[SOLVED] AMD FX6300 and Asus970 bottleneck

[Elpasadobl]

Hello dear friends! I come to you in dire need of help, hence my electronic calculating machine is unable to give me joy in playing games no more.

I have a rusty bucket with junk old machine and lately was sometimes even unable to watch videos in youtube - constant stutters, same in games. After getting 2 "Nvidia driver was recovered" one after the other, I cleaned my PC, reinstalled all drivers, and installed additional coolers.

I noticed that I had memory on JEDEC by default on 1066Mhz, I changed it to 1333Mhz, and then Isince it tasted nice, I also overclocked everything (Memory at 1333Mhz, CPU at 4Ghz, NB at 2600 Mhz, HT at 2400 Mhz, working stable). The results were incredible, everything worked like a charm, but after some more tweaking, stutters returned.
I thought that perhaps I overclocked too much and downclocked, but the problem continues, although not that noticeable.
One of the problems was the hot VRM and NB, so I put a coller right on them.

In most of benchmarking software, I get good results from OC, except for some strange ones: In Furmark I once got stutters and a bad result for my GPU and in Novabench I got good results for my memory and CPU but bad ones for my Disk and GPU.
Best and worst Novabench score with same specs.
Specs:
Windows 7 64bit
Asrock 970 Pro3 R 2.0 motherboard
AMD FX 6300 at 3.7 Ghz, 1.35V
North Bridge at 2600 Mhz, 1,38V
PCIE at 101Mhz (I read at another forum that it helped someone.)
Memory 2x GSkill 4Gb 1600Mhz and 2x cheap unknown manufacturer 4Gb 1333Mhz in dual channel, at 8-8-8-23 ticks and 1333 Mhz. Even the slightest overclock of memory above 1333 makes the system unstable, despite the cheap one having XMP profiles for 1500 and 1600 Mhz.
EVGA GTX1050 Ti 4Gb, no overclocking, new.
Kingston 240Gb SSD, new.

All drivers are up to date.

TLDR: I got good benchmarking results in RAM and CPU overclocking but overall worse performance. at first it looked like VRM and insufficient cooling, but after solving that the problem still stands. In Novabench I get good results for CPU and RAM even if I go over the stable theshold, but the GPU and disk benchmarking numbers get worse. In other benchmarking software I was unable to find abnormalities. Also Windows takes a long time to load at startup, but for some brief time it worked like a lightning.
My suspicions:
The NB is tricky
The HT is tricky
The MB is crap



Please please help, my friend who got a FX8350 and an Asus M5A78L-M LX3 with a GTX1660 has same symptoms in Novabench. And has even worse fps in "Escape from Tarkov".

Hoping to hear your wisdom soon.

 

[ChevetteSCx]

FIRST things First. DO NOT MIX-AND-MATCH RAM!! Especially Name Brand with an ''unknown manufacturer'' (junk). You can't be sure that the timings and voltages will even be close to the same!!!
Does the ''Unknown RAM'' even have heat Spreaders?* If you need 16GB, then go find/purchase a set of 4x4GB, or 2x8GB...but for the time being, GET THE JUNK OUT, and let the GSkills RUN FREE!!

 

[zx128k]

Whats your PSU? Make and model. Normally you would need a 300 Watt or greater power supply. For the 1050ti.

 

[Karadjgne]

Bah. Nonsense. There's absolutely nothing wrong with 'no-name' ram or mixing ram. Mixing ram is a bad idea if buying ram, since there's no guarantee that it will work or create headaches. OP already owns the ram, and it works. There's no such thing as 'no-name' ram. All ram, from Gskill to Adata to Kingston is made by a handful of OEMs, who do nothing more than slap a heatsink with a brand on it. You'll find the same Samsung IC's in Patriot Elite as Adata XPG as Gskill Trident-Z, or the same Micron IC's in HP's or Crucial or Dell. Lack of a branded heatsink is meaningless to performance, just looks. Mixed ram runs at the lowest speeds of all the sticks, so you won't get 1600 when it's mixed with 1333, all will run at 1333 or all will run at 1600 if you set them to. You may have to bump the ram voltages and/or NB VID, NB FID or CPU/NB to get them stable. But that can also affect VRM/NB temps.

What op should start with is checking windows event viewer, and not only for any critical errors (red flagged) but also for any repetitive errors (multiple yellow flagged) that could point in a general direction if there's a serious amount of conflicts affecting performance. A clean pc doesn't always mean a healthy OS and Windows has had more than its fair share of conflicts with every driver out there.

