Great stuff! I learned to figure out power phases on PCBs from just looking at them with the help of your videos. I have a dead 580 DirectCUII from ASUS lying around - I'll be taking it apart soon and can get you some pictures if you're interested.
Hey buildzoid, I just want to thank you so much for making these great videos, really interesting to watch, and really helpful as I'm trying to learn more about VRMs, and these videos are really helpful. I was wondering if you maybe shed some light on a question that I have: I don't know if it is like this for gpus as much (still a lot more of your videos to watch), but I know for motherboards, on lower-mid end Z170 boards, they'll have a "8 or 10" phase VRM for the core voltage, which we know is really just a 4 phase channels from the pwm controller. However, upon closer examination, each real phase from the pwm goes into a driver, but then is split between two different phases (inductors, high and low side mosfets.) I was reading Sin Hardware's VRM guide, and he briefly touched upon this, saying that splitting two phases from one driver was not the correct way to do it, and is doesn't cut the switching frequency in half like a doubler does. Because of this, I'm assuming that without a doubler, you don't really have control over each independent phase, instead the driver treats both of the phases like one phase in total. I'm guessing they do this to use much cheaper Mosfets and Inductors, but am at a loss why the doubler is really of use (I know in your PCB analysis of the 1080 FTW Hybird on Gamer's Nexus, you said the doubler could do more advanced things like use the phase with the least amount of current going through.) Still I'm not really solid on this, it would mean tons to me if you could respond, or maybe regard using a true doubler + drivers, or just driving two phases from one driver in one of your videos. Sorry in advance if you've already covered this!
Basically if you use a doubler you can do some very primitive current balancing. But you can also run a much higher clock into the doubler and then feed the MOSFETs half that to get very close to a real phases' accuracy. Taking the output of the driver and shoving it into 2 phases worth of components is a popular marketing stunt but does absolutely nothing for power quality. You just get a higher current through put and better thermals due to the extra surface area. However they could do the same if they increased the amount of FETs in each phase and just used 1 large choke.
Can you do one for GTX 970 STRIX? Please.... I need one really bad since der8auer haven't replied my message and there aren't any videos out there in regard to the breakdown.
Hi! Thanks for the info! I'm modding an EVGA GTX 960 FTW with the ACX cooler, though I swapped it out for a CPU AIO with an adapter that I modded to get good VRM cooling. I'm wondering how to tell what the new GPU Vcore is once doing the mod, though, since after I turned it up, HWinfo and GPU-z still report 1.275 V at full load. I measured that pin on the controller to ground and got 20k ohms (your video said you measured 2 Kohm on your card?) I added another 20k ohm varistor. In some positions it works and others the card is unstable. I guess I can do the math for a resistor in parallel to figure out what voltage the NCP* thinks it's getting on that sense pin, but not sure if it's linearly increasing output to the GPU in response to the artificially imposed voltage dip. I also noticed there's still a "Performance Limit: Power" despite that I soldered a wire between the two sides of the current shunt (I only found one, big resistor, by the +12V power connectors). Despite this, I'm free to keep cranking up clocks and don't see any actual throttling happening. Thanks for any tips!
excuse me if I bother you with the question but here it go: accidentally I sorted the fan pins of a gtx 970 and since then it doesn't work. where should I start seeking on the pcb? I did not spotted any burned.
This is amazing, thank you! Could I request you do the OcUK's GTX 970 "Reference Cooler Edition" please? If you need images, I would be willing to take apart and photo my own card if necessary!
