• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

New 3080 and suddenly problems

It's not a problem, separate cables mean it couldn't possibly be a problem. This myth about multi rail CPUs being a problem goes back over ten years when multi rail PSUs were new technology and some manufacturers designed them so poorly that there were issues as it was easy to overload a rail (single rail is now used as clever marketing by manufacturers as they know the early issues with multi rail are burned on enthusiasts' brains; of course single rail is easier and cheaper for them to design and manufacture while offering fewer protections for the user, so it suits them down to the ground). This is not an issue since over ten years ago and certainly not from reputable brands.
Only myth is PSUs with single transformer/current source having multiple rails, when marketing says so.
Actually having multiple 12V sources/transformers would be extra cost in components besides making PSUs bigger.

While this schematics is very antiquated in some details, overall structure and functionality is fully valid for basically every PSU:
https://www.smpspowersupply.com/600W_power_supply.jpg
Controller simply monitors 3mV per A voltage drop in R61 and R62 and if voltage drop rises over set limit, shuts down whole PSU.
Only load separation is what length of wires and those current monitoring resistors achieve.
Haven't yet seen PSU with even single filtering capacitor after those resistors.
(and anyway having extra cap in cable near load would be more effective)


Whole mess was cooked up to cover Intel's doings.
Intel's designers referenced earlier standards when making ATX specification.
And one of those standards happened to say you can only output max 240VA (20A @12V) from any single wire/lead.
With that becoming fast inadequate designers invented adding second group of 12V wires with own current limit counted separately.

But instead of PSU makers honestly saying that user has to balance load between leads connected to different limiters, marketing invented BS about multiple rails.


And those earlier standards actually covered also power supplies with higher output power.
For example UL classifies those as Level 6 devices.
That UL Level 6 marking used to be in sticker of many PSUs decade ago.

While having only single high 12V current limit can certainly be considered less safe for HW failure situations/short circuits, safety difference is clearly marginal in normal size PSUs.
After all most quality PSUs used heavily by enthusiasts have been such for dozen+ years.


Though for ~1kW and above PSUs I would certainly consider some current limits sensible.
If max current is 100A, it will get hard to trigger protections by something shorting in end of single 12V wire cable:
Already 0,12 ohm loop resistance would limit current to that preventing PSU from "seeing" short circuit.

But with very high 500+W transients of power hungry cards like RTX 3080, they would certainly need to be fed from multiple current limited circuits, or "fuse" would have be slow, to avoid triggering PSU shutting down.
 
HWInfo now has a handy voltage readout for each GPU power connector and the PCIE slot. Can you post a screenshot at idle and another one after the PC has been under gaming/graphics load e.g. 3DMark?

Here is a reading of the GPU section after PC has been booted up and is sitting idle.

Capture.png

I will post another in a bit after I game for a bit on completely stock settings.
 
I had to reduce my ram voltage as I experienced crashes after upgrade to a 3080 and the issue was my ram was overheating in games and causing errors and instability due to the heat output of the card, this was undetectable in memtest as the GPU isn't loaded so the ram stayed cooler.
 
Thanks, at idle your voltages look good from what I can see.

With every back to stock I just gamed for 3hrs with no crashes.

GPU felt sluggish though in Metro Exodus, but ram is down at 2133MHz currently so it could be down to that.

I need a few more game sessions before I can say the PC is stable and it is just the ram not being stable.

Below is a screenshot having just excited my gaming session.

Untitled.png

So tell me then if it is temperature related, how are people with the 8 Pack CL14 3600MHz ram going to fair?

Surely at 1.45v out of the box it is going to be unstable.

At the moment temperature conditions could not be better in my PC. It's winter, cold ambient temps and I have a case with good ariflow and I am having temperature related issues...

Leaves me a bit in no mans land if I choose to change my ram sticks as if I bought say the 8 pack CL14 3600MHz stuff, how do I know it wouldn't exhibit the same temperature issues?
 
