of thermo.
Thermodynamics states water in a closed loop try to goto equilibrium, however the carrying potential of liquid is determined by FLOW.

So you want to keep that cold plate as cool as possible, you need greater flow, without a bottleneck somewhere else as the system is trying to attain equilibrium.

350W of heat @ 1gpm gives roughly water 1C increase in temperature.

Going by Math here a 360 radiator will be able to dissipate 450W while keeping roughly a 5C delta @ 1800rpm Yate Loon fans.

Its about keeping bottlenecks in check. Eventually in a LCS, the bottleneck will be either @ flow, or @ ambient.
You can never bypass that bottleneck called Ambient... If you do, you just broke the physics of our world, unless your cheating and running a chiller.

If your going to have that much heat in one spot, then your only solutions are to:

1. Increase Flow -< because again, thermo states Flow is the value in which water can potentially move X heat.
2. Keep the loop shorter <-- because a shorter loop has less restriction and increase flow.
3. Change Coolant to something more efficient... so you can change the carrying potential of the liquid. (like liquid gallium)


You guys are making him go excessive...
And yes i fully know what excessive is.. i used to be called the king of excessive..
My last system had 4 radiators in 3 loops using 6 pumps, holding essentially 1 full gallon of distilled h2o.

Did i need it? No... absolutely not... But its a bad habbit i ended up taking from XtremeSystem, which i learned to mellow out a lot.

But what does he need it for a single CPU loop.

He just needs a D5 + a thick high performance 360 radiator with great fans + a great cpu block, to keep that TR2 chip in check while being under constant load for 24/7.

Also the more complex a loop becomes, the long the downtime, and mark is not the type of person that appreciates having his PC down (its a crunching machine, so each wasted time means loss of points.). So keeping a simple loop, will be easier to service, and easier to setup.

Simply put, you cannot expect those temps to stay this low over long times. If you actually load the system heavily for many hours, you will see coolant temps skyrocket. Increasing flow beyond 1 gallon per minute will do nothing to stop this. You must keep coolant as close to ambient as possible. 5C delta is borderline mediocre performance! 2-3C should be the goal.

Great fans suggest loud. Some people are OK with this. One of the joys of liquid cooling is the quietness of the system. Larger rads with slow fans will accomplish this. Nobody wants to hear three 3000 rpm fans on a 360 rad which is NOT going to keep a 32 core system adequately cool under 24/7 load.

And again, it's only excessive if you design it to unity (liquid temps match ambient with measurement points taken anywhere in the system) AND you're looking to add more radiators!

And Xtremesystems, ugh. Too much koolaid drinking over there. From dye in coolant to completely un-necessary additives doing more harm than good.

 

[aigomorla]

And Xtremesystems, ugh. Too much koolaid drinking over there

Im lost in this statement...
What time period are you talking about?

Most of the LCS we have in today's hobby was born from XS.
Cathar the father of the first focused nozzle waterblock, to Marci, the developer of the thermochill PA series radiator.
Even all the Presidents from LCS, like Gabe @ Swiftech, Eddy from EK, Paul from XSPC, Vincent from Bitspower, all have traces to XS.

We even had linus who would try to rep NCIX, and Martin who designed the very first flow calculator. Skinee and Vapor who also did a lot of testing.

True its not what it used to be now, but back when i was active on it, it was one of the front leaders who shaped LCS to be what it is today.

 

[Markfw]

Simply put, you cannot expect those temps to stay this low over long times. If you actually load the system heavily for many hours, you will see coolant temps skyrocket. Increasing flow beyond 1 gallon per minute will do nothing to stop this. You must keep coolant as close to ambient as possible. 5C delta is borderline mediocre performance! 2-3C should be the goal.

Great fans suggest loud. Some people are OK with this. One of the joys of liquid cooling is the quietness of the system. Larger rads with slow fans will accomplish this. Nobody wants to hear three 3000 rpm fans on a 360 rad which is NOT going to keep a 32 core system adequately cool under 24/7 load.

And again, it's only excessive if you design it to unity (liquid temps match ambient with measurement points taken anywhere in the system) AND you're looking to add more radiators!

