Asus Crosshair iv Formula issues
I put together a pc with used parts, here are the exact components:
ROG Crosshair IV Formula
Phenom II x6 1090T black edition @ 4ghz
Corsair Vengeance Lp 1600 16go
Gigabyte r9 290x
Cooler Master 212 evo
Seasonic 650w psu
I overclocked the cpu without a problem, it runs at ~20° idle and 40° under Prime95.
But I still have a few minor issues I'd like to understand.
When I press the power button it takes a while before the computer gets going.
The red lights (cpu, gpu etc.) blink once each time and only then the fans start spinning and the computer starts.
I did some cable management afterwards, don't mind the mess on the video ;)
I also ran a userbenchmark test and it says my ram is performing way below average.
I set the timings right I believe, so I don't understand where the issue could come from.
Then I used AHCI for my ssd but it doesn't appear in the bios?! It works like this and windows boots without a problem, but it seems the bios only wants to see IDE drives.
Also the gpu temperature is at 60° idle. Seems quite high to me?
Thanks for your answers :)
I have a M4A89GTD PRO (same bios versions, 3029 was last released with beta bulldozer support under AM3 - few boards were capable as these) running an OC'd 965 that does something similar. It probably has to do with requiring a reboot after first initialization to set stuff like nb speed or other roms (even my X470-F does it, but completely turns off fans in the process). When I had a ODD installed that would make a tray noise when starting, on a cold boot I'd hear it twice (instead of once) before POST showed up on screen.
That sata port page is ok, it works like that to show what sata 5-6 are detecting when they are set to legacy IDE mode separately from 1-4 in AHCI, as they aren't initialized (and devices listed) on the ahci rom anymore. With all ports set to ahci, it won't show any drive in there.
If you go to BOOT menu> Hard Disk Drives you'll see the remaining detected drives when you press enter in one of the ordered entries (ignore the "IDE" prefix that will show).
Thanks for your answer!
Is there anything I can do then to get it to start 'normally'?
Im used to computers starting when you press that switch;) its not a big deal but it bugs me.
Also, what about the ram? Anything wrong with my current settings?
To shorten first POST, you might try the latest bios and disable the most stuff you can like firewire and serial controllers, lan boot room, set all sata ports to ahci (you can still boot from your sata dvd by pressing f8 on post, after setting hdd/sdd as the only boot device), use the pair of slots most away from the CPU as its the better one in case of a memory related trouble. If applicable on that board, make sure you don't have the cpu unlocker dip switch enabled.
If bios are exactly like on mine and your board behaves the same, if you update your bios to a more recent one like 3029 you'll only be able to control cpu and cpu_nb voltage via offset (the advantage of it is that you can have lower idle voltage under CnQ on top of lower clocks). If you wish for those better idle temps and power consumption, CnQ requires a cpu mult only up to 18.5x, else it won't work for the Windows Balanced profile. OTH, by using something like 16.5x242 you'll gain a bit more performance at the same cpu clock (ht link/mem/cpu_nb mults will have to be adjusted), but slight voltage adjustments might be required as it works a bit harder. With 20X you can use K10Stat to control idle clocks and VID, but it reacts slower than native CnQ.
Btw, those temps (if core) are too good to be true under stress testing 6 cores at >1.4V over extend time (1-4h). For socket you should only go up to ~57ºC, and ~62ºC for core until errors start to appear on long runs. Either your testing is very short, or that bios versions doesn't report Thuban core temperatures correctly, because a 212 isn't that great to manage >150W.
That cpu_nb voltage left on auto at 1.356V is terribly high, for only 2600 I doubt you'll need something over 1.25-1.275V with an average CPU (my inferior Deneb die only needs ~1.225V for 2640 prime stable with cpu at 4080). To make it more difficult on testing, cpu_nb needs to be raised a bit along with cpu speed.
These cpus like good timings (ideally CL6 or 7 when at 1600 unganged), so you'll have to search on the particular revision/date of your sticks to know what ICs do they have; your IMC might be good up to ~1800-1900ish. That Corsair LP series wasn't particularly know use good bins/ICs, so when you settle on a CPU speed, you might try to keep raising htt and lower cpu_mult to see how high they can go on speed if timings can't be lowered at 1600.