The pictures look pretty legible to me, just about the same as the ones you usually see for modern cards www.ixbt.com/video2/images/x550/sapphire-x550-scan-front.jpg That said, the card is old school! I agree that it's probably not worth discussing, you can do your own analysis by looking up the datasheets of the chips on it. Most of the concepts are similar (look at the power ratings of the mosfets, see if you can find a datasheet for the controller, etc) Edit: ugh you can't read half the chips. You're right : too blurry
The scenarios for those questions are 4000, 5000 and 6000 Intel CPUs with high - but not extrem - OC (lets say at 1,35 - 1,45vcore with something like a nh D15 or a watercooler with a 3x120mm radiator): a) Does RAM with high clock speeds (respectively low timings) negatively impact CPU overclocking potential / require a higher vcore? If yes, how big of a deal is it? b) What should I look for when buying a mainboard too achieve a high CPU OC? When I bought my last mainboard some time ago, I used a search machine and filtered for Z97/ Z87, at least 8 CPU Phases, 8x/8x PCIe Multi-GPU support and then sorted by price. Skipped a few mainboards on that list, because the VRM radiators on the mainboard looked small and ended up with a ASUS Z97-A for ~120€. Is there a better way to shop for a mainboard? c) I run this mainboard with a i5 4690k @ 4,7ghz at 1,36vcore; can't get a stable 4,8ghz OC, even tried up to 1,425 volts. Lets say I had spend 250€ on a mainboard, would that have had any impact on my overclock? Does having more than 8 Phases help with overclocking a i5 or i7? How about i7 extreme Cpus? edit: I'm running 2 sticks of 4gb 1333mhz cl10 ram, which are overclocked to 1400 cl8 (still got those from back then when I bought my 1st pc, which had an i7 920)
There aren't any good pics. I do plan to do some Nvidia dualies though. I have a 590 to torture and I really want to get my hands on a 690 since I love dual core cards.
Is a Evga 960 ssc pretty similar on the pcb, im guessing not. i have an old one laying around that sat at 1575-1610 mhz so im really interested in bios modding and voltage mods. the 1070 i just got has me sitting 1st in my hardware configuration on the free firestrike fx-8320 and a single 1070, think it still says uknown gpu but its a g1 gaming and that has an average boost that stays above 2100 but tops at 2168 edit. but i fucking hate gpu boost 3
You ought to stack a couple of TECs to cool a low power chip sometime. I tested a single 60w TEC 40mm on a waterblock and it went all the way down to -25c on the cold side. That thing cooled d own by a 400w TEC could probably push a low power chip sub zero without much issues.
Actually Hardcore Overclocking take it easy there bud, ur triggering my jimmies over here. Also as a side note, PC desktop mining is profitable again. 2016 = return of desktop mining, even CPU mining.
I'm guessing the profitability only applies to ETH and only if you don't pay an insane power bill. Power in the UK isn't cheap. If you're worried about how long till I do stuff with TECs I'm aiming to do something with them within the next month or two and this won't change unless something dies on me.
Actually Hardcore Overclocking I'm making around 80-90 dollars a month whilst pulling 350w, I'd call that pretty decent. But I don't suggest people go out and buy hardware just to mine with, it's decent if you already got the hardware laying around. Also one thing when it comes to TECs, the ones you buy online from ebay and such seem pretty fishy, lower binned ones. I see huge differences in deltaT between the ones I got, some spots on the TEC can be defect and so on.
My "big" TEC project is going to be using 4 undervolted 24V TECs rated for 200W. At 12V I'm gonna get a good DT and a relatively low power draw along with a Qmax of 500-600W. That's going to be used for a water chiller with the goal to do 0C water under load with 200-300W heat loads. Ideal for freezing mid range GPUs and 4 core intel CPUs. I can always go out and get some of those cheap 12V TECs that do like 300W Qmax while pulling 400W but I want to do something a little more sophisticated first.
** QUESTION ** Im at 1520 mhz at msi gtx 960 ok ? if i go higher ex. 1530 - 40 and start a game I don't get artifacts for sometime 1-2 min and then im geting some artifacts and even crashes , im at 1.275 Vcore Air Cooling so... the question do you think this is from the VRMs that erm.. that the vrms geting warm or even hot ( i don't know what temp they are) that i get artifacts after a some time or is from the core becouse im at 60 *C at core.
"do your insane overclocking slightly more safely" yes prudence
Glad someone requested this particular card, it's the one I have :D
Great stuff! I learned to figure out power phases on PCBs from just looking at them with the help of your videos.