And Xtremesystems, ugh. Too much koolaid drinking over there. From dye in coolant to completely un-necessary additives doing more harm than good.

Well guess what. Anandtech and others say that at 4 ghz, its a 500-600 watt chip !! @1.375 vcore.

As of the reviews this am. And they are using the motherboard I have on the way, the MSI 19 phase one.

 

[aigomorla]

500-600 watt chip !! @1.375 vcore.

my statement exactly...

500W power draw chip does not mean 500W heat draw...

that would make it one hell of a inefficient chip, where 100% of the power given is all lossed due to heat...

At worst id say its 33% efficient... so a 600W chip would at most put out 400W in heat... hence doing the math, a 360 is enough, since a 360 can handle 450W with a 5-7C delta... liquid coolant temp - ambient when 1800rpm fans on a MCR320 class.

The Nexxxos is a much more efficient radiator then a MCR320, so, id expect a drop in deltas to be another 2-3C's with Gentle Typhoons powering them.

 

[Markfw]

my statement exactly...

500W power draw chip does not mean 500W heat draw...

that would make it one hell of a inefficient chip, where 100% of the power given is all lossed due to heat...

At worst id say its 33% efficient... so a 600W chip would at most put out 400W in heat... hence doing the math, a 360 is enough, since a 360 can handle 450W with a 5-7C delta... liquid coolant temp - ambient when 1800rpm fans on a MCR320 class.

The Nexxxos is a much more efficient radiator then a MCR320, so, id expect a drop in deltas to be another 2-3C's with Gentle Typhoons powering them.

I got the 2150 rpm typhoons. The rad is due by Friday, pump thursday, waterblock and fittings, no idea. CPU is supposed to ship today, but not yet.

 

[StefanR5R]

500W power draw chip does not mean 500W heat draw...

x W power draw at the processor socket does exactly mean x W heat transfer through the cooler (or melting the hardware).

But x W at the processor socket < y W at the mainboard power sockets < z W at the wall.
y = x + other board components + VRM losses
z = y + peripherals + PSU losses

How and where Ian Cutress took the 500 W is not yet published, as this part of his review is still in the works.

 

[dlerious]

Since it's your first time, do a leak test. Get everything set up, but don't plug power into your system - you don't want to fry anything. Put paper towels around or near your fittings. use a wire/paper clip to jump the green and a black wire on your 24-pin cable - with PSU off. Fill your reservoir nearly full and turn your PSU on until the water level is almost gone. Refill and repeat until water level is steady. Top off your res and let run overnight - unless you notice something right away. You can use something like this https://www.ekwb.com/shop/ek-atx-bridging-plug-24-pin instead of wire and you could also use a different power supply if you wanted, as long as there's no power going through your system.

I'd also flush the rads, haven't used Alphacool, but I've had some that had flux and residue in them. Fill the rad 1/2 or 3/4 full with warm tap water. Plug and vigorously shake. After you've determined there's nothing in them, flush with distilled water a couple times.

 

[guskline]

Markfw, in the for whatever it's worth column, I have 3 XSPC RX480 v3 rads in series with 12 Gentle Typhoon 1850 fans in my 5960x rig OC'd to 4.4 with a custom watercooled GTX1080TI. I have 2 D5 Vario pumps in series. The fans are connected to a fan controller so I can run them much slower @600-800 rpms, and even at slowest speeds, when maxing out system the temp of the water loop vs my ambient temp is 8 degree difference. When I crank the fans up I can drop the difference to 4 to 5 degrees!

You had a question of how to power up the pump and leak test? If you have an older power supply, mod the power connector by shorting two of the leads (Google it and they will walk you through it). Then you use the on/off button of your power supply to activate the pump.

When you have connected the cpu block, radiator and pump/ reservoir combo with tubing and all of the fittings are snug (don't over tighten) start filling the reservoir with distilled water. When it is nearly full, turn on the power supply connected to the pump a few seconds then turn off. Fill more water and repeat this sequence until the loop is filled. Keep the reservoir @3/4 full minimum and when this happens keep the pump running.you will notice a lot of little air bubbles at first. You can add a little more distilled water while it's running if it gets lower. I put pieces of paper towel near all of my connectors to pick up any leaks. If you find leaks, turn off the pump, tighten the coupling that appears to have leaked, dry the area with paper towel and restart the pump. Some people leak test 24 hours but I do @3 hours and if no leaks I let it run another 6 hours and then fill the reservoir nearly full and connect all the power cables, including the pump, to the main power supply and fire up the system.

