2 Ram module makes PC restart

[Harlow]

I've just taken a look at your Patriot PSD48G266681 memory on the Amazon web site and the picture shows what I consider to be an unusal feature (i.e. I've not seen it before). If you look closely at the lower edge of the DIMM, you'll see the length of the gold plated contacts is greater in the middle and shorter at either end. It looks like Patriot have chamfered (ground down) the left and right-hand ends of the DIMM, presumably to make it easier to plug into the motherboard. My guess is you need less pressure initially during fitting, when only the middle section of the DIMM meets the socket contacts. This probably helps to get the DIMM seated, before the final hard push to get the latches engaged.

If the outline of your memory is identical and the metal contacts in your motherboard's DIMM sockets are particularly low down, there's the vague chance you're not getting good contact on all the gold DIMM contacts. It's a wild hypothesis, but I did have a similar problem with an old plug-in VGA graphics card. In this instance the opposite was true. There was too much blank fibreglass below the bottom of the gold plated contacts, so the GPU wasn't recognised by the motherboard. I had to file 2mm of blank fibreglass off the bottom of the card, so it seated lower in the socket and made electrical contact with the motherboard. If you do have this specific problem, the only solution is new RAM with no chamfer. I'd advise against filing down the centre of your DIMMs.

 

[amrgul1]

I also forgot to mention that sometimes (rarely) I notice a lag with mouse movement. The guy in the service said that this also is the issue of the ram.

chance you're not getting good contact on all the gold DIMM contacts

Uhm... if they don't get all contacts would it still work?!

 

[VuurVOS]

It seems you have Nayna NT5AD1024M8A3-HR:

https://club.dns-shop.ru/review/t-103-operativnaya-pamyat/56393-obzor-operativnoi-pamyati-patriot-signature-line-psd48g266681-8/

Can you try this configuration?

 

[amrgul1]

It seems you have Nayna NT5AD1024M8A3-HR:
https://club.dns-shop.ru/review/t-103-operativnaya-pamyat/56393-obzor-operativnoi-pamyati-patriot-signature-line-psd48g266681-8/
Can you try this configuration?

Tried but things got even worse - I got restarts and also freezes soon after windows loads. Reverted bios to defaults.

 

[Harlow]

We know the Patriot PSD48G266681 DIMMs are working most of the time when both modules are installed, but result in the occasional crash. As you surmise, all connections between the DIMMs and the sockets seem to be working, otherwise the computer wouldn't start. My concern is the shorter gold-plated PCB contacts on each end of the DIMMs may result in higher resistance, if the contact area on some connections is sub-optimal. As I said earlier, this is just a guess on my part. I don't have any facts to back it up.

Good news. The replacement RAM fitted by the IT guy worked fine, so you know the mobo, CPU, GPU and SSD are functional. The fault has now been isolated to the Patriot RAM, regardless of the fact it passes MemTest.

Did the computer work perfectly with the Patriot RAM in the past or have you always had BSODs with this setup? If the problems only started recently, the RAM has deteriorated to the point where it causes instability. Changing RAM Voltages or timings might effect a short term fix, but probably not a permanent cure.

The problem might also be due to degradation in the CPU's memory controller due to prolonged overclocking or just general use. I pushed an Athlon X4 620 hard with a high overclock for 6 years, but eventually I had to knock it back down to stock speed. My guess is electro migration set in. A price you have to pay for all that extra speed and heat.

I know you'd like to fix the problem for free just by tweaking the BIOS, but fitting two new DIMMs is a simple (more expensive) solution. Alternatively, ditch one DIMM and carry on with half the RAM.

It all depends on how patient you are. Welcome to the world of computer repairs. Best of luck.