I have a dead 580 DirectCUII from ASUS lying around - I'll be taking it apart soon and can get you some pictures if you're interested.
Sure that would be cool.
Cool! I'll drop you a note with the links in the coming weeks.
Hey buildzoid, I just want to thank you so much for making these great videos, really interesting to watch, and really helpful as I'm trying to learn more about VRMs, and these videos are really helpful.
I was wondering if you maybe shed some light on a question that I have: I don't know if it is like this for gpus as much (still a lot more of your videos to watch), but I know for motherboards, on lower-mid end Z170 boards, they'll have a "8 or 10" phase VRM for the core voltage, which we know is really just a 4 phase channels from the pwm controller.
However, upon closer examination, each real phase from the pwm goes into a driver, but then is split between two different phases (inductors, high and low side mosfets.) I was reading Sin Hardware's VRM guide, and he briefly touched upon this, saying that splitting two phases from one driver was not the correct way to do it, and is doesn't cut the switching frequency in half like a doubler does.
Because of this, I'm assuming that without a doubler, you don't really have control over each independent phase, instead the driver treats both of the phases like one phase in total. I'm guessing they do this to use much cheaper Mosfets and Inductors, but am at a loss why the doubler is really of use (I know in your PCB analysis of the 1080 FTW Hybird on Gamer's Nexus, you said the doubler could do more advanced things like use the phase with the least amount of current going through.)
Still I'm not really solid on this, it would mean tons to me if you could respond, or maybe regard using a true doubler + drivers, or just driving two phases from one driver in one of your videos. Sorry in advance if you've already covered this!
Basically if you use a doubler you can do some very primitive current balancing. But you can also run a much higher clock into the doubler and then feed the MOSFETs half that to get very close to a real phases' accuracy. Taking the output of the driver and shoving it into 2 phases worth of components is a popular marketing stunt but does absolutely nothing for power quality. You just get a higher current through put and better thermals due to the extra surface area. However they could do the same if they increased the amount of FETs in each phase and just used 1 large choke.
Thank i'm learning about gpus problem to fix, and you cleared my mind about this gpu.
Thank you. This is a very helpful video.
Can you do one for GTX 970 STRIX? Please.... I need one really bad since der8auer haven't replied my message and there aren't any videos out there in regard to the breakdown.
Hi! Thanks for the info! I'm modding an EVGA GTX 960 FTW with the ACX cooler, though I swapped it out for a CPU AIO with an adapter that I modded to get good VRM cooling. I'm wondering how to tell what the new GPU Vcore is once doing the mod, though, since after I turned it up, HWinfo and GPU-z still report 1.275 V at full load. I measured that pin on the controller to ground and got 20k ohms (your video said you measured 2 Kohm on your card?) I added another 20k ohm varistor. In some positions it works and others the card is unstable. I guess I can do the math for a resistor in parallel to figure out what voltage the NCP* thinks it's getting on that sense pin, but not sure if it's linearly increasing output to the GPU in response to the artificially imposed voltage dip.
I also noticed there's still a "Performance Limit: Power" despite that I soldered a wire between the two sides of the current shunt (I only found one, big resistor, by the +12V power connectors). Despite this, I'm free to keep cranking up clocks and don't see any actual throttling happening. Thanks for any tips!
you do damn good work
excuse me if I bother you with the question but here it go: accidentally I sorted the fan pins of a gtx 970 and since then it doesn't work. where should I start seeking on the pcb? I did not spotted any burned.
This is amazing, thank you!
Could I request you do the OcUK's GTX 970 "Reference Cooler Edition" please?
If you need images, I would be willing to take apart and photo my own card if necessary!
Speaking of old cards, can you do a breakdown of an ATi x600* Hypermemory?
