Coming back to AMD...

[EnJoY]

Hey Guys,

I decided to sell my Core i7 3930K, ASUS RIVE and GTX680 4GB FTW+ and downgrade to an FX 8350, ASROCK Extreme9 990FX and HD7950 that I got a great deal on. I'm saving $1,000 in this swap, so it was really a no-brainer for me.

I'll be carrying over my 4x4GB G.Skill 2133MHz Ripjaws and Corsair H100 with me, as well as my Seasonic 1KW Platinum.

I wanted to get some impressions on the ASROCK. I haven't seen much talk about it because the Extreme9 seems to be fairly new, but the specs looked very solid and the price was better than anything in the same league as it. I'm somewhat of the belief that 990FX boards don't have much to differentiate themselves from one another in terms of performance or overclocking potential.

Anyhow, I will post my results with the chip and board once I get them humming. Hoping I can peek 5GHz, but I don't do enough on my system anymore to require that for 24/7.

EDIT: Got the chip today, stepping: FA 1252PGT - batch: 0507

[FlanK3r]

5 GHz with Vishera is possible, but you need good cooling (ideal wattercooling or allinone at least)

[EnJoY]

5 GHz with Vishera is possible, but you need good cooling (ideal wattercooling or allinone at least)

Yea, I'm not terribly concerned if I can get it stable, I'd just like to be able to run some tests at 5+ GHz.

[informal]

Stable at 5Ghz unlikely but benchmarks at 5+Ghz are very possible with good cooling.

[EnJoY]

Can anybody tell me anything about the stepping? FA 1252PGT - batch: 0507

Google is coming up with zilch except one guy on this board, but no OC results.

[tbone8ty]

i have fx-8350 1249pgt and it does 4.7ghz on my nzxt havik 140 heatsink

[l0ud_sil3nc3]

1244PGT here 5090 Cinebench R11.5 with 1.41v

[AMDforME]

The Asrock Extreme9 is quite similar to the Asrock 990FX Fatality Pro mobo with minor PCI/PCIe slot differences. There isn't much difference in any mobos now days with one chipset maker per CPU brand and with the memory controller now within the CPU. The mobo VRM and features are what tends to differentiate the products. A proper VRM design with a good heatsink tends to allow overclocking operation without chronic mobo failures. An highly overclocked FX-8350 draws a lot of power and the E9/Fatality Pro seem to be up to the task, IME. Testing has shown that there are a lot of BIOS adjustments on some models that serve no useful purpose other than for marketing hype.

[Daveburt714]

I wanted to get some impressions on the ASROCK. I haven't seen much talk about it because the Extreme9 seems to be fairly new, but the specs looked very solid and the price was better than anything in the same league as it. I'm somewhat of the belief that 990FX boards don't have much to differentiate themselves from one another in terms of performance or overclocking potential.

I have an Ex9 although I haven't done much testing with it yet... It's a nice board, but it does have a few quirks.

The Negatives:
1) As with any new board, the bios needs a little work. Sometimes it requires a couple times to get settings to stick in bios, especially CPU Multi.
2) Even with LLC set to 100% it still suffers from VDroop, not a big deal, but it does require you to set VCore a little higher to compensate.
3) The windows OC'ing tool is pretty bad... About the only thing it's good for is changing the HTRef/voltages (not uncommon with a lot of boards though )

The Positives:
1) the PWM section of the board seems strong and runs cool to the touch.
2) 2 USB3 headers are a nice feature
3) It has an Intel NIC

Like I said, I haven't pushed it yet... IMHO, the Asus C5F is a better board, but it's also a lot more expensive!
I think you'll like the Ex9, and I'm pretty sure you won't find a better board in the price range...

I know you have mad skills man! I'm actually glad to see you back on an AMD rig, maybe you can learn me somethin'.
I'm not sure about the stepping of your chip, but in the end, it's always luck of the draw regardless (Intel or AMD).

Most 8350's scale really well with voltage/cooling, but become power hog's over 1.45 Vcore (with diminishing returns).
With the H100 and a decent sample you should be able to hit the 4.5/4.7 range for 24/7 stable use while staying under 1.45v...

Other than benchmarks, Vishera's MT potential is decent! They're really nice for Video encoding (just one example).