 

[amrgul1]

Did the computer work perfectly with the Patriot RAM in the past or have you always had BSODs with this setup

These Rams I purchased not long time ago. They seem to be new shielded with paper on its box. I don't have problem to buy new Rams but the problem is, looking at my local stores, I can't find with exact part numbers which are 100% matched to my mobo's QVL list, well I didn't look for all firms out there but Corsair and Gskill does not match.
I also tried The King's solution and increased Ram voltage (Vdimm) to 1.25 and it worked partially - enlarging video file sizes didn't give restarts anymore but for only 6-7 tries, after that it still restarted, so, it seems it increased stability but not for 100%. I just can't find the other setting - VSOC in bios yet to try.

 

[VuurVOS]

Tried but things got even worse - I got restarts and also freezes soon after windows loads. Reverted bios to defaults.

Can you try with only one module installed in DIMM_A1. If you still have issues then put the other module in DIMM_A1. It could be that one of those modules is faulty.

 

[amrgul1]

Can you try with only one module installed in DIMM_A1. If you still have issues then put the other module in DIMM_A1. It could be that one of those modules is faulty.

Only one ram module works well, meaning it does not restart PC and does not matter which one it will be - both separately work without restarts.

 

[BetrayerX]

Try a second memory kit. I got an Asus B450 mobo atm that REFUSES to boot with memory on dual channel. Tried 3 different kits that all work on my x570 MSI board:
Team Force Xtreem 3600 32GB
GSkill RipJaws 3600 32GB
Team Group 3600 16GB

All do the EXACT same behavior. 1 memory alone work fine in any slot. Any will do. XMP works for 1.
2 MIGHT work if you put them in a single channel configuration, sometimes wont boot and no XMP, just basic speeds.
2 memories on dual channel, not even a boot check, just a yellow light on the mobo and that's it.

Went to RMA and the a$$h0le in turn actually scoffed the situation and denied the claim in quite a very condescending manner, saying that the memory kits are not approved for that specific mobo. I hate ASUS support, it's absolute garbage and this is not even close to other experiences I've had with them.

 

[amrgul1]

2 memories on dual channel, not even a boot check, just a yellow light on the mobo and that's it

It seems your case is even worse than mine. It's strange so much different rams not compatible to Asus motherboard.

 

[VuurVOS]

@amrgul1
Can you share a screenshot of ZenTimings when only one module is installed. I am curious if it is different configured when only one module is installed.

 

[amrgul1]

@amrgul1
Can you share a screenshot of ZenTimings when only one module is installed. I am curious if it is different configured when only one module is installed.

There one guy on the Techpowerup had simillar problem with restarts and he solved that by changing PCIE 16 from auto to Gen3. I also found it in bios and changed but for me this didn't solve restarts. You can see that article here and final post from OP :

PC reboots while playing games in dual channel config.

I think im fine with this don't want more problems :D You can try, just for the fun of it. I understand you had a share of issues to the point you wanted "just to work" in any possible way... I know tthe feeling! Well keep it as is for a while, enjoy it, make sure its stable and then you can...

www.techpowerup.com

 

[Lew Zealand]

Try a second memory kit. I got an Asus B450 mobo atm that REFUSES to boot with memory on dual channel. Tried 3 different kits that all work on my x570 MSI board:
Team Force Xtreem 3600 32GB
GSkill RipJaws 3600 32GB
Team Group 3600 16GB

All do the EXACT same behavior. 1 memory alone work fine in any slot. Any will do. XMP works for 1.
2 MIGHT work if you put them in a single channel configuration, sometimes wont boot and no XMP, just basic speeds.
2 memories on dual channel, not even a boot check, just a yellow light on the mobo and that's it.

Went to RMA and the a$$h0le in turn actually scoffed the situation and denied the claim in quite a very condescending manner, saying that the memory kits are not approved for that specific mobo. I hate ASUS support, it's absolute garbage and this is not even close to other experiences I've had with them.