 

[CountMike]

For some of GPU glitches I would also suspect PCIe bus over 100.

 

[Karadjgne]

It's also possible it's a bios glitch with CSM. These new cards from Maxwell on far prefer uefi bios and can have issues with legacy bios that's unpredictable. I've seen everything from flat out refuses to work to working flawless and it changes from card to card.

 

[Elpasadobl]

Whats your PSU? Make and model. Normally you would need a 300 Watt or greater power supply. For the 1050ti.

TACENS something something 750 Watt, the 1050ti doesnt require additional power supply in my case. I checked the voltage on 5V and 12V, seems fine.

 

[Elpasadobl]

Bah. Nonsense. There's absolutely nothing wrong with 'no-name' ram or mixing ram. Mixing ram is a bad idea if buying ram, since there's no guarantee that it will work or create headaches. OP already owns the ram, and it works. There's no such thing as 'no-name' ram. All ram, from Gskill to Adata to Kingston is made by a handful of OEMs, who do nothing more than slap a heatsink with a brand on it. You'll find the same Samsung IC's in Patriot Elite as Adata XPG as Gskill Trident-Z, or the same Micron IC's in HP's or Crucial or Dell. Lack of a branded heatsink is meaningless to performance, just looks. Mixed ram runs at the lowest speeds of all the sticks, so you won't get 1600 when it's mixed with 1333, all will run at 1333 or all will run at 1600 if you set them to. You may have to bump the ram voltages and/or NB VID, NB FID or CPU/NB to get them stable. But that can also affect VRM/NB temps.

What op should start with is checking windows event viewer, and not only for any critical errors (red flagged) but also for any repetitive errors (multiple yellow flagged) that could point in a general direction if there's a serious amount of conflicts affecting performance. A clean pc doesn't always mean a healthy OS and Windows has had more than its fair share of conflicts with every driver out there.

Thanks for the tip, I will check the Windows event viewer. As for "no-name" RAM - the chip manufacturer doesnt show in Aida64 nor anywhere else, the only thing i got was "Module Name A Force 51264Y133I". But whatever.

Today I will check what was recommended - I will get a perfectly fine working power source just in case and report back.

 

[Elpasadobl]

For some of GPU glitches I would also suspect PCIe bus over 100.

I put int over 100 only recently, I had glitches before that. I will report back soon

 

[King_V]

My first suggestion would, honestly, be to back off all the overclocking. Do not try to crutch the system with an overclock when you don't know for sure what is causing your performance issues.

The increase in heat production and power draw is disproportionately greater than the increase in performance from overclocking. That can cause throttling.

TACENS something something 750 Watt, the 1050ti doesnt require additional power supply in my case. I checked the voltage on 5V and 12V, seems fine.

That could be your problem right there. I would not trust that PSU. What little I was able to find (from about 4 years ago) suggests it's not a good unit. It may CLAIM to be a 750W PSU, but I am guessing it likely doesn't make close to that on the 12V rail, and is likely to be poor quality in terms of how "clean" the power it provides to your components would be.

If you can get a photo of the power specs label from it, that would be useful information.

 

[Karadjgne]

A Force Manufacturing Ltd is a brand of ram. Nothing fancy. The 133 denotes it as 1333MHz. Because of its 'low-end' ness, it's more than likely not going to be able to handle 1600 well, if at all, but shouldn't have any issues at 1333.

Good idea replacing thy Tacens unit, it's a 750w claim that encompasses all the rails, and then adds more wattage for flavor. The guts are closer to 350w rated. Any decent 450w - 550w will easily handle your pc, a good substitute would be something like the Seasonic S12-II or M12-II 520w. Older unit design, but fits perfectly with your older equipment on a budget.

 

[Elpasadobl]

Thanks for the tip, I will check my PSU again.

 

[ChevetteSCx]

@Karadjgne

TLDR: I got good benchmarking results in RAM and CPU overclocking but overall worse performance. at first it looked like VRM and insufficient cooling, but after solving that the problem still stands. In Novabench I get good results for CPU and RAM even if I go over the stable theshold, but the GPU and disk benchmarking numbers get worse. In other benchmarking software I was unable to find abnormalities. Also Windows takes a long time to load at startup, but for some brief time it worked like a lightning.
My suspicions:
The NB is tricky
The HT is tricky
The MB is crap

You said it...