WAAAAY too old. Like so old IDK what it is
This:
www.google.com/url?sa=i&rct=j&q=&esrc=s&source=images&cd=&cad=rja&uact=8&ved=0ahUKEwiQi8jQzvTOAhVm5YMKHRwVBvMQjRwIBw&url=http%3A%2F%2Fixbtlabs.com%2Farticles2%2Fvideo%2Fx550.html&bvm=bv.131783435,d.cGc&psig=AFQjCNFbuDBA_uMMYhS5_ld21qWARizS_g&ust=1473040748671311
Yup way too old. Also those pics are pretty blurry. So I can't see anything too useful in them.
The pictures look pretty legible to me, just about the same as the ones you usually see for modern cards www.ixbt.com/video2/images/x550/sapphire-x550-scan-front.jpg
That said, the card is old school! I agree that it's probably not worth discussing, you can do your own analysis by looking up the datasheets of the chips on it. Most of the concepts are similar (look at the power ratings of the mosfets, see if you can find a datasheet for the controller, etc)
Edit: ugh you can't read half the chips. You're right : too blurry
Thanks!
Rx 470 red devil pls?
I'll do it if I find a good pic.
Like the ones here? videocardz.com/review/powercolor-radeon-rx-470-4gb-red-devil
Hey MAN.. can i have info?? what is actualy chip marked as U501 on back side?? Werry aprecipated :)
Do you have a recommendation for better cooling on the 8800GT from Nvidia? I have the single slot version and it runs hotter than I want it to.
My q15 burst where could I get a new one ?
Do gtx 960 g1 gaming pcb bd please.
Or which mosfet is at q15
can you also do the MSI 970 Gaming 4 g and there are 3 pcb versions?
Chao, The Roaring Lion
Hello.can u tell me where pcie sense pin is connected.my card have power from pcie but not detect pcie connected
I got shutdown and conncect pcie cable.
Please help me
Can u do 970 pcb
The scenarios for those questions are 4000, 5000 and 6000 Intel CPUs with high - but not extrem - OC (lets say at 1,35 - 1,45vcore with something like a nh D15 or a watercooler with a 3x120mm radiator):
a) Does RAM with high clock speeds (respectively low timings) negatively impact CPU overclocking potential / require a higher vcore? If yes, how big of a deal is it?
b) What should I look for when buying a mainboard too achieve a high CPU OC?
When I bought my last mainboard some time ago, I used a search machine and filtered for Z97/ Z87, at least 8 CPU Phases, 8x/8x PCIe Multi-GPU support and then sorted by price.
Skipped a few mainboards on that list, because the VRM radiators on the mainboard looked small and ended up with a ASUS Z97-A for ~120€.
Is there a better way to shop for a mainboard?
c) I run this mainboard with a i5 4690k @ 4,7ghz at 1,36vcore; can't get a stable 4,8ghz OC, even tried up to 1,425 volts.
Lets say I had spend 250€ on a mainboard, would that have had any impact on my overclock? Does having more than 8 Phases help with overclocking a i5 or i7? How about i7 extreme Cpus?
edit: I'm running 2 sticks of 4gb 1333mhz cl10 ram, which are overclocked to 1400 cl8 (still got those from back then when I bought my 1st pc, which had an i7 920)
have you tried HBM cards on LN2 ??/
Not yet but I plan to.
Dose that work on The gtx 950?
hey, can you do a review Nvidia Titan Z pcb? Please
There aren't any good pics. I do plan to do some Nvidia dualies though. I have a 590 to torture and I really want to get my hands on a 690 since I love dual core cards.
+Actually Hardcore Overclocking same here, and thx though.
Is a Evga 960 ssc pretty similar on the pcb, im guessing not. i have an old one laying around that sat at 1575-1610 mhz so im really interested in bios modding and voltage mods. the 1070 i just got has me sitting 1st in my hardware configuration on the free firestrike fx-8320 and a single 1070, think it still says uknown gpu but its a g1 gaming and that has an average boost that stays above 2100 but tops at 2168 edit. but i fucking hate gpu boost 3
hoping for zen lol
You ought to stack a couple of TECs to cool a low power chip sometime.