I have an Asrock B450 board and long story short, it would only boot at 3200 MHz when I bought RAM that's qualified for the Mobo at Asrock's website (Corsair was cheapest qualified, so bought that), even though I already had other 3200MHz G.Skill that works fine in a couple of MSI Mobos. This is the first time I've encountered this incompatibility on a Mobo, highly annoying but I'm reusing that other 3200 in a B465 Asrock board that seems happy with it. I pay closer attention to these Mobo RAM qualifications now...

 

[Lost_Troll]

The issue is that your DDR modules are probably at the CPU's dram controllers max on timings and when in dual channel mode and it is causing the controller to have fit. The one thing that points to this, is when the tech ran the slower DDR 2133 and it ran without rebooting.

I would try to set your DDR speed to 2400 (1200 MCLK) (Only change the DDR speed and nothing else) and see if it has the same reboot issue. If it does not, then you know that the DDR timings for the speed are too tight for the CPU dram controller to handle.

If it still continues, then I would look for a new RAM set. As for a replacement, you will have a hard time finding a set that is on the list since most manufactures update the ram chips with faster ones, which will always change the part number of the ram module or set. So, you may want to pick a manufacture and email their support and see what part number they recommend for your system. By doing this, it will make it easier to RMA if they too have the issue.

 

[amrgul1]

The one thing that points to this, is when the tech ran the slower DDR 2133 and it ran without rebooting

Indeed his Gskill Trident was 2133MHz (though I don't remember mentioning it).

I would try to set your DDR speed to 2400 (1200 MCLK) (Only change the DDR speed and nothing else) and see if it has the same reboot issue. If it does not, then you know that the DDR timings for the speed are too tight for the CPU dram controller to handle.

Already tried with 2400 and 2133MHz but still restarts. With so much restartings now, Win10 boot got damaged, have to repair it. I btw dual booting with Seven on mbr disks.

 

[BetrayerX]

I pay closer attention to these Mobo RAM qualifications now...

Oh, as I said, got a few more ASUS stories. Once installed an ASUS b550 (one of the cheapo ones) and the client bought certified DDR 4 3200 memory for it. Machine was unstable, got a few blue screens and even windows failed to install in one sit. I got the memory from one of my machines (Teamgroup 16gb ddr4 3200) and it ran without any issues with XMP on. Machine runs like a charm with them. The punchline is that I installed the "failed" memory on my machine, ran with XMP, and no issues whatsoever.

Saying again, the ASUS board did not work properly with the certified memory and ramdon memory kit ran fine with it, while further test proved that the original memory kit ran fine somewhere else.

 

[Wirko]

I've just taken a look at your Patriot PSD48G266681 memory on the Amazon web site and the picture shows what I consider to be an unusal feature (i.e. I've not seen it before). If you look closely at the lower edge of the DIMM, you'll see the length of the gold plated contacts is greater in the middle and shorter at either end. It looks like Patriot have chamfered (ground down) the left and right-hand ends of the DIMM, presumably to make it easier to plug into the motherboard. My guess is you need less pressure initially during fitting, when only the middle section of the DIMM meets the socket contacts. This probably helps to get the DIMM seated, before the final hard push to get the latches engaged.

If the outline of your memory is identical and the metal contacts in your motherboard's DIMM sockets are particularly low down, there's the vague chance you're not getting good contact on all the gold DIMM contacts. It's a wild hypothesis, but I did have a similar problem with an old plug-in VGA graphics card. In this instance the opposite was true. There was too much blank fibreglass below the bottom of the gold plated contacts, so the GPU wasn't recognised by the motherboard. I had to file 2mm of blank fibreglass off the bottom of the card, so it seated lower in the socket and made electrical contact with the motherboard. If you do have this specific problem, the only solution is new RAM with no chamfer. I'd advise against filing down the centre of your DIMMs.

View attachment 285417

That shape of the DIMM is standard in DDR4 (and DDR5 too).