 

[dlerious]

Markfw, I'd add a T fitting and a drain valve as well. Makes draining a lot easier. For filling, you can use a funnel, syringe, or fill bottle.

 

[Fir]

my statement exactly...

500W power draw chip does not mean 500W heat draw...

that would make it one hell of a inefficient chip, where 100% of the power given is all lossed due to heat...

At worst id say its 33% efficient... so a 600W chip would at most put out 400W in heat... hence doing the math, a 360 is enough, since a 360 can handle 450W with a 5-7C delta... liquid coolant temp - ambient when 1800rpm fans on a MCR320 class.

The Nexxxos is a much more efficient radiator then a MCR320, so, id expect a drop in deltas to be another 2-3C's with Gentle Typhoons powering them.

Um, no. A CPU is not a motor. If you're pulling 500W at the socket, you sure better had a thermal solution that can handle 500W. If not, it's going to heat up. No solution is 100% efficient hence the reason why sub ambient cold plate design is required if you want your core temps to be lower than 30C over ambient, etc.

And I will say it again, the delta you're shooting for is on the high side. Not everyone is just playing games or running benchmarks. Under 24/7 load (folding/crunching/rendering large, etc.) your temps are going to drift too high for comfort UNLESS you're running stock. This is exactly why a big tower cooler can actually outperform the largest AIO. In a benchmark test, the AIO wins because its fluid temp stays cool enough to act as a buffer for the excess capacity that's not there.

And not everyone wants to hear fan noise as well. So use twice the radiator design, 1/2" ID (always) tubing and fittings throughout, and slow the fans down to under 1000 rpm and your hardware will be happier and certainly your ears too!

Leak tests, yes I do recommend doing this. I don't use compressions (always clamp over barb!) and never had an issue with a leak. I always flush out new radiators too as they will have flux and metal shavings in them. It's a PIA but the fins on the waterblock is where this debris winds up and that's the last place you want it! When I was cranking out 1-2 builds a week I found it was easier to make a rig with a pump and 1 micron water filter (like the ones at Home Depot for whole house water) and let it circulate distilled water overnight. After a month, it was scary to see how much crap was in that filter bowl!

 

[StefanR5R]

I don't use compressions (always clamp over barb!) and never had an issue with a leak.

So far I have been using both compression fittings and barb fittings + clamps, always with EPDM tubing (ID 3/8", OD 5/8"). I always had perfect seal between tubing and fitting, with either kind. What I didn't always get right at very first attempt was the seal between fitting and G1/4 port (to be sealed by the O-ring), e.g. as I don't want to over-tighten if it is an acetal port rather than a brass port.

BTW, the first and only AIO that I had was a modular AIO from EKWB. In addition to O-rings, they used a paste to improve their seals (their 1st gen modular AIOs had leak issues; I had their 2nd gen). That was a bad idea of theirs, because some of this paste got into the loop and clogged up my waterblocks eventually. (When least desired, during the 2017 SETI@home Wow!-Event.)

 

[Fir]

I've never had a compression leak under normal use.
Some custom builds with hoses that can be moved by the user, OTOH, scare me. Especially with long "legs" and high pressure pumps like the RD30.
And if a hose pops on a rig like that it's going to be bad. Might as well hose the box with a garden hose full on, smooth bore!

Those SLI fittings that slide inside each other like an expansion coupling are bad news too! Best looking and sturdy was EK connector blocks. Even now SLI is becoming a thing of the past. I miss dual GPU cards.

I was using spring clamps and Herbies and decided to go back to trustworthy worm drive clamps. Those Lamptron clamps, while looking great, turned out to be a disaster because they are made from soft aluminum and instead of using a steel threaded insert the screw is threaded into the soft aluminum making the chance of stripping a given. And forget about re-use.