I know I'm rambling, but here are a couple pics of my testing with the Ex9:
This one is shot of the rig, air cooled, using a Xigmatek Thor's Hammer @ 75f:



Here's a shot of the clocks achieved with this setup/ambient temps (cooling obviously not up to the task ):


Power consumption under load (Prime Blend) was ~350w. Actual CPU VCore was set @ 1.475 in bios, so you can see the VDroop acording to CPU-z...

[I.nfraR.ed]

2) Even with LLC set to 100% it still suffers from VDroop, not a big deal, but it does require you to set VCore a little higher to compensate.

Not sure how it is on the Extreme9, but on other AsRock boards the "Asus 100% LLC" corresponds to LLC disabled on Asrock - quite the opposite you'd expect.

[AMDforME]

^^^

FWIW, it appears that Asrock label's their LLC backwards compared to how it's suppose to work. My extensive testing of Vcore voltages with and without LLC shows that "25%" works the best and maintains the Vcore slightly above the set voltage when under load. 50%-75%-100% all lower the Vcore progressively. When "auto" mode is selected the Vcore can skyrocket way beyond what it's set to or should be allowed. This has been my experience and that of others using the Asrock 990FX Fatality Pro which has a very similar 12-phase VRM design to the E9.

It's worth noting that unlike other CPUs that I have seen, IME the FX processors will become unstable if you use too much Vcore or CPU-NB voltage even when the increased voltage does not cause the CPU to run hotter than normal. There seems to be a Vcore/CPU-NB voltage combination that works best and more voltage does not improve the OC'ing capabilities and actually reduces stability - once you find the optimum combo.

[EnJoY]

The Asrock Extreme9 is quite similar to the Asrock 990FX Fatality Pro mobo with minor PCI/PCIe slot differences. There isn't much difference in any mobos now days with one chipset maker per CPU brand and with the memory controller now within the CPU. The mobo VRM and features are what tends to differentiate the products. A proper VRM design with a good heatsink tends to allow overclocking operation without chronic mobo failures. An highly overclocked FX-8350 draws a lot of power and the E9/Fatality Pro seem to be up to the task, IME. Testing has shown that there are a lot of BIOS adjustments on some models that serve no useful purpose other than for marketing hype.

I agree, there hasn't been a significant difference in quite a while. You could even say the big differences stopped with the nForce 3 chipset and A64's. Most differences are VRM and firmware related.

I have an Ex9 although I haven't done much testing with it yet... It's a nice board, but it does have a few quirks.

The Negatives:
1) As with any new board, the bios needs a little work. Sometimes it requires a couple times to get settings to stick in bios, especially CPU Multi.
2) Even with LLC set to 100% it still suffers from VDroop, not a big deal, but it does require you to set VCore a little higher to compensate.
3) The windows OC'ing tool is pretty bad... About the only thing it's good for is changing the HTRef/voltages (not uncommon with a lot of boards though )

The Positives:
1) the PWM section of the board seems strong and runs cool to the touch.
2) 2 USB3 headers are a nice feature
3) It has an Intel NIC

Like I said, I haven't pushed it yet... IMHO, the Asus C5F is a better board, but it's also a lot more expensive!
I think you'll like the Ex9, and I'm pretty sure you won't find a better board in the price range...

I know you have mad skills man! I'm actually glad to see you back on an AMD rig, maybe you can learn me somethin'.
I'm not sure about the stepping of your chip, but in the end, it's always luck of the draw regardless (Intel or AMD).

Most 8350's scale really well with voltage/cooling, but become power hog's over 1.45 Vcore (with diminishing returns).
With the H100 and a decent sample you should be able to hit the 4.5/4.7 range for 24/7 stable use while staying under 1.45v...

Other than benchmarks, Vishera's MT potential is decent! They're really nice for Video encoding (just one example).

I know I'm rambling, but here are a couple pics of my testing with the Ex9:
This one is shot of the rig, air cooled, using a Xigmatek Thor's Hammer @ 75f:

Here's a shot of the clocks achieved with this setup/ambient temps (cooling obviously not up to the task ):

Power consumption under load (Prime Blend) was ~350w. Actual CPU VCore was set @ 1.475 in bios, so you can see the VDroop acording to CPU-z...

Thanks very much for all of the info here Daveburt, some good insight. I should receive my board on Friday and if I do, I'll be playing with it on Saturday and then again Sunday night.