Mixed ram runs at the lowest speeds of all the sticks, so you won't get 1600 when it's mixed with 1333, all will run at 1333 or all will run at 1600 if you set them to. You may have to bump the ram voltages and/or NB VID, NB FID or CPU/NB to get them stable. But that can also affect VRM/NB temps.

I have spent the last year-or-two building some ''newer'' systems, and playing with various Manufacturers and Models of AM3/AM3+ Motherboards (ASUS, Gigabyte) and CPUs, as well as DDR3 RAM (G.Skill, Corsair, Patriot, OCZ and Kingston), for fun as well as a learning experience! I can say with certainty that, REGARDLESS of whose ''chips'' are under those pretty heat spreaders, the support components and build quality of those DRAM CAN and WILL affect performance. OCZ and G.Skill tend to run their DDR3 @1.65V for the highest timings, resulting in WAY MORE heat in the CPU and VRMs, typically resulting in throttling. I have managed to get some of their DDR3 pairs to run at top clocks, with voltages closer to 1.55V, but I still find that unacceptably high. Corsair, Patriot and Kingston will run their highest XMP timings with no issues @1.50V, and can typically be manually overclocked to the next level, at the same voltage. I can also tell you, that when timings and voltages for the DDR3 get ''unstable'' the system doesn't always just quit working. I have had it :
1.) post to BIOS;
2.) boot to Windows, and crash @The ''START" screen;
3.) boot into Windows and run normally, but reboot when stressed;
4.) boot to windows and throttle under load;
5.) boot to Windows and run CRAZY SLOW and WEIRD (ReTRY ReTRY ReTRY, Delay, Delay, Delay)

You appear experienced and knowledgeable in this arena, so mixing-and-matching RAM may be easy for you, but since the poster of this thread essentially appears to be a little new to this level of system tweaking, we have to go back to basics!! Have you checked to see if the

As for "no-name" RAM - the chip manufacturer doesnt show in Aida64 nor anywhere else, the only thing i got was "Module Name A Force 51264Y133I"

even shows up in ASRock's Memory QVL??

Furthermore, none of this precludes that the DDR3 RAM in question may be failing...or the Motherboard..or the PSU!!

Memory manufacturing is similar to a PSU, just because is says "EVGA", doesn't mean that a myriad of manufacturers (some pretty skeptical in component selection and build quality) haven't built it poorly! AND, on that note...poster GET A QUALITY, NEW PSU!! (and read this posters experiences, all because of his PSU...)

Question - Problems with Sapphire 5700 xt nitro+ special edition , change again for rx580 and still the same issues

Hello gents Hope everything is alright with you ! Before anything else , im going to leave my rig specs in here : PSU: cooler master 600w Motherboard: b450 tomahank msi CPU : Ryzen 5 2600 RAM: 16gb corsair 2800mhz SSD 1tb GPU : 1º rx 580 armour OC 8GB MSI , after Sapphire rx5700 xt nitro+...

forums.tomshardware.com

As for the Motherboard...it does appear to be a UFEI BIOS :

ASRock > 970M Pro3

Support for Socket AM3+ / AM3 Processors; Solid Capacitor design; Supports Dual Channel DDR3 2400+(OC); 2 PCIe 2.0 x16, 1 PCIe 2.0 x1, 1 PCI; 7.1 CH HD Audio (Realtek ALC892 Audio Codec), ELNA Audio Caps; 6 SATA3; 4 USB 3.1 Gen1 (2 Front, 2 Rear), 10 USB 2.0 (4 Front, 6 Rear); 1 TPM Header, 1 IR...

www.asrock.com

I stand by my original assessment, get the no-name RAM O-U-T!! (and reset EVERYTHING to factory, so we have a baseline...)

 

[Karadjgne]

Kinda depends on the ram. Gskill snipers in DDR3 ran low voltages, 1.35v. Most newer model ram ran 1.5v, it was only older model designs that ran 1.65v, as did all DDR3 when it was first adopted. But that has little or nothing to do with ram performance, and shouldn't affect amd cpu performance since the SB and NB chipsets are seperate from the cpu. It was only in later Intels, after the swap from Northbridge to pcie hub and moving the memory controller to the cpu that ram voltages had some affect on cpu temps. That's when VRM's could possibly be affected, but on those old am3 designs, it's the changes to vcore voltages and vid which affect the VRM's the most.