 

[aigomorla]

Um, no. A CPU is not a motor. If you're pulling 500W at the socket, you sure better had a thermal solution that can handle 500W. If not, it's going to heat up. No solution is 100% efficient hence the reason why sub ambient cold plate design is required if you want your core temps to be lower than 30C over ambient, etc.

And I will say it again, the delta you're shooting for is on the high side. Not everyone is just playing games or running benchmarks. Under 24/7 load (folding/crunching/rendering large, etc.) your temps are going to drift too high for comfort UNLESS you're running stock. This is exactly why a big tower cooler can actually outperform the largest AIO. In a benchmark test, the AIO wins because its fluid temp stays cool enough to act as a buffer for the excess capacity that's not there.

And not everyone wants to hear fan noise as well. So use twice the radiator design, 1/2" ID (always) tubing and fittings throughout, and slow the fans down to under 1000 rpm and your hardware will be happier and certainly your ears too!

(insert face palm here)

Sigh... No you do not plan a 500W cooling solution for a system which is pulling 500W.
You will NEVER get 100% loss in heat... that again, is not an efficient chip, and no chips for that matter are that leaky.
You also did not account for the draw on the board, and other components on the motherboard.

Also, incase you missed it, mark got Gentle Typhoons 2150.
Im fairly sure he will have them running at 100% all the time.
The numbers i gave mark and everyone are numbers which were calculated by Skinee and Martin back when science was used more heavily then bling for liquid cooling.

And no a big tower outperforming an AIO was clearly explained by me in the above post.
Its more about FLOW and efficiency overall in the loop.

If mark wants to go all out insane, then he would need a new case, a Core X9 / Tower 900 / Caselabs Magnum / Mountain mods.

Which im pretty sure he doesnt want to get.
And he can always run dual loops, or expand his loop later on should he feel that he needs it, but i have already clearly explained the point in diminished returns in LCS, and how poor the results are once u get to them.

You can spend infinite amount of money on your loop, however the improvements will just get smaller and smaller at a exponential cost.

Lastly, the more fans you have sound gets multiplied greater then having the fans run a tad bit faster due to acoustics. Again something science has shown when martin / skinnee / vapor were fore fronting tests on the ideal setup.

 

[Fir]

I did NOT say PULLING 500W.
I said 500W INTO the CPU!
A system that pulls 500W (total) is very low power.
I build systems that pull over 3000W that run on three phase power.

If a CPU is drawing 500W from the socket, you will need to design a cooling system that can remove 500W period.

 

[aigomorla]

I did NOT say PULLING 500W.
I said 500W INTO the CPU!
A system that pulls 500W (total) is very low power.
I build systems that pull over 3000W that run on three phase power.

If a CPU is drawing 500W from the socket, you will need to design a cooling system that can remove 500W period.

500W on the socket does not translate to 500W heat...
Again, there is NO CHIP IN EXISTENCE that will do that....
Not even running current though copper wire straight would do that....

And no comment about the 3000W.... ive planed and advised a large watercooling solution of over 50 PC's in a PC Gaming center, running off a 100G modified fish tank we used as a res, being cooled off 3 Koyo copper racing radiators we pulled from motorsports, and being fed to each PC though a 1 inch main all done professionally by a plumber.

Sigh.... i know what im talking about when it comes to LCS....

 

[Fir]

500W on the socket does not translate to 500W heat...

Wrong! The 500W has to go somewhere!

Again, there is NO CHIP IN EXISTENCE that will do that....
Not even running current though copper wire straight would do that....

Wrong again. I assure you, the square of the current times the resistance. Copper wire can certainly heat up given enough current. Copper is a good conductor so it makes a poor heater. Nichrome is far better at heating hence its use in heaters.

And yes, my 7890XE system at 4.7GHz 1.3Vcore pulls north of 550W from the socket. Before de-lidding and custom cooling the AVX-512 full load would hit >100C and throttle in less than a second! So yes, it is possible. This is why these HEDT boards have crazy power delivery systems. If it doesn't stop, they are going to have to expand beyond typical dual 8 pin on the motherboard. My SR-2 actually used four pin PCI-E connectors for each socket in addition to the 8 pin for this reason. And those processors maxed out at just six cores!