Btw, I don't know where you heard of my "mad skills" but I have no such thing. I've been away from the scene for awhile.

[FlanK3r]

5350 MHz my max ambient Cinebench with 1.57V. No problem stable at 5 GHz 1.525V (AOD stability or OCCT)

[BlindShot]

Hey Guys,

I decided to sell my Core i7 3930K, ASUS RIVE and GTX680 4GB FTW+ and downgrade to an FX 8350, ASROCK Extreme9 990FX and HD7950 that I got a great deal on. I'm saving $1,000 in this swap, so it was really a no-brainer for me.

I'll be carrying over my 4x4GB G.Skill 2133MHz Ripjaws and Corsair H100 with me, as well as my Seasonic 1KW Platinum.

I wanted to get some impressions on the ASROCK. I haven't seen much talk about it because the Extreme9 seems to be fairly new, but the specs looked very solid and the price was better than anything in the same league as it. I'm somewhat of the belief that 990FX boards don't have much to differentiate themselves from one another in terms of performance or overclocking potential.

Anyhow, I will post my results with the chip and board once I get them humming. Hoping I can peek 5GHz, but I don't do enough on my system anymore to require that for 24/7.
as far
EDIT: Got the chip today, stepping: FA 1252PGT - batch: 0507

No love for the 3930k? I made the switch to x79 and now I'm clueless at overclocking lol.

[Daveburt714]

InfraR.ed & AMDforME: Thanks for the tips folks! This is my first ASRock board....

I did a little more testing tonight and apparently LLC is not working at all on this board.
(latest bios v1.10)

I'm using CPU-z to monitor voltages and after testing every LLC setting I came up with the same results:
With multi @ 22.5 and VCore set to 1.4625 in bios the resulting voltages were the same regardless of LLC settings (including disabled).
Idle voltage was 1.44v and load voltage was 1.424v. Power option in Win7 was set to "High Performance" and Turbo Mode disabled in bios.

Not sure how they are with bios releases, hopefully there will be a new one soon...

[AMDforME]

^^^ Feel free to e-mail Asrock Tech Support with your findings. That is how you get quick BIOS updates. It takes quite a few BIOS updates to address the many issues that exist these days. That however doesn't explain why LLC is not functioning properly or at all on your mobo.

To eliminate all possible voltage adjusting options with the FX processors, you need to disable:

1. Cool and Quiet
2. Turbo Core Mode
3. Application Power Management
4. CPU Thermal Throttle - if you don't plan to run near the thermal limit
5. C6 mode
6. C1E

Not all BIOS will offer these settings but most should even though the names might vary slightly.

Setting the fan to "max" or 100% is also required - as you did.

From personal experience with Asrock BIOS, I also suggest that you have the latest two BIOS for your mobo on a flashdrive. This is because when you are looking for the maximum operating frequency for the CPU or RAM and you get a lock-up or BSOD, the BIOS can go into a "stupid loop" where no matter what settings you make in the BIOS, it uses the default settings. The only cure for the "stupid loop" is to install another BIOS version then after it is confimed good and it boots to Windoze without issue, go back and install the latest BIOS version, again. Gigabytes dual-BIOS prevented me from every having to do this, but each mobo maker has it's own little quirks IME.

I also found that in the 990FX Fatality Pro BIOS that the CPU-NB frequency reverted to the default 2200 MHz. even when set higher, when you run the RAM @ 2133+ MHz. Asrock has a beta BIOS to fix this flaw.

[EnJoY]

Got the system all setup guys. My chip's VID is 1.375v with all power saving features, apm and turbo core disabled.

I'm going to do some preliminary CPU multiplier overclocking tonight.

[EnJoY]

InfraR.ed & AMDforME: Thanks for the tips folks! This is my first ASRock board....

I did a little more testing tonight and apparently LLC is not working at all on this board.
(latest bios v1.10)

I'm using CPU-z to monitor voltages and after testing every LLC setting I came up with the same results:
With multi @ 22.5 and VCore set to 1.4625 in bios the resulting voltages were the same regardless of LLC settings (including disabled).
Idle voltage was 1.44v and load voltage was 1.424v. Power option in Win7 was set to "High Performance" and Turbo Mode disabled in bios.

Not sure how they are with bios releases, hopefully there will be a new one soon...