I had to reduce my ram voltage as I experienced crashes after upgrade to a 3080 and the issue was my ram was overheating in games and causing errors and instability due to the heat output of the card, this was undetectable in memtest as the GPU isn't loaded so the ram stayed cooler.

Good to know.

Seems to be consistent with what I am experiencing.

Which 3080 did you get?
 
With every back to stock I just gamed for 3hrs with no crashes.

GPU felt sluggish though in Metro Exodus, but ram is down at 2133MHz currently so it could be down to that.

I need a few more game sessions before I can say the PC is stable and it is just the ram not being stable.

Below is a screenshot having just excited my gaming session.

Untitled.png

So tell me then if it is temperature related, how are people with the 8 Pack CL14 3600MHz ram going to fair?

Surely at 1.45v out of the box it is going to be unstable.

At the moment temperature conditions could not be better in my PC. It's winter, cold ambient temps and I have a case with good ariflow and I am having temperature related issues...

Leaves me a bit in no mans land if I choose to change my ram sticks as if I bought say the 8 pack CL14 3600MHz stuff, how do I know it wouldn't exhibit the same temperature issues?
Voltages still look fine, but I meant to leave HWInfo open while the game was running so we could see if there were any voltage drops under load.

With your RAM is there an XMP profile at 1.35V or lower and 3200MHz or something like that? You could try reducing RAM to 3200MHz, also try reducing the Infinity Fabric speed independently of the RAM and see if that's the problem. Samsung B-Die is notoriously sensitive to temperature, Micron E-Die like the Crucial Ballistix series (check whether all Crucial Ballistix is E-Die before buying) works well with Ryzen and is much less sensitive to temperature. Of course you won't be getting tight timings out if it, more like 3600MHz CL16 and anything above 3200MHz is technically overclocking and dependent on your CPU's memory controller and Infinity Fabric anyway.

I wonder whether the RTX 3080 using PCIE 4.0 on the CPU has pushed it over the edge and that's why the RAM is no longer stable at 3600MHz, otherwise it is a temperature issue.
 
Last edited:
Good to know.

Seems to be consistent with what I am experiencing.

Which 3080 did you get?
I'm using the ventus, ram was set to 3800/14 with 1.49v and was fine for a year when I was using a 1070ti soon as I switched to the 3080 I started getting random CTD which after some trial and error with different settings I pinned down to the ram so relaxed the timings to 16 and voltage to 1.44 and its been fine since.
 
Very strange all in all though as all the reviewers of these cards all tested if the new airflow design of the FE cards would impact on ram/cpu temperates and they all said there was nothing to fear.

Our experiences seem to be the opposite.
 
Voltages still look fine, but I meant to leave HWInfo open while the game was running so we could see if there were any voltage drops under load.

With your RAM is there an XMP profile at 1.35V or lower and 3200MHz or something like that? You could try reducing RAM to 3200MHz, also try reducing the Infinity Fabric speed independently of the RAM and see if that's the problem. Samsung B-Die is notoriously sensitive to temperature, Micron E-Die like the Crucial Ballistix series (check whether all Crucial Ballistix is E-Die before buying) works well with Ryzen and is much less sensitive to temperature. Of course you won't be getting tight timings out if it, more like 3600MHz CL16 and anything above 3200MHz is technically overclocking and dependent on your CPU's memory controller and Infinity Fabric anyway.

I wonder whether the RTX 3080 using PCIE 4.0 on the CPU has pushed it over the edge and that's why the RAM is no longer stable at 3600MHz, otherwise it is a temperature issue.

I do have a DOCP 3200MHz profile on the ram which is at 1.35v but I need to run it at a minimum of 3600MHz as for me this is the minimum acceptable speed on Ryzen 3000 and up.

Luckily my ram is officially specced for 3600MHz CL16 as if says so on the team group website. The only issue is I have no DOCP profile for it.
 
Very strange all in all though as all the reviewers of these cards all tested if the new airflow design of the FE cards would impact on ram/cpu temperates and they all said there was nothing to fear.

Our experiences seem to be the opposite.
Maybe it's worth rigging up a fan directly over the ram or getting one of those ram coolers.