I tested a single 60w TEC 40mm on a waterblock and it went all the way down to -25c on the cold side.
That thing cooled d own by a 400w TEC could probably push a low power chip sub zero without much issues.
Oh I have plans for a lot TEC cooling experiments but I also have a lot of other stuff to get done before that.
Actually Hardcore Overclocking take it easy there bud, ur triggering my jimmies over here.
Also as a side note, PC desktop mining is profitable again.
2016 = return of desktop mining, even CPU mining.
I'm guessing the profitability only applies to ETH and only if you don't pay an insane power bill. Power in the UK isn't cheap. If you're worried about how long till I do stuff with TECs I'm aiming to do something with them within the next month or two and this won't change unless something dies on me.
Actually Hardcore Overclocking I'm making around 80-90 dollars a month whilst pulling 350w, I'd call that pretty decent.
But I don't suggest people go out and buy hardware just to mine with, it's decent if you already got the hardware laying around.
Also one thing when it comes to TECs, the ones you buy online from ebay and such seem pretty fishy, lower binned ones.
I see huge differences in deltaT between the ones I got, some spots on the TEC can be defect and so on.
My "big" TEC project is going to be using 4 undervolted 24V TECs rated for 200W. At 12V I'm gonna get a good DT and a relatively low power draw along with a Qmax of 500-600W. That's going to be used for a water chiller with the goal to do 0C water under load with 200-300W heat loads. Ideal for freezing mid range GPUs and 4 core intel CPUs. I can always go out and get some of those cheap 12V TECs that do like 300W Qmax while pulling 400W but I want to do something a little more sophisticated first.
Could you email me a that picture of the 960 Pcb especially the vrm
You might have it already after 6 months, but www.techpowerup.com/review/msi-gtx-960-gaming/4.html
Did you recive my message?
woops missed it somehow.
** QUESTION ** Im at 1520 mhz at msi gtx 960 ok ? if i go higher ex. 1530 - 40 and start a game I don't get artifacts for sometime 1-2 min and then im geting some artifacts and even crashes , im at 1.275 Vcore Air Cooling so... the question do you think this is from the VRMs that erm.. that the vrms geting warm or even hot ( i don't know what temp they are) that i get artifacts after a some time or is from the core becouse im at 60 *C at core.
i had an gtx 960 sc running at 1575 mhz with +100mv and 110% power usage. stable
1620 mhz on csgo but it was artifacting nightmare.
I have a msi gaming r9 290x 8gb
it has a accelero xtreme 3 cooler on it and the vrms go up to 110c. should I worried
* be worried
my sapphire 390x hits 70 on the vrm
there's 2 VRMs on the card. Which one is hitting 110C? because usally the minor VRM is the one that runs hot.
its the VRM 1. VRM 2 stays always at 55c even if the card is cold or Iam gaming for 4 hrs
if my memory serves me right VRM1 is the minor VRM.
I've already watched this about a dozen times so I've got this down but my PCB is different I think. I have a PNY XLR8 960
gtx 1060 attached mini please :) Want to get to know my card better
zotac* mini
What about some older Cards?? Like my 680´s or 6950´s...would be nice to know why these little Fellas wont Run quik =)
I've actually measured 140W @ 1.242 vcore
Can you do a 780 Ghz (from Gigabyte)? It's my card :)
heres a picture images.bit-tech.net/content_images/2014/03/gigabyte-geforce-gtx-780-ghz-edition-review/gig780ghz-8b.jpg
I can't read the numbers on the FETs nor the voltage controller in the photo. I'll try find some close ups of those.
+Actually Hardcore Overclocking I can open mine up and take some close up pictures
that would work great.
Here ya go imgur.com/a/fochR
why should someone raise Corevoltage on a 960? At roomtemp more volts wont help you anyway and on LN2---who wants to LN2 a GTX 960?
I do
hwbot points
Screw you! Hahahahha nice man. I also need this info. Got this card and gotta fix it.
That huge empty space on the pcb just to make it look bigger is a joke.