DDR4 differences and advantages over previous generation modules

Physically, a DDR4 module, looks very similar to a DDR3. However, DDR4 has 288 pins compared with DDR3’s 240 pins; DDR4 SO-DIMMS have 260 pins instead of 204 in DDR3

www.atpinc.com

 

[Lost_Troll]

Already tried with 2400 and 2133MHz but still restarts. With so much restartings now, Win10 boot got damaged, have to repair it. I btw dual booting with Seven on mbr disks.

I would just look for another ram kit and sell the set you have now. It sucks that you have to fix your win 10 installation and hopefully you do not have to do again.

 

[Harlow]

Many thanks to Wirko. I didn't realise DDR4 and DDR5 DIMMs had a different profile. My only excuse is the systems I repair use old-fashioned DDR3. I have only two machines with DDR4 and one with DDR5 and I haven't disturbed the DIMMs since they were fitted. I've just pulled a DDR4 DIMM out of my Gigabyte rig and the bottom edge is "chamfered". You learn something new every day.

Also thanks to BetrayerX for info about Asus RAM compatibility issues. I had occasional BSODs on my new rig with an Asus Prime X670-P WiFi mobo + 7950X and 2 sticks of Kingston Fury KF548C38-32 RAM. Kingston was the cheapest dual-DIMM 64GB DDR5 kit I could find back in December 2022. It runs at stock 4800MHz and there are no faster XMP settings. I wasn't prepared to shell out more cash for a faster 6000MHz kit. The BSODs only happened with one particular program, but the 7950X has been running crash-free, flat out, 24/7 processing videos after updating the offending app. The computer room is now the warmest place in the house!

My other Asus rig has been stable for nearly 4 years with a ROG Strix X570-F Gaming mobo + 3800X and 4 sticks of Corsair LPX CMK32GX4M2D300016. Despite running four DDR4 DIMMs, the system is happy running the tightest 16-20-20-38 1500MHz XMP setting, i.e. memory @ 3000MHz. I was more cautious with the 7950X and bought two 32GB DDR5 sticks instead of four 16GB DDR5 sticks which were cheaper. No point in loading the memory bus unnecessarily. It also means I can expand to 128GB RAM if my video editor's demands increase.

As for the OP's trouble with RAM errors corrupting Windows 10, I'd be inclined to use Macrium Reflect Free (or similar) to create an image of the bootable drive. Then, if Windows 10 dies again, boot the PC from another drive and restore the Windows 10 image to the main drive. It saves at least an hour reinstalling Windows and all essential programs.

As several people have already mentioned, don't get too stressed trying to find RAM on the QVL. These lists are rarely kept up-to-date for boards more than one year old. The cheapest option for new RAM is probably mail order. If you buy from a reputable online company, you should be able to RMA items that don't work, although it doesn't hurt to check their 'Terms and Conditions' first. If new RAM is too expensive, check out second-hand RAM on eBay. Look for someone with a good history rating and keep your fingers crossed. If you get "stiffed", eBay will refund your money. I've bought loads of second-hand DIMMs on eBay, including ECC Server RAM.

 

[amrgul1]

Hi again. In case you are still interested in my situation, I bought 2 Corsair Vengeance 3000MHz and no restarts and everything seems to be good (though it's hard to be 100% sure). Just what bugs me about this new rams, is that motherboard made their default frequencies - 2133 which is low and not desirable, while its highest supported default is, as I already discussed - 2666MHz. If anyone could explain me why mobo did so, I would be much grateful. I went and manually set frequencies to its highest default and is working normally (at least until now), also tried to the highest cpu supported - 2933MHz and also seems to be working. I decided I will use 2800MHz, a bit overclocked but I think should be normal. When I set it to 2933 and also 2800, I noticed that DRAM voltage didn't changed - it's still on 1.2 though latencies increased from 15, 15... to 20, 19, 19, 19... (on 2800). Is this how it should look like? meaning DRAM voltage not increased but latincies did. Maybe I should also manually increase DRAM voltage?! Is 1.2 enough for 2800MHz?

edit: Oh I just saw Aida64 shows memory speed (xmp) - 3000MHz and memory speed - 2133MHz. So does this mean that 3000MHz is only overclock profile and the default is 2133?! I'm a bit confused now... Also Aida shows memory timings on 1405 (xmp) - 15-19-19-36, should I set this manually?

