Question PhenomII x4 975BE mild OC to fis HT bus issue

May 25, 2019
46
6
35
Asus AM3+ motherboard
ASUS M5A78L-M Plus/USB3 DDR3 HDMI DVI USB 3.0 760G MicroATX Motherboard
CPU Phenom II x4 975BE
8gb AMD radeon DDR3 1866
Radeon HD3650 PCIE 512mb (OC 750/400) pathetic I know but just base video till I can upgrade to a r9 390x

problem is the system bus underclocks to 199.5 mhz x18 multi, cpu 3590 mhz and yes all spread spectrum funtions are disabled and bios lists HT link as 200
Possible solution raising system bus to 201 multi still 18 as my HS/F is not up to any major OC but does a good enough job @ stock
This actually results in a mild oc as CPU-z reports the system bus at 200.5 x 18, 3609 cpu
I know the difference in 3590 and 3609 is trivial but still I like to get the full speed out of it.
question is should this mild of an OC just to get essentially the speed I paid for affect temps very much, and can the moterboard tolerate .5mhz higher reference clock for extended use. (when I actually get a better heatsink and truly OC I'll be using the multi as thats the most effective on a black edition chip).

OS is windows 7 home premium x64 (dont care for the bloat and inbuilt spyware of win 10 not to mention the issue of app incompatibility)

I realize this probably seems like a pathetic oc, but as I said when the new heatsink arrives I'll kick the multi up to 20 at first, and see how it goes. I'd actually like to push it to 4.0 or maybe a bit more if all goes well with the new cooler.

I'm not new to this by any means. I been overclocking AMD chips since the 486DX series amd graphics cards since the radeon x850. I overclocked my 3650 by adding ramsinks and the cooler off an old dead 3850 then editing and flashing the bios with radeon bios editor to run at 750/400 and 1.3v instead of 1.25, still a weak card but the improvement is clearly noticeable.

the only reason I ask about the possibility of such a wimpy oc damaging motherboard components, is it also very slightly overclocks the 2000 HT link and the pcie bus as well and I know that can lead to trouble (though I doubt it with such a slight increase) That's why I usually just up the multi but this is the first board I've seen that ran the main ref clock under spec by default with all spread spectrum disabled.
 
Last edited:
May 25, 2019
46
6
35
Yeah I figured at the price point the 760G chipset wasn't the best probabaly why set to default speed it was underclocking the ref clock, but it was all I could afford. I did put some nice copper heatsinks on the VRM's to help with the 125w chip snd I've yet to see it downclock (aside from the constant .5mhz less ref clock constant that is) My case has good airflow and I added a fan to the chipset, so hopefully this board will be good enough for at least 3.8 (x20 multi)
 
It seems you are on the right track. You might be able to stabilize even a little bit more. The 975s were later chips so the silicon was a little more mature than the 965. Obviously vrms overheating will be your limiting factor of how hard you can push it. After you get some heat sinks on them, you could do the finger touch test. Hot is nornal, but if it burns you are too hot lol.

Also overclock the CPU-NB which is the memory controller for Phenom II. The higher you can push that, the better your latencies will be. 2.8-3.2GHz is a good goal on air for that. Pushing HT Link won't give you much result until you get a much stronger graphics card, but even then your core frequency will likely still be more bottleneck.
 
May 25, 2019
46
6
35
Thanks
The heatsinks I got on the VRM's seem to be working well. they get warm but not burning hot, and have a case exhaust fan drawing air across them as it evacuates heat from the case (Mid range Delta 92mm 102cfm 25mm thickness)
Power supply is 750w EVGA 80+ Platinum

That's a good tip about upping the cpu-NB clock. I'd read mixed information, some saying push it to 3.2 and others saying anything over 2.6 was unnecessary unless running DDR3 2400.