I do have a DOCP 3200MHz profile on the ram which is at 1.35v but I need to run it at a minimum of 3600MHz as for me this is the minimum acceptable speed on Ryzen 3000 and up.

Luckily my ram is officially specced for 3600MHz CL16 as if says so on the team group website. The only issue is I have no DOCP profile for it.

If you can run 3600/16~3800/16 than you should be able to relax the voltage quite a bit, I found going from 16 to 14 needed quite a bit more voltage.
 
Got my PSU in multi rail. As has been said lots of historic bad feeling toward multirail that is now misunderstood.

Both those games are not shining examples of stability mind?
 
OK, so it seems slackening the timings on my ram has done the trick.

Yesterday I ran the system with 2133MHz ram and today I luckily found one of my old CL16 3600MHz profiles at 1.35v and I have just had an extended gaming session with no issues.

So thank goodness it wasn't an issue with my PSU or GPU.

But does beg the question why the 3080 has affected my ram overclock.

Mostly heat or perhaps the ram overclock I had was never really stable (despite working for a year)

But all the reviewers of these cards addressed this concern and said there was nothing to fear and the cards where actually very efficient.

Makes me wonder as more of these cards get in to peoples hands if the faster ram kits at 1.45v will start to exhibit problems as well due to overheating.

This is also in a chilly room, mid winter in a very well ventilated case.

So best case scenario!
 
"I do have PBO turned on though using motherboard limits, scalar 2x and boost 500Mhz"

that boost sounds very high, my advice is put PBO=Enabled but dont mess with Advanced settings like the curve, scalar and boost if you are having stability issues
 
"I do have PBO turned on though using motherboard limits, scalar 2x and boost 500Mhz"

that boost sounds very high, my advice is put PBO=Enabled but dont mess with Advanced settings like the curve, scalar and boost if you are having stability issues

Yes it's high. But the cpu will never boost that high before it hits one of the pbo limits.

It's not the CPU that's the problem it was my ram overclock. I've toned it down to cl16 and so far it's all good.

But it makes me wonder why it worked before and now after installing 3080 it doesn't.

Heat?

But if that's the case what of ram sticks like the 8 Pack cl14 3600 ram that is 1.45v out of the box?

They gonna overheat too? :confused:
 
Sounds like its sorted. I will mention that some games are fussy about Windows page file size, I always used to set a fixed page file size but some games really don't like that, Hitman I think is one, so now I've just set that to default/auto and no more crashes.

Your RAM issue may be that although it was stable before the 3080 it was right on the edge of stable and it just took something to push it into unstable.

Sorry I missed, what case and cooling do you have? This is also a factor when it comes to enough cooling inside the case. Make sure the fans are ramping up enough when gaming.
 
In the past I had problems with a GTX 1080 black screening and goodness knows what but turned out to be driver issues I've had zero issues with my 3090 FE which I'm pleased about, your issues especially blue screening point towards a hardware issue to me

I agree the most likely culprit here is the ram.

It's at 1.47v in bios (more like 1.48-1.5v observed voltage. It oscillates) plus heat from 3080 could be pushing it over the edge.
But I thought it was well tested by reviewers that the new FE design wasn't negatively affecting CPU / RAM temperatures.

I don't know about reviews but the heat from my FE card was definitely affecting CPU temperature maybe the 3080 is different but theres not a lot of difference 320w vs 350w iirc I had to fit an AIB to suck in some cool air from outside the case to keep the cpu under control and I've actually undervolted mine because the sheer amount of heat thrown out by the card is frankly ridiculous

I had to reduce my ram voltage as I experienced crashes after upgrade to a 3080 and the issue was my ram was overheating in games and causing errors and instability due to the heat output of the card, this was undetectable in memtest as the GPU isn't loaded so the ram stayed cooler.

If RAM overheats it will throw out errors that was established in one or more of the overclocking threads in the RAM forum it won't necessarily harm the ram itself but it can cause significant errors and if necessary fit a fan to cool it
 
Back
Top Bottom