And no comment about the 3000W.... ive planed and advised a large watercooling solution of over 50 PC's in a PC Gaming center, running off a 100G modified fish tank we used as a res, being cooled off 3 Koyo copper racing radiators we pulled from motorsports, and being fed to each PC though a 1 inch main all done professionally by a plumber.

Ok I get the hobbyist stuff. If you want to compare experience level I designed thermal storage systems back in the 1980s for financial buildings. These got popular fast because they could run the chillers at night (when electric was cheap!) and the tubes carrying refrigerant in a tank of brine would accumulate a thick layer of ice. This would keep the water cool that cooled the building and computer systems. They saved about $700k yearly and the system cost $1.2m. Regarding the central system you referenced above you could have gotten away with a much smaller reservoir if you sized your radiators appropriately!



Sigh.... i know what im talking about when it comes to LCS....

I never questioned your LC knowledge but I'm just pointing out the mistake about thermals.

 

[Markfw]

HELP !! I am having a problem getting the compression ring to slide back then forward then tighten. So you put something on the tubing before you start this ?

 

[aigomorla]

sigh i am done with your in your arguement in regards to the 500W.
If you got 500W going to socket, its not all going to the socket.
Most of it infact is going to the VREG's and Msofets along with all the many # phase power regulators the board has that feeds the cpu power.

Why do you think Derbaur blew his board's vregs? Do you think the cpu is direct injected of some sort?
*sigh*

Many of it is lost in heat before it even gets to the cpu, so 500W to the socket is not even 500W.
Also there is no way to messure how much wattage your getting exactly at the socket unless you have a super expensive machine to connect the cpu directly to, as there are way too many points on the cpu u would need to get a reading from.


@ Mark...
Did you get the correct the compression and tubing?
If you got the wrong size for either one, it will be difficult to fit?
Since i see you got XSPC, are you sure you didnt flip the compression ring backwards?
One side has a notich which applies pressure on the lip, so the ring is not reverseable.

How about this setup for CPU ONLY ? The 600 watt is a guess.

XSPC G1/4" to 1/2" ID, 3/4" OD Compression Fitting V2 for Soft Tubing, Black Chrome, 8-pack
$42.99
Only 5 left in stock.
Quantity: 1 Change quantity of XSPC G1/4" to 1/2" ID, 3/4" OD Compression Fitting V2 for Soft Tubing, Black Chrome, 8-pack from 1
Sold by: BE Equipment
Add a gift receipt
and see other gift options

PrimoFlex Advanced LRT 1/2in. ID x 3/4in. OD Tubing Bundle (10ft pack) - Bloodshed Red
$26.95
Only 8 left in stock.
Quantity: 2 Change quantity of PrimoFlex Advanced LRT 1/2in. ID x 3/4in. OD Tubing Bundle (10ft pack) - Bloodshed Red from 2
Sold by: Tyler Direct
Add a gift receipt
and see other gift options

Mark check to see if you got the correct sizes, of if the vendor botched something.
If its as listed above, the ring should slide without any issues, and all you need to do is hand tighten the compression over the tubing.

Make sure u use a quarter tho to screw in the barb on the block/rad.
That is why the barb has a little notch, for you to use a quarter to tighten them.

 

[dlerious]

HELP !! I am having a problem getting the compression ring to slide back then forward then tighten. So you put something on the tubing before you start this ?

You shouldn't need to. There can be a variation with tubing size between batches (not a lot, but enough to cause a really tight fit). I've never used XSPC fittings before, but I've had other brands that I couldn't really use with my Primochill LRT and others that didn't work with my EK ZMT. Sorry I don't recall which were which, I'm using EK, Monsoon, and Bitspower mainly. I have some extra EK fittings here I'd send you if they were the correct size - unfortunately I'm using 3/8 x 5/8.

 

[Markfw]

sigh i am done with your in your arguement in regards to the 500W.
If you got 500W going to socket, its not all going to the socket.
Most of it infact is going to the VREG's and Msofets along with all the many # phase power regulators the board has that feeds the cpu power.

Why do you think Derbaur blew his board's vregs? Do you think the cpu is direct injected of some sort?
*sigh*

Many of it is lost in heat before it even gets to the cpu, so 500W to the socket is not even 500W.
Also there is no way to messure how much wattage your getting exactly at the socket unless you have a super expensive machine to connect the cpu directly to, as there are way too many points on the cpu u would need to get a reading from.