Regarding the GPU, that idle temp really seems quite high for the default 300/150 (?) 2D clocks under a big cooler. You might try to apply fresh paste.
Damn thanks for the detailed answers. I'm quite a novice in overclocking, I just followed a youtube guide and that was it.
Here are my userbenchmark results if that's of any interest!
UserBenchmarks: Game 65%, Desk 51%, Work 42%
CPU: AMD Phenom II X6 1090T - 58.2%
GPU: AMD R9 290X - 75.7%
SSD: Corsair Neutron 128GB - 68.3%
RAM: Corsair Vengeance LP DDR3 1600 C9 4x4GB - 43.1%
MBD: Asus Crosshair IV Formula
I ran it a few times and I actually got 100th percentile with my cpu and 99th with the gpu.
The temp issue got better since I installed afterburner overclocked it and configured the fans with it.
I have 43° at idle with fans blowing at 45% though. Oc values are 1130 core clock and 1450 memory clock. I tried 1150/1500 but it was unstable.
So yeah I guess I should replace the paste, because hwmonitor says they run at 2.3k rpm at idle! :thumbsdow
I did as you told, lowered cpu nb voltage to 1.26 and it seems stable.
I don't quite understand what I should do with the ram though yet. I'm gonna run Prime95 now for a while to see how it goes and then change some things to the ram.
I should first try to raise HT Link Speed then?
I haven't updated the bios either, I thought it was up to date until you told me about version 3029. Will look into that.
edit: after an hour of Prime95, max temp is 47° and it usually is just over 40 on average. Blend test with defaults settings.
I actually got an error:
[Mar 1 12:21] Test 3, 4000 Lucas-Lehmer iterations of M19374367 using AMD K10 type-2 FFT length 1M, Pass1=256, Pass2=4K, clm=4.
[Mar 1 12:23] FATAL ERROR: Final result was A9120061, expected: D1045A66.
[Mar 1 12:23] Hardware failure detected, consult stress.txt file.
[Mar 1 12:23] Torture Test completed 25 tests in 31 minutes - 1 errors, 0 warnings.
[Mar 1 12:23] Worker stopped.
Oh, it's a 290x, not a 280x. Should run hotter than a Tahiti, but still, you should replace paste.
Don't worry about ht link speed, high values can actually hinder your performance or cause instability. Just keep it at 2000, or ~2000-2200Mhz when you're using a htt OC.
If you're still at the printscreen settings, you could try to figure if that p95 error is related to imc/ram (lack of vcore usually ends up in a bsod). To isolate it, lower cpu mult to 19.5x, raise cpu_nb to 1.3V just to check if that crash is related to the lowered cpu_nb voltage (those 16GB might require a bit more than I'm used to), if it passes, go to back to 20x, and slowly raise cpu_nb by 0.0125 steps until it stops crashing.
If you keep having problems, depending on CPU 1.425V at 4.0GHz might not be enough to be prime stable, even if it's a common value for a good clocker, it wouldn't be strange if you require ~1.45V. Beware that as you near 1.5V, temperatures and power consumption will raise a lot.
I doubt your ram at 1600 C9 would be the direct cause failing, but if so, try to raise it to 1.55V and/or 2T, in case it's a imc limitation with a big amount of ram. But if possible, you should avoid 2T if you can with only 2 sticks, because of the bandwidth drop.
After everything is stable, there are many things you can do about ram. One is to tighten timings a 1600, maybe up to 1.6V to be safe (if it's micron it can go higher), by going to 8-9-8-27, 8-8-8-27, 7-8-7-24, 7-7-7-24 etc. The other approach is to use a higher base htt (over 240) and lower the cpu mult, ht link mult and ram mult, so you'll go above 1600mhz with more relaxed timings.
When you go this route, using a higher than base htt for the same cpu and cpu_nb speed (242x16.5 instead of 200x200) will result in better performance, but might require a tad more cpu voltage because the cpu works harder. If htt OC is used in combination with CnQ instead of a locked p0 state, you should really set vcore with offset, so that p1,p2... states that will end up overclocked are also running with a higher voltage.
The amount of setting combinations in this platform, and the way they interact can take you some weeks until you find an optimal config.