Update: I now noticed in bios an "Overclocked Tuner" appeared with D.O.C.P and D.O.C.P Standard profiles. Now I see that I should set that and Cpu target frequency - for example 2800MHz and D.O.C.P will adjust other parameters automatically. Just what's the difference between D.O.C.P and D.O.C.P Standard?!

 

[Harlow]

Don't get annoyed because your motherboard booted up with the new RAM at 2133MHz. As I mentioned in an earlier posting, this action ensures that most pairings of RAM and mobo will POST (start up) normally. The "low" frequency of 2133MHz is the JEDEC default for DDR4. All DDR4 motherboards are guaranteed to work at 2133MHz. There is no guarantee the motherboard will boot at higher (XMP) frequencies. That's for you to determine llater.

After fitting new RAM, it's a good idea to boot into Windows (Linux, etc) and run a few tests to make sure the system is stable. There's no sense applying a huge memory overclock if the RAM is unstable at the JEDEC default 2133MHz. Once you're satisfied the RAM works at 2133MHz, you can try out some XMP settings. Faster XMP frequencies increase memory bandwidth, but you might not be able to detect a significant improvement in most programs.

As you've spotted in Aida64, your DIMMs have multiple XMP settings for the same clock frequency. When you pick an XMP frequency greater than 2133MHz, the UEFI/BIOS usually picks the fastest timings programmed into the SPD (Serial Presence Detect) chip at that frequency. If the fastest timings (lowest numbers) prove unstable, you can try slower timings (higher numbers) at the same frequency, but you may have to copy these numbers shown in Aida64 over into the Primary, Secondary and Tertiary sections in the UEFI/BIOS (both DIMMs).

If you invoke D.O.C.P. (easiest option) to overclock RAM, it will scan the XMP timings in the SPD and make the all necessary changes for you. The BIOS might not use all the SPD timings for a particular XMP profile. For example, it could set a different tRFC value that's not in the SPD table. This is quite normal.

After applying a new XMP setting, it's wise to run a stress test, e.g. the System Stability Test under Tools in Aida64. Your RAM might seem to be running fine at 3000MHz, but BSOD during a stress test.

If you plan to overclock the CPU, keep an eye on the temperatures during a stress test. I don't know if you're using the stock AMD air cooler on the 2600 or something like a triple 140mm Corsair H170i. High processor overclocks benefit from a decent cooler to avoid CPU throttling. I like the big Noctua NH-D14 and NH-D15 air coolers. Don't expect too much from your A320M mobo because the VRM's are not designed for heavy CPU overclocking.

 

[amrgul1]

Don't get annoyed because your motherboard booted up with the new RAM at 2133MHz. As I mentioned in an earlier posting, this action ensures that most pairings of RAM and mobo will POST (start up) normally. The "low" frequency of 2133MHz is the JEDEC default for DDR4. All DDR4 motherboards are guaranteed to work at 2133MHz. There is no guarantee the motherboard will boot at higher (XMP) frequencies. That's for you to determine llater.

After fitting new RAM, it's a good idea to boot into Windows (Linux, etc) and run a few tests to make sure the system is stable. There's no sense applying a huge memory overclock if the RAM is unstable at the JEDEC default 2133MHz. Once you're satisfied the RAM works at 2133MHz, you can try out some XMP settings. Faster XMP frequencies increase memory bandwidth, but you might not be able to detect a significant improvement in most programs.