 

[ChevetteSCx]

Alot of manufacturers adopted ''LV'' series near the EOL of DDR3. But, since the poster did not specify ACTUAL MODEL #s or DDR3 VOLTAGES in their system specs, we're kinda flying blind, here. Also, if they mixed LV RAM, like the G.Skill RipJaws X F3-1600C9S-8GXLL, with some ''off brand'' older model, 4th tier 1.65V or worse 1.7V-to-1.9V DDR3, then they will have problems...as some RAM JUST DOES NOT respond well to higher voltages, especially ''chip only'' models, without pretty heat spreaders. We simply DO NOT HAVE enough information, nor does the poster appear to have the knowledge or experience to ''sync up'' these modules so they function properly...

 

[Elpasadobl]

Hello guys! Thanks everyone for the help! In the end it was VRM zone overheating, the air current was not reaching it despite a fan being pointed right at it. Was solved by attaching a 100mm fan closer to the VRM mosfets and downclocking a bit. As right now system seems stable under load and works fast. However, my friend still has this issue, and his VRM is not hot.

 

[Elpasadobl]

Hello! Everything is working at 1333 Mhz 8-8-8-23 at 1.5V

 

[Elpasadobl]

Alot of manufacturers adopted ''LV'' series near the EOL of DDR3. But, since the poster did not specify ACTUAL MODEL #s or DDR3 VOLTAGES in their system specs, we're kinda flying blind, here. Also, if they mixed LV RAM, like the G.Skill RipJaws X F3-1600C9S-8GXLL, with some ''off brand'' older model, 4th tier 1.65V or worse 1.7V-to-1.9V DDR3, then they will have problems...as some RAM JUST DOES NOT respond well to higher voltages, especially ''chip only'' models, without pretty heat spreaders. We simply DO NOT HAVE enough information, nor does the poster appear to have the knowledge or experience to ''sync up'' these modules so they function properly...

You are right, I am pretty new to it.
Before syncing I attempted to make the simple green ram (the no name one) work at the stated in Aida64 for that RAM XMP profiles 1.1/1.2 of 1500Mhz@1.6V or 1600Mhz@1.65V, however despite trying different clocks and voltages (up to 1.75V) I was never able to go above 1400Mhz, and that was done overclocking the BCLK, and the system wasnt stable.

 

[Elpasadobl]

@Karadjgne


You said it...


I have spent the last year-or-two building some ''newer'' systems, and playing with various Manufacturers and Models of AM3/AM3+ Motherboards (ASUS, Gigabyte) and CPUs, as well as DDR3 RAM (G.Skill, Corsair, Patriot, OCZ and Kingston), for fun as well as a learning experience! I can say with certainty that, REGARDLESS of whose ''chips'' are under those pretty heat spreaders, the support components and build quality of those DRAM CAN and WILL affect performance. OCZ and G.Skill tend to run their DDR3 @1.65V for the highest timings, resulting in WAY MORE heat in the CPU and VRMs, typically resulting in throttling. I have managed to get some of their DDR3 pairs to run at top clocks, with voltages closer to 1.55V, but I still find that unacceptably high. Corsair, Patriot and Kingston will run their highest XMP timings with no issues @1.50V, and can typically be manually overclocked to the next level, at the same voltage. I can also tell you, that when timings and voltages for the DDR3 get ''unstable'' the system doesn't always just quit working. I have had it :
1.) post to BIOS;
2.) boot to Windows, and crash @The ''START" screen;
3.) boot into Windows and run normally, but reboot when stressed;
4.) boot to windows and throttle under load;
5.) boot to Windows and run CRAZY SLOW and WEIRD (ReTRY ReTRY ReTRY, Delay, Delay, Delay)

You appear experienced and knowledgeable in this arena, so mixing-and-matching RAM may be easy for you, but since the poster of this thread essentially appears to be a little new to this level of system tweaking, we have to go back to basics!! Have you checked to see if the

even shows up in ASRock's Memory QVL??

Furthermore, none of this precludes that the DDR3 RAM in question may be failing...or the Motherboard..or the PSU!!

Memory manufacturing is similar to a PSU, just because is says "EVGA", doesn't mean that a myriad of manufacturers (some pretty skeptical in component selection and build quality) haven't built it poorly! AND, on that note...poster GET A QUALITY, NEW PSU!! (and read this posters experiences, all because of his PSU...)

