Question Got a secondhand 1650 Super and now I regret it

Page 2 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Jan 14, 2023
23
0
10
Bought a secondhand MSI GTX 1650 Super VENTUS XS OC to replace my Powercolor Radeon R7 360 with it. The card itself looks like new, the seal is not broken, no visual defects. I've put it in my PC(had to buy a DVI-I-to-VGA adapter), started it. It was fine until I started FurMark. It couldn't last more than 30 seconds(it reboots PC with no BSOD, Kernel Power 41 (63) and BugCheckCode 0). Okay, I thought it must be the PSU(Aerocool VP-550(not a 80+ Bronze model) was inside at this moment). Changed it to MONTECH GAMMA II GOLD, which is also 550W and has 80+ Gold certificate. Everything stayed the same, but! The card could stay 1 minute and 30 seconds and then restart my PC again. Tried delete drivers with DDU and install them through GeForce Experience. No changes. Tried to turn on FurMark with each of two RAM sticks by changing them. Still no changes. Tried to underclock it. Made the core clock around 1300Mhz. Didn't work. Tried to make the power limit lower. I set it on 70% and voila, now it seems can work in FurMark with no troubles, but it's core clock is almost on the same core clock which R7 360 has(1100Mhz).
My config:
CPU: AMD FX-6300(Turbo Boost is turned off)
Motherboard: Asus M5A78L-M LX3
RAM: Kllisre KRE-D3U1600M/8G(DDR3-1600 8GB) + Kingston 99U5471-058.A00LF(DDR3-1600 8GB)
Two fans on the side, one is on the back, 2 HDDs

With R7 360 everything works pretty fine
I'm very hoping that it's not GPU's problem but motherboard's one

Edit: I played Apex Legends with it with textures on Medium. PC rebooted in 10 minutes likely. Final Fantasy 7 Remake didn't rebooted my PC, I went through the whole first chapter(I forgot to check what graphic settings were)
 
Last edited:
Jan 14, 2023
23
0
10
Yep, this one
That's what I have made(addtional small radiator on the top to hold it because the card is sticked to the motherboard upside down)
mPlFJ7X.jpg
 
Jan 14, 2023
23
0
10
Switching from AMD to Nvidia may require a run of DDU.
Download and follow instructions.
https://www.guru3d.com/files-details/display-driver-uninstaller-download.html
Install new drivers as an administrator.

Then search for your "Shader Cache" folder and delete its contents.
Retry your games. First load will rebuild the shader cache so be patient.
I was using DDU and the drivers themselves are asking for adminstrator rights, so I was deleting with using DDU and was installing as an admin. Do you really think this is the Shader Cache trouble if it PC can't last fine in Superposition on 1080p High preset(the first time after applying a custom radiator was fine with no reboots, I repeat!!)
 
Long story short version.
In games and benchmarks ( gaming) the CPU must calculate the size, shape, place on map, orientation etc for everything you see in every frame.
Then the video card takes the points and creates a "Wire Frame" and does its magic.
Colors,textures,skins,shading,lighting, etc..... all visual effects are then placed on the "Wire Frame" , post processing effects are added and sent to the output frame buffer and sent to the monitor.
To see this , run heaven in demo mode.
At the top click camera. Select Wire Frame.
This map is what the CPU must create for each frame.
Adding cooling to the video card VRMs probably reduced their waste heat by 5 to 10 watts if you are running high clock speeds.
The cooler they run the more efficient they run.
But a faster video card gives the CPU more to do . More frames to generate, more workload on CPU.
 
And when cpu runs on high loads, this stresses vrms on motherboard.
AM3 platform is notoriously famous for vrm overheating causing crashes with AMD FX line of cpus (95W and higher TDP).

VRMs on your motherboard have no heatsinks at all. This is obvious source of problems.
 
Jan 14, 2023
23
0
10
Yahoo!!! I'm back with big updates for ya. The problem isn't the GPU itself, because I've tested in another machine with Intel CPU. And it works fine. 10 minutes in FurMark - passed. 1080p Extreme in Superposition - passed too! So.... what's next, where are we looking now?
 
Last edited:

Karadjgne

Titan
Ambassador
Montech Gamma II is a CWT GPS series. Same as Thermaltake Smart and a few others. Basically it's performance is worse than a Corsair CX, but a bit better than Evga GQ. It's an OK Budget psu, that shouldn't have issues with a gtx1660 or below. It's an improvement over the Aerocool, but then again, every even halfway decent psu, is.

Unless ypu have seriously crap-tastic cpu cooler, there should be no reason to disable Turbo, that'll do more for single thread performance with the 4.1GHz possible boost, and not use any more voltage than currently used. Only reason to kill Turbo is if running a dedicated OC.

Speaking of cpu cooler, with that Asus board, a Good downdraft cooler will do a lot more than a halfway decent tower. A Good downdraft cooler provides direct airflow to the motherboard, including the VRM's. The lack of decent heatsinking in that area isn't necessarily a bad thing, but can be a limitation at higher power use.