I was experiencing similar issues with LLC not working properly after flashing BIOS 1.10. I cleared CMOS and now LLC is working just fine.

Leaving LLC disabled with overvolt by 1-1.25v under load, so be very careful with this! With LLC at 100%, it will droop by ~.25v. I set it to 75% and now it doesn't budge under load from what it is set too (perfect).

Btw, I am getting temps of 47c core, 68c CPUTIN (socket?) running Prime95 small FFT at 4.3GHz, 1.4v under Corsair H100 with stock fans. Seems a little high, no? Room is average room temp, but I don't have an accurate ambient measurement.

[AMDforME]

Enjoy- If your ambient room temp is around 68F/20C, then 47C core temp with an H100 is quite reasonable as you're over-volting 8-cores, which ups the power consumption and thermal load significantly. As you up the frequency more or increase the CPU-NB some, the core temp will also likely creep upwards. The FX/Vishera 8-core CPUs are rated to run 61C max. for 24/7 so your temps are fine.

[Daveburt714]

I was experiencing similar issues with LLC not working properly after flashing BIOS 1.10. I cleared CMOS and now LLC is working just fine.

I could have sworn I reset bios after the flash to 1.10.... Anyway, I did it tonight and it did help a little, but LLC still acts wonky compared to the C5F.

We're both pretty new to this board, so I thought I'd offer a suggestion that did seem to help!
After massive testing on the Asus C5F, bumping the IMC (CPU-NB) up helped stability a lot (currently running @ 2.5 1.275v on that rig (VERY Stable!)).

Even though the default IMC voltage on the Ex9 was 1.875, I took a page from my C5F experience and bumped that voltage up to 1.225.
For whatever reason, doing that also made CPU LLC work better and allowed me to lower Vcore...

It might be something you want to try on your rig!

[AMDforME]

^^^ IME there is a very specific combination of vcore and CPU-NB that works well for each CPU and it's not necessarily consistent. Sometimes bumping one or the other voltage compensates for the other but eventually you can hone in on the best combination.

I do think Asrock has some issues with how they label their LLC as it's not the way it's suppose to work.

[EnJoY]

I performed some preliminary testing on the CPU last night. The results don't look so good.

With VCore set to stock (1.375v) and LLC set to 75% so that load stays solid at what is set in BIOS, I can get 4.3GHz stable. 4.4GHz, requires 1.4v, and it's very downhill from there. 4.5GHz is Cinebench stable at 1.4375v, but failed prime after 4 hours. I haven't tried higher yet, but if the pattern stays true, I am looking at just 1.48+ vcore to get 4.6GHz stable.

My temps were hitting 60c on the cores at 4.5GHz with 1.4375v, so my cooling is already seemingly at capacity with this heat load. I would definitely need beefier fans to push higher.

Furthermore, unless I substantially improved my cooling to a high-end water cooling setup or phase change, I am not going to get better returns for my voltage increases, and therefore I'm already at the point where I'm not sure it's worth the extra heat and power consumption to go further. I may even have to settle in at 4.4GHz 1.4v, because there's a 10c difference in core temp between that and 4.5ghz 1.45v.

I'll fiddle around with it more of course, but obviously this is a bit disappointing.

[AMDforME]

^^^ You may want to up the CPU-NB to ~1.30v, the HT to ~1.25v and lower the vcore a bit and see if that makes it easier to get 4.5-4.6 GHz. stable. Make sure the RAM is also a little higher than the default voltage. IME too high of voltages on the Vishera CPUs can make the system unstable even if heat is not an issue. You can also use F-Stream to see if the VRM circuit is thermally cutting phases and vcore after an hour or so of P95. While it's not suppose to cause errors, it can under stress testing with OC'ed CPUs, IME.

An H100 is only the equivalent of a high end HSF so your temps are not unusual IME.

[Jethro]

Prime 95 sucks for FX cpu's anymore. I would suggest OCCT. 60c at 1.43 isnt gonna net great clocks IMO. I've had alot of FX's and they all flake for me at high 50's~60c. If you really want 4.8+ I would recommend water.

[demonkevy666]

Why not use turbo ?

4.4ghz on all core
4.6ghz on all cores for 4 core load ( one core loaded in each module)
4.8ghz+ on a module for single core/dual core in module loads ?