As you've spotted in Aida64, your DIMMs have multiple XMP settings for the same clock frequency. When you pick an XMP frequency greater than 2133MHz, the UEFI/BIOS usually picks the fastest timings programmed into the SPD (Serial Presence Detect) chip at that frequency. If the fastest timings (lowest numbers) prove unstable, you can try slower timings (higher numbers) at the same frequency, but you may have to copy these numbers shown in Aida64 over into the Primary, Secondary and Tertiary sections in the UEFI/BIOS (both DIMMs).

If you invoke D.O.C.P. (easiest option) to overclock RAM, it will scan the XMP timings in the SPD and make the all necessary changes for you. The BIOS might not use all the SPD timings for a particular XMP profile. For example, it could set a different tRFC value that's not in the SPD table. This is quite normal.

After applying a new XMP setting, it's wise to run a stress test, e.g. the System Stability Test under Tools in Aida64. Your RAM might seem to be running fine at 3000MHz, but BSOD during a stress test.

If you plan to overclock the CPU, keep an eye on the temperatures during a stress test. I don't know if you're using the stock AMD air cooler on the 2600 or something like a triple 140mm Corsair H170i. High processor overclocks benefit from a decent cooler to avoid CPU throttling. I like the big Noctua NH-D14 and NH-D15 air coolers. Don't expect too much from your A320M mobo because the VRM's are not designed for heavy CPU overclocking.

First of all, I want to express my respects that you are so helpful and don't hesitate responding with so much comprehensive explanations. Yes, I remember you said 2133MHz is Jedec standard as also 2400MHz you said. That's why logically (imo) would be for the motherboard to apply the highest supported and stable frequency for better results. I only say this because, not every user who does not have a clue in such details, would go and search - is his PC parts working and giving back the quality and value, the user paid for. Well, maybe there is not big difference, but for my - as a user standpoint it would be more logical that motherboard give the highest supported results (I don't want to make debates in this, I learned the things just now ).
D.O.C.P. profiles only appeared now with Corsair, previous rams didn't have it, so I set it with Cpu 2800MHz frequency and when applied D.O.C.P. it increased DRAM voltage to 1.35 and lowered CL latency to 16 (others are still on 20). I never had in mind to overclock something, just when encountered this problem with ram and learned so much here, made a "brave" step to increase frequency a bit. Yesterday I run Memtest 7 pass again on these new rams, will do Aida stability test also (30 min at least?!). I'm now using stock Amd cooler because it fits normally, before had Thermaltake 120 mm but was too large for my microatx board, so I changed it. Lol. Noctua is hard to find here not to mention its price
Also want to have couple of words about the real reason why previous rams didn't work. It's because, as I think now - one of them may be broken and it was not the case with motherboard incompatibility ( I hardly imagine all people in the world buying ram, first go and look QVL list). I once mentioned here that though they work separately without restarts but sometimes I noticed mouse freezes and I think this only happens with one of them specifically. So one of them might be somehow faulty and if this is true, then this also means (don't want to hurt someone's feelings) that Memtest is not reliable (I even did 10 pass after).

Thanks again to everyone, I really learned much from you

 

[Harlow]

MemTest86 is a good guide but not infallible. If a DIMM still causes BSODs after passing MemTest, don't use that DIMM. I would say though that MemTest86 usually weeds out bad DIMMs. I only had 3 fails using MemTest86 on a batch of 60 second-hand 4GB DDR3 DIMMs (computer pulls) and they work OK in numerous old PCs. I only run 3 complete passes per DIMM when bulk testing old RAM.

Regarding stress testing after applying a RAM, CPU or GPU overclock, some people advise 5 minutes for a "quick and dirty" test, before increasing the overlock setting up one notch. Others say you should stress test your final overclock for up to 24 hours. How long you test your RAM overclock is entirely up to you. If you can tolerate an occasional unexpected BSOD, that's fine. If your system's stability is important, be more cautious. You don't see many overclocked CPUs in 'mission critical' server computers, although this may be changing with a few specific server CPU releases.