Honestly I'd look into overclocking that cpu, at least to turbo limits, and get that vcore/vid down. Do Not rely on standard software temp readings, they are not anything close to accurate nor applicable. You will and should be using AMD Overdrive, period. It has a learning curve, but basically responds in the opposite to standard Intel way of thinking. 0 (zero) = thermal limit. Any number bigger is cooler running, you should idle closer to 40, game closer to 20ish. Only time you'll ever need worry is if the thermal margin is in the 10-0 range, but even then, you are still ok.
 
Jan 14, 2023
23
0
10
Unless ypu have seriously crap-tastic cpu cooler, there should be no reason to disable Turbo, that'll do more for single thread performance with the 4.1GHz possible boost, and not use any more voltage than currently used. Only reason to kill Turbo is if running a dedicated OC.
I use a tower cooler from Aliexpress with two coolers on it. It should blow 100-110W away. Maximum FX-6300's TDP is 95W as far as I can remember. But with Turbo Boost turned on (while I was using my R7 360) it was turning off my PC so I have decided to turn it off and I could use my FX-6300 + R7 360 nicely
 

Karadjgne

Titan
Ambassador
The fx 6300 is a Thermal Design Power cpu of 95w. Realistically ypu should be running a cooler capable of 1.5-2x TDP. That means 150w-200w range coolers. When a cooler is closer to a 1:1 ratio, 100w = 100°C roughly. Thermal limit on Any FX cpu is 62°C as measured by AMD engineers.
 
Jan 14, 2023
23
0
10
The fx 6300 is a Thermal Design Power cpu of 95w. Realistically ypu should be running a cooler capable of 1.5-2x TDP. That means 150w-200w range coolers. When a cooler is closer to a 1:1 ratio, 100w = 100°C roughly. Thermal limit on Any FX cpu is 62°C as measured by AMD engineers.
It means the temp sensors are lying?
 

Karadjgne

Titan
Ambassador
It means the temp sensors are lying?
Yes. Totally. For the cpu. Inside a cpu, on the cores of Every Intel is a thermal diode, there's also more spread around the entire cpu, 26 in all I think on some models. The FX has 0. No thermal diodes at all. Cpu temp is totally false. I could not even imagine Why Amd did that, but they had to get engineers and modify engineering samples with strips, just to get the FX thermal limits of 62°C.

So as it stands, only Amd Overdrive and CoreTemp (gotta fix the settings in it) can 'read' an FX cpu.

They don't actually read temps. There's a complex algorithm that consists of load per core, voltage per core, combined voltages, socket temps and other variables that when calculated out come up with a number. That number is the Thermal Margin.

So take a game like CSGO on a 3570k. It's using only 2 cores, but both cores are about maxed. Cpu temp might register as 80°C, but realistically that's only 2 cores at 80°C, the remaining 2 cores are closer to 40°C. With an FX, because of the load calculations, you might get a TM of 20. Or, you could run a game that uses all 6 cores, with only 1 heavy, the other 5 about half way used, and still get a TM of 20, whereas the Intel hottest core would be closer to 60°C, the other 3 at 50.

So what it's telling you is roughly how much thermal headroom you have left. Think of it like the gas gauge in a car, tells you how much fuel you have left, but for an intel it'd say how many gallons have been used and it's then on you to figure out how much gas is left.

But understand this, the actual number is not important. What's important is what that number represents. There's zero difference between 15 and 25, or 0 and 5 etc. The numbers will change. It's what the numbers stand for that means everything. You see anything close to 40, you are cold, ton of headroom. 30's, same thing. 20's is working as it should, no worries. Teens is warm, but still no cause for alarm. Single digits, now you're running hot, keep an eye on things. Negative numbers, you are cooking, what's the problem with the cooler fan, time to freak out.
 
  • Like
Reactions: Tac 25

Karadjgne

Titan
Ambassador
Reboots are difficult without knowing exactly what happens. Many have the bios set to reboot automatically on power loss, there's also the Oops error, we are sending info to Microsoft and then reboots, there's also craptastic psus that trip protections but reboot anyway, which they should not etc.

Simply stating there's a reboot doesn't say much, is it a reboot with no bluescreen or blackscreen, sudden reboot with no apparent reason, Oops, etc. And is it a reboot or reset?

VRM's are a possibility. That board has minimal phases, 3+1, on a 760G chipset. With more modern boards, if a phase drops for any reason, there's usually space in the 8,10,12+ phases where the power isn't interrupted, but in a 3+1, you drop a phase for any reason, or even limit a phase, and there isn't enough power to support the cpu.

It could also be a bad gpu socket, a pin out of alignment, bad or busted solder joint and when pushing hard power limits with furmark, it's getting warm enough to short, which goes right back to the cpu (eventually) on the pcie buss, and the cpu says Nope! And shuts down. Which allows the solder to cool, and resume normal, non-furmark, functionality.

Could also be a failing capacitor.

It's a very old board, and was on the lower end of budget to begin with, and there's a dozen or more possibilities, all of which point to the motherboard in some way.