Speaking of graphics, I planned to get a r9 390x as I know that card has windows 7 drivers, but does anyone know if the newer RX series like a RX 580 has drivers for windows 7 64 bit
 
On the older C2 revision chips 2.6GHz NB was about the ceiling. Yours should be a later C3 revision. E0 went even higher. You can check yours in CPU-Z. You can test how much the L3 and memory latency improves in AIDA64. Gaming performance is very latency dependant and you will see those minimum frames improve as the pipeline gets less bogged down.
 
  • Like
Reactions: RodneyDawn
May 25, 2019
46
6
35
Yeah mine is a C3 revision
New heatsink arrived today. Idling @ 35c 3.8 stock vcore with cpu/nb @ 2.8 (slight voltage bump on cpu-nb) (AS Ceramique2 needs 25 hours to cure so temps may improve. no stability issues in windows so far, and only hit 54c stressing with OCCT for an hour but the real test will be overnight stressing once thermal paste hits max efficiency.
lapped the heatsink and cpu IHS to 3000 grit.
I'm thinking at this point 3.8 is probably the practical limit on air. At least with the 760G chipset. given the temps.

Might be able to get to 3.9 or possibly 4.0 by borrowing an old trick from the Athlon XP days and replacing the heatsink's fan with a 120mm Delta. (Heatsink uses a standard 120mm fan so Delta 120x25mm should fit just fine for 152cfm)
Noise isnt an issue as I've laready got a 92mm delta as an exhaust in addition to psu exhaust, a 40mm delta on the chipset and a 120mm delta as an intake.

Other peripherals include a zip 750 drive with pata to SATA adapter, a SATA Blu-Ray burner A WD Black SATA 6gbps 1tb SSD (Main system disk) and 3x WD Velociraptor 500gb HDD's (Data storage)
 
  • Like
Reactions: damric
May 25, 2019
46
6
35
The copper heatsinks that I attached to them with artic silver thermal adhesive get pretty warm but not uncomfortably hot. They have the exhaust fan and PSU fan both drawing air across them but I can easily add a dedicated fan to blow directly on them. 60mm should fit perfect between the back of the case and the large CPU cooler. I'll have to shut down and break out the JB-weld to secure it but then the vrm's should be good for a bit more.
 
May 25, 2019
46
6
35
UPDATE: With CPU fan upgraded from 120x25 basic low noise PWM to 120x25 Delta 152 cfm full load burn temps dropped to 49c

ran Ungine valley on max (yeah gpu not up to it capped frames pretty bad) no artefacts (further proving my already tested proven GPU overclock ) no hangs, crashes, etc.
tested for stability and microstutter playing games-so far no problems.
tested booting several linux distros from live usb drives (Kali, Debian, Ubuntu) no problems
3.8 is solid.

going to reboot and try for 3.9 as I finish posting this.
 
May 25, 2019
46
6
35
Booted into windows at 3.9, slight vcore bump 1.365 (up from 1.3500)
CPU-NB at 2.9
Load temp 52C
Passed overnight Prime95.
Slight improvement in Ungine Valley (still GPU limited)

I added a 60mm fan to blow directly on the VRM heatsinks and they feel considerably cooler now. still warm but not hot.
Going to try for 4.0 (3.0 CPU-NB)

UPDATE: 4.0 @1.365 vcore, 3.0 cpu-NB
Load temp 54c
still have to do prime95 0vernight to check stability but it passed an hour in OCCT and ran ungine valley.
VRM heatsinks are noticeably warmer than at 3.9 even with the fan blowing directly on them so that's about as high as the multi is going to go. have to fine tune with ht ref clock though.

If it proves to be truly stable I'll leave the multi there and bump up the ht ref clock a bit.
 