Edit: You forgot to disable both spread spectrum options, this signal modulation can cause instability under OC.
I disabled both spread spectrum options!
I actually have 4 sticks of ram, 4x4 then. I put cpu nb voltage back to 1.3.
T2 doesnt get better results on userbenchmark. I ran a memtest86+, did 1 pass without an error.
I'll try some other things tomorrow, thanks again for the help! I learned a lot [IMG class=inlineimg]/forum/images/smilies/thumb.gif[/IMG]
With 4 sticks you might require 2T for stability. On my setup with a deneb, which has a imc that usually clocks ram less, running 12 mismatched gb (4+4+2+2) I can't get it free from memory errors at 1600 C7-8-7 1T under a 240htt no matter ram voltage (while 2T does fine), even if it's OK at 1T with a 200htt, and in the past one of the pair of sticks alone did fine at Cl6 1T (good old psc). The other problem with 4 sticks is that you might require a tad more cpu_nb to keep things stable, but that also varies a lot with cpu.
To test ram in a harder way, you can use hci memtest (free demo) under windows, by launching 4 instances testing ~3456MB each (ensure you have that free ram so you don't swap from system drive, else reduce it a bit), and let it complete at least 100% coverage without errors (400-800% is ideal if you have the time). But keep in mind that everything else must be stable before this ram tweaking part.
If you're afraid to corrupt the windows partition when applying tighter timings and failing to boot it (it's extremely rare, but can happen), you can just run memtest86 after exiting bios, and set it only for test5 (it's the faster to find errors, but they only show after finishing a pass) and do 1 or 2 passes before booting windows and using the usual stress tests.
Also two quirks about this platform/bios:
1)when you go into bios and make a ht-link or nb_cpu multiplier change, then save and exit, it's possible that most times you'll fail POST on the next boot (with a stable setting) if your board doesn't force an extra reset (my 890GX and 790FX don't). You'll just have to press reset when it happens, and then the system will boot normally every time.
2)these asus ami bios have a annoying tendency to auto change boot order after you enter bios with a pen drive plugged and exit+save, by adding it to the bootlist in 1st position. So it might happen to you sometimes with your memtes86 drive, by showing a boot error message after you remove your pen drive when attempting to boot from hdd. This will require you to enter bios and select your correct drive again as the first boot device.
Ok I updated the bios to the same version you have.
I did the memtest in windows and got no errors. I forgot to mention that before I oc'd the cpu, ram was running in single channel.
It only changed to dual once I changed things in the bios. Is that an issue?
It seems it's more stable with 1.3v cpu nb and 1.42 cpu. I still have way below average ram results in the benchmark, I could tweak things I guess but at 9-9-9-24 I should already have decent scores I think? Lowering timings wont give me +20% will it?
Yeah I noticed that memtest86 would start automatically if the usb drive was plugged which is a bit annoying but not a problem in the end.
I'm quite happy with what I've got except for the ram and the gpu temp which should be solved with new paste.
If you're talking about the benchmark linked above and this is your result at:
You simply can't expect the same results with your PhII vs z97 based systems (or Z68, Z77, etc) running the same ram, hence the significantly lower scores. It'd be different against other AM3 systems only, but almost certainly there aren't many in their database with that ram config.
And well, 1600 at C9 is about value ram territory, but even if you had C6,C7 capable sticks the gains would be minimal to make a difference in 2019. For your reference, on aida64 5.99 I'm getting about 20098MB/s read spead, write is 10579MB/s, all this under [email protected], [email protected], 12GB [email protected] 7-8-7-24 2T Unganged (secondaries are just default), 57.4 ns latency on a late 965BE C3+890GX. It's just a tad higher than your score as expectable for C7 vs C9, but scores can't be really compared directly between different programs (the latency test seems completely different). tl;dr, your score seems normal for your system.
Btw, showing single channel with 4x4gb sticks with all memory detected must have been a software mistake, it shouldn't be even possible to do so.
|All times are GMT -7. The time now is 01:42 PM.|
Powered by vBulletin® Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
User Alert System provided by Advanced User Tagging (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
vBulletin Security provided by vBSecurity (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2019 DragonByte Technologies Ltd.