N.B. The default settings in the Aida64's Stability Test include 'Stress CPU', 'Stress FPU', 'Stress cache' as well as 'Stress system memory'.

You might consider removing the check marks from the first three boxes (Stress CPU, FPU & cache) to reduce the power dissipated in the CPU. If you leave these three boxes checked, CPU power will ramp up towards a nominal 95W (2600's TDP with no overclock) and the processor might even reach thermal throttling (95C junction temp) with your stock AMD heatsink and fan.

For initial testing where you want to concentrate on RAM, it might be better to disable the Aida's CPU stress tests and run them at a later date, when you're more confident the memory overclock is stable.

 

[amrgul1]

You might consider removing the check marks from the first three boxes (Stress CPU, FPU & cache) to reduce the power dissipated in the CPU. If you leave these three boxes checked, CPU power will ramp up towards a nominal 95W (2600's TDP with no overclock) and the processor might even reach thermal throttling (95C junction temp) with your stock AMD heatsink and fan.

Though removing first three checkboxes and leaving only stress memory still stresses cpu 100%. I did that for 30 min and it passed the test and temperatures were quite normal - 51 C Cpu, 46 C - Ram. Speed of fans also increased - on a stock Amd Cpu cooler it reached 1800+, which in normal situation is about 1300-1400.
So, everything looks normal now The problem was to take steps and diagnose so far, at least for me as a user and not a technician was not easy, I was feared that it might be Motherboard or even worse - Cpu, glad it was only Ram. Well, my PC is not very strong of course so overclocking is the concern at all. Good usability also might be achieved with taking good care on the PC, meaning disabling unnecessary things, diagnosing what slows it down, removing junk files. Antivirus suits also might become big pain if all their secure features are turned on...

 

[Harlow]

I've just run a quick Aida64 Stress System Memory test on an old i5-4670K system and checked CPU power and temperature using CPUID HWMonitor.

With Aida64 'Stress CPU, FPU and cache' disabled, the max CPU power is 81.83W and the max temp is 53C.
With 'Stress CPU, FPU and cache' enabled, max CPU power increases to 97.93W and the temp rises to 71C.

The difference between the two tests is not huge, but power does increase by 16.1W from 81.83W to 97.93W and temperature rises by 18C from 53C to 71C. The full set of AIDA64 stress tests does push your CPU hard and I don't see any point running it for too long. No point in breaking your PC if it's your only machine.

Note, I've set x43 CPU multiplier in the BIOS for a nominal 4.3GHz overclock. The default clock speed for the i5-4670K is much lower at only 3.4GHz.

CPU World shows the maximum temp for the i5-4670K is 72.72C with a TDP of 84W. My HWMonitor results show the full Aida64 stress test has pushed the CPU up near the maximum temp (71C is close to the 72C limit) and the overclocked power level (97.93W) is 14W over the stock TDP level (84W) at 3.4GHz. N.B. I am not using the standard Intel heatsink fan on this rig. It could not cope with a 26% overclock from 3.4 to 4.3GHz. Instead I use a much larger heatsink.

Removing junk files from your disk drive will not speed things up, especially if they're on the SSD (with the proviso your SSD isn't nearly full). A spinning hard disk drive might benefit from some pruning followed by a "defrag", because files stored near the centre of the disk, run at half the speed of files saved near the outer edge of the platters. I'd advise against defragging an SSD.

Disabling unnecessary programs from loading when Windows boots up (using the Startup tab in Task Manager) can improve OS loading times marginally and reduce the background tasks slightly. It's amazing what clutter accumulates when the OS installation is several years old.

To improve Windows Desktop refresh times with really old GPUs, I tweak 'Performance Options' as shown below. Search for "Adjust the appearance and performance of Windows" to reach this setting. What you lose in fancy screen animations, you gain with improved Desktop redraw times, but it won't make your games run faster. For that you'll more powerful GPU.