@ Mark...
Did you get the correct the compression and tubing?
If you got the wrong size for either one, it will be difficult to fit?
Since i see you got XSPC, are you sure you didnt flip the compression ring backwards?
One side has a notich which applies pressure on the lip, so the ring is not reverseable.



Mark check to see if you got the correct sizes, of if the vendor botched something.
If its as listed above, the ring should slide without any issues, and all you need to do is hand tighten the compression over the tubing.

Make sure u use a quarter tho to screw in the barb on the block/rad.
That is why the barb has a little notch, for you to use a quarter to tighten them.

I bought exactly what you linked to me. If I use a little water, it sort of works, but not sure if its tight enough. Maybe a little dish soap as a lube ? Its the correct size.

Here is one I tried to tighten, and on just over the "barb"

 

[aigomorla]

you can use a little bit of dish washing soap diluted in water if that works for you.
You dont need to screw them in all the way like that tho...
It needs to be tight hand screwed, and you shouldnt have to use a wrench.

Your not working with nearly enough pressure for you to require having to use a wrench to screw them on.

 

[Markfw]

you can use a little bit of dish washing soap diluted in water if that works for you.
You dont need to screw them in all the way like that tho...
It needs to be tight hand screwed, and you shouldnt have to use a wrench.

Your not working with nearly enough pressure for you to require having to use a wrench to screw them on.

I am just rying to get ready for when the radiator comes. It says to screw the pump directly to the radiator. I will have the fans on the other side of the rad, then the fans will be how it attaches to the case. Or I could find another way to mount it, and then the rad goes on the frame and the fans inside of that. Whats best ?

And with just my hands, its all but impossible to get the ring up to where it engages the threads,.or tighten it, THATS the problem. I think the soap may fix that problem, only on the outside of the tubing.

 

[Fir]

Mark:

Don't use dishwashing fluid!!! YIKES!!!!!

You do NOT want to contaminate your tubing. It's a closed loop system it needs to be kept clean! If your tubing is stiff you can use a hair dryer to soften it. I've never seen fittings like that before. Bitpower and Koolance compressions work smoothly and by hand don't have this issue. If any sharp metal is present and cuts into the tubing walls it can/will cause a leak!

 

[Fir]

sigh i am done with your in your arguement in regards to the 500W.
If you got 500W going to socket, its not all going to the socket.
Most of it infact is going to the VREG's and Msofets along with all the many # phase power regulators the board has that feeds the cpu power.

Why do you think Derbaur blew his board's vregs? Do you think the cpu is direct injected of some sort?
*sigh*

Many of it is lost in heat before it even gets to the cpu, so 500W to the socket is not even 500W.
Also there is no way to messure how much wattage your getting exactly at the socket unless you have a super expensive machine to connect the cpu directly to, as there are way too many points on the cpu u would need to get a reading from.

Power delivery is quite efficient actually. If your demand is 500W and your motherboard power delivery train is 95% efficient this means 25W needs to be dissipated. Easily handled by the heatsinks onboard. This is why motherboards don't really need big heat sinks. ALL of the power goes through the VRMs. There is inherent inefficiency in switching supply topology. It's no mystery. These parts do NOT have an infinite capacity which is strained by the enthusiast pushing the limits of the hardware. Just as a race car driver can blow his engine if pushing it too hard. That's what happens when the safe operating area (SOA) is exceeded. The part will fail, often catastrophically with flames and release of smoke! That's what happens when Derbaur blew up his board.

There is no need to measure power delivery to the cpu. It can be calculated and the power delivery systems on modern boards have just about every setting needed and reading of said values for one to calculate this precisely. It's also possible to calculate the switching losses in the VRMs. It varies widely particularly if the hardcore enthusiast begins tinkering with the PWM frequencies to get the absolute max clock speed that's stable enough for a few CB runs, etc.

And yes, it's no secret, there are examples of CPUs pushing 1000W. Why do you think they need liquid N2 or even He? At those levels, they would immediately collapse and crowbar the rails and its another paycheck (or two!) down the drain.