Last edited:
  • Like
Reactions: damric
Booted into windows at 3.9, slight vcore bump 1.365 (up from 1.3500)
CPU-NB at 2.9
Load temp 52C
Passed overnight Prime95.
Slight improvement in Ungine Valley (still GPU limited)

I added a 60mm fan to blow directly on the VRM heatsinks and they feel considerably cooler now. still warm but not hot.
Going to try for 4.0 (3.0 CPU-NB)

UPDATE: 4.0 @1.365 vcore, 3.0 cpu-NB
Load temp 54c
still have to do prime95 0vernight to check stability but it passed an hour in OCCT and ran ungine valley.
VRM heatsinks are noticeably warmer than at 3.9 even with the fan blowing directly on them so that's about as high as the multi is going to go. have to fine tune with ht ref clock though.

If it proves to be truly stable I'll leave the multi there and bump up the ht ref clock a bit.

If that's stable, that's a really nice clock for phenom II. Pretty low voltage too. I had about a dozen different Phenom II samples, and I swear all of them were leaky voltage pigs that would eat over 1.5v to stabilize 4GHz and ran very hot due to that fact. 3GHz is way above average overclock on the NB so that's very nice also since L3 cache runs at that frequency plus it reduces RAM latency.
 
May 25, 2019
46
6
35
3.9 is stable, still waiting to see if 4.0 will be. Got Prime95 running on it now. I'll update in the morning whether it proves stable or not but so far it seems to be.

The newer revision Phenom II's like my 975 scale better than the old Phenom II revisions (up to the 965) were voltage sponges, and muniature heaters because of it but the newer chips clock higher with a lot less voltage from what I've seen.
970 975 and 980 are the newer revision chips.
 
May 25, 2019
46
6
35
Passed overnight prime95 amd subsequently ran memtest for 5 full passes to check for memory controller issues.
No errors in memtest.
Bumped voltage up to 1.375 and increased HT ref clock to 203 for 4060mhz cpu speed. (4050 in cpu-z as my motherboard slightly under-drives the ht ref clock)

full load temp @4060 (4050) 1.375v 56c
If this proves to be stable I'm going to leave it there. I'm getting close to the max my cooling can cope with, and the VRM's are pretty warm as well.
no issues in ungine valley or ungine heaven, though framerate was increased an average of 1-2fps upping the HT ref clock
Ran OCCT for an hour without errors, and now running prime95
 
  • Like
Reactions: damric
May 25, 2019
46
6
35
upper limit on RAM for stability is 2133. its rated for 1866 and will boot windows @ 2400 with a slight voltage bump but fails memtest unless timings are relaxed to 10-10-10-28

@2133 with the same voltage bump it runs stock timings 9-9-9-24 and passes memtest every time (I let it keep running overnight and it never failed a pass)
I'm using a corsair memory cooling fan to keep the temps down.

1 fps increase in ungine valley benchmark as a result
(GPU is the bottleneck)

Not much more can be done for the GPU (Besides replacing it with a R9 390x which I ordered today (concern about whether the RX series is win 7 compatible is why I went with a R9)

the Radeon 3650 is the older DDR2 model and the ram is already pretty warm at 400mhz despite the added ramsinks, which is why I didnt bump it up when I boosted the core to 750 with Radeon Bios Editor (tested in catalyst control center overdrive function first) Cheap aftermarket cooler (was 15.00 on amazon) and Artic Silver Ceramique. never sank much into this card as it was intended from the start to simply be a ''basic'' 3d card for Aero until I could order a R9 390x, but all things considered it performs pretty well for what it is.

Why I went with the 975BE as a CPU for this system was, it was more cost-effective than the 980, and generally overclocks better than the Thuban core series (Phenom II x6)
I considered a FX8350 as well, due to the similar price point with the 975, but after doing a lot of research I noticed that A:, windows 7 needs a hotfix for Bulldozer to work optimally, and B: the Phenom II's have higher instructions per clock than the FX series, not to mention very few games I run can fully utilize all four cores of the phenom, so the FX's additional cores wouldn't make up for the lower instructions per clock.
True, the newer technology would make the FX the ''better'' of the two chips, but for my particular usage the phenom ii was the more effective option.
 
  • Like
Reactions: damric