Question - Problems with Sapphire 5700 xt nitro+ special edition , change again for rx580 and still the same issues

Hello gents Hope everything is alright with you ! Before anything else , im going to leave my rig specs in here : PSU: cooler master 600w Motherboard: b450 tomahank msi CPU : Ryzen 5 2600 RAM: 16gb corsair 2800mhz SSD 1tb GPU : 1º rx 580 armour OC 8GB MSI , after Sapphire rx5700 xt nitro+...

forums.tomshardware.com

As for the Motherboard...it does appear to be a UFEI BIOS :

ASRock > 970M Pro3

Support for Socket AM3+ / AM3 Processors; Solid Capacitor design; Supports Dual Channel DDR3 2400+(OC); 2 PCIe 2.0 x16, 1 PCIe 2.0 x1, 1 PCI; 7.1 CH HD Audio (Realtek ALC892 Audio Codec), ELNA Audio Caps; 6 SATA3; 4 USB 3.1 Gen1 (2 Front, 2 Rear), 10 USB 2.0 (4 Front, 6 Rear); 1 TPM Header, 1 IR...

www.asrock.com

I stand by my original assessment, get the no-name RAM O-U-T!! (and reset EVERYTHING to factory, so we have a baseline...)

Thanks! I will see to it. I already ordered new 2 sticks of 8GB DDR3 at 1866Mhz 2 weeks ago. The no-name RAM is legacy from my old PC. Unfortunately the tacens doesnt have any stickers except for "750W". I will try to find a good PSU, but I am a little short on a budget, is there any way to tell a good PSU outside of voltage/current measurment?

 

[Elpasadobl]

I would like to thank everyone again for taking their time to help me out.
Just in case, after I finish working I will set everything to default and we will start testing again.

 

[drea.drechsler]

Hello dear friends! I come to you in dire need of help, hence my electronic calculating machine is unable to give me joy in playing games no more.

I have a rusty bucket with junk old machine and lately was sometimes even unable to watch videos in youtube - constant stutters, same in games. After getting 2 "Nvidia driver was recovered" one after the other, I cleaned my PC, reinstalled all drivers, and installed additional coolers.

I noticed that I had memory on JEDEC by default on 1066Mhz, I changed it to 1333Mhz, and then Isince it tasted nice, I also overclocked everything (Memory at 1333Mhz, CPU at 4Ghz, NB at 2600 Mhz, HT at 2400 Mhz, working stable). The results were incredible, everything worked like a charm, but after some more tweaking, stutters returned.
I thought that perhaps I overclocked too much and downclocked, but the problem continues, although not that noticeable.
One of the problems was the hot VRM and NB, so I put a coller right on them.

In most of benchmarking software, I get good results from OC, except for some strange ones: In Furmark I once got stutters and a bad result for my GPU and in Novabench I got good results for my memory and CPU but bad ones for my Disk and GPU.
Best and worst Novabench score with same specs.
Specs:
Windows 7 64bit
Asrock 970 Pro3 R 2.0 motherboard
AMD FX 6300 at 3.7 Ghz, 1.35V
North Bridge at 2600 Mhz, 1,38V
PCIE at 101Mhz (I read at another forum that it helped someone.)
Memory 2x GSkill 4Gb 1600Mhz and 2x cheap unknown manufacturer 4Gb 1333Mhz in dual channel, at 8-8-8-23 ticks and 1333 Mhz. Even the slightest overclock of memory above 1333 makes the system unstable, despite the cheap one having XMP profiles for 1500 and 1600 Mhz.
EVGA GTX1050 Ti 4Gb, no overclocking, new.
Kingston 240Gb SSD, new.

All drivers are up to date.

TLDR: I got good benchmarking results in RAM and CPU overclocking but overall worse performance. at first it looked like VRM and insufficient cooling, but after solving that the problem still stands. In Novabench I get good results for CPU and RAM even if I go over the stable theshold, but the GPU and disk benchmarking numbers get worse. In other benchmarking software I was unable to find abnormalities. Also Windows takes a long time to load at startup, but for some brief time it worked like a lightning.
My suspicions:
The NB is tricky
The HT is tricky
The MB is crap



Please please help, my friend who got a FX8350 and an Asus M5A78L-M LX3 with a GTX1660 has same symptoms in Novabench. And has even worse fps in "Escape from Tarkov".

Hoping to hear your wisdom soon.

Did you enable 'high performance' mode in BIOS? I can't remember the exact name, but something like that.

What it does is it disables processor APM. When the processor is overclocked it will try to adhere to it's rating or TDP, that means it will down-clock periodically when under load. My son's Pro3 motherboard has that setting and it ran my FX6300 at 4.5Ghz great once I put it into high performance mode.