A little birdie told me...

[BeepBeep2]

There seems to be plenty of misunderstanding around the 8.722GHz run on FX-8370.

Unlike some of the news falsely claim not all of the 4 compute units (8 cores) were operating at the frequency of 8.722GHz.
I have never claimed otherwise and honestly no one except Christian Ney even bothered to ask me.

The other compute units were clocked down to a lower PState, yet they were neither clock or power gated but in full power state (effectively C0 state).
Clock and / or power gating the other compute units or cores is quite common now days but thats not something I do.
Just as shutting down 3/4 of the compute units of the processor is just not my thing.
The way I currently do it mimics the normal behavior of Turbo Core. During low loads the idling cores will be clocked down.

Thank you for the clarification. This is pretty much exactly what I thought was happening.

Also, thank you for your contributions to the community and your research. The information you're providing about the new Vishera models is quite intriguing... I think I may just buy one!

EDIT:
I'd like to add also that overclocking is a hobby for me and it's something I always look at with a light heart. I congratulate great talent and results and whenever I come up short I'm still proud that I accomplished something. I surely don't care if I'm 1st or 120th in some HWBOT ranking because I really only compete against myself. I wish all overclockers looked at it this way. Like Movieman said, it's how you play the game that matters the most...

[Daveburt714]

Hey Stilt....

As usual you offer some of the best AMD info on XS!
I'm no dummy, and I know there aren't any really good options on the AMD side of things.

My current chip is a 9370 that's stable @ 4.8Ghz/1.488v. MicroCenter has lowered the 9590 to $220...
I'm jonesing for a new AMD chip to play with, and since you mentioned that the 8370 has some (minor) revisions, do you think I'd be better off to wait until they have an 8370, or just go ahead and purchase the 9590 @ $220?

I'm a decent OC'er, but not a hardcore (HWBot/sub zero) guy. Either way, I know it's kind of silly and I'll be paying close to $1 per Mhz but I really need something new to play with!

If I could get 5ghz stable (on my serious) custom water, do you think I'd have a better chance with the 8370 or the 9590?
Maybe the 8370E would be the best option considering my cooling?

I know it's all luck of the draw bro... Just wondered what you would choose in my situation...

My cooling specs are slightly improved from whats in my sig, but it's pretty close.

Thanks Bud!
Dave

[The Stilt]

Dave, my personal preference is: FX-8370E > FX-8370 > FX-8350 > FX-8320 > FX-9370 > FX-9590.
The FX-8320E would probably go right after the FX-8370 models, however I cannot comment about it since I have not tested one personally.

Go for the FX-8370E if you have air or AIO cooling.
Thats the best model to do some tinkering with too

[vario]

@STILT
I know its kinda OTm but as you seem to know alot, any idea if better VRM efficiency is doable ? Or mainboard manufacturers just dont care because none seems interested in this parts effciency.I have platinum PSU and i find it weird that VRM in these days are so inefficient :-/.Are any manufacturers better than others ? Or one solution is better than other, what i am looking for when buying mobo to get decent vrm efficiency ?
Also, thank you for the NB tests, 2700 is more than any stable result on air ive seen really .Mine does 2600 but im not 100% sure about stability and run it day to day at 2400, also what test is best for NB stability ?

[The Stilt]

@STILT
I know its kinda OTm but as you seem to know alot, any idea if better VRM efficiency is doable ? Or mainboard manufacturers just dont care because none seems interested in this parts effciency.I have platinum PSU and i find it weird that VRM in these days are so inefficient :-/.Are any manufacturers better than others ? Or one solution is better than other, what i am looking for when buying mobo to get decent vrm efficiency ?
Also, thank you for the NB tests, 2700 is more than any stable result on air ive seen really .Mine does 2600 but im not 100% sure about stability and run it day to day at 2400, also what test is best for NB stability ?

The ODMs are most definitely interested in improving the efficiency, however they are still far more interested in reducing the cost.
The motherboard industry is rather unhealthy in terms of competition so everyone is trying to save every single penny possible.
And that means the newest and more advanced parts will not be implemented in the designs immediately.

This doesn't mean that there is no progress what so ever.
ASUS, Gigabyte and MSI have recently started to implement more advanced and efficient components in the VRM of their high-end boards.
ASUS has used fully digital VRMs even in their mainstream boards for couple years now and the other ODMs are following.
Gigabyte was the first one to use the state of the art IRF PowIRStage modules which combine all of the essential components to a single package (high & low fets + gate driver).
These solutions are highly efficient and with proper tuning can achieve efficiency up to 95%. ASUS and MSI have used the same modules in some of their ultra high-end designs.

Recently ASUS has used Texas Instruments PowerBlocks in their high-end designs. The PowerBlocks contain both high and low fets in the same package and can reach up to 93% efficiency.
Each PowerBlock cost around three times a traditional n-channel mosfet, however since each PowerBlock replaces three traditional mosfets (high + 2x low) the BOM will remain nearly the same.
I am sure the PowerBlocks will be used in most of the new designs regardless the targeted market segment.
ASUS has made some efforts in reducing the core loss of the inductors. The new ASUS BlackWing inductors along with the Trio "knuckle" inductors increase the overall efficiency by having lower core loss but they also
have higher maximum and saturation current ratings.

The area of peak efficiency of a traditional switching buck VRM circuit is rather narrow.
Usually the peak efficiency is achieved at 80-90% of the maximum load, while anything higher or lower results anything between little to significantly worse efficiency.
Because of this it is important to use the phase control features available on the advanced controllers. When the phase-shedding is enabled the unneeded phases will be disengaged
and only the required phases will remain active. This allows the VRM to have higher efficiency at low loads when all of the enabled phases are operating at their optimal load / efficiency region.

E.G.

Your motherboard has six phases, each designed to deliver up to 25A of current.
Your CPU consumes 155W of power when fully loaded but requires only 25W in idle with all of the power saving features enabled.
Lets say the core voltage is static 1.2V in both cases

Phase-shedding disabled

Load:

VRM TDC: 6 * 25 = 150A
Active phases: 6 (150A)
CPU PMax: 155W
CPU VDD: 1.2V
CPU IMax: 129A
VRM Load: 129 / 150 = 86% >> Peak efficiency

Idle:

VRM TDC: 6 * 25 = 150A
Active phases: 6 (150A)
CPU PMax: 25W
CPU VDD: 1.2V
CPU IMax: 20.8A
VRM Load: 20.8 / 150 = 13.9% >> Poor efficiency

Phase-shedding enabled

Load:

VRM TDC: 6 * 25 = 150A
Active phases: 6 (150A)
CPU PMax: 155W
CPU VDD: 1.2V
CPU IMax: 129A
VRM Load: 129 / 150 = 86% >> Peak efficiency

Idle:

VRM TDC: 6 * 25 = 150A
Active phases: 1 (25A)
CPU PMax: 25W
CPU VDD: 1.2V
CPU IMax: 20.8A
VRM Load: 20.8 / 25 = 83.2% >> Peak efficiency

So the short answer is: The efficiency can be improved significantly, it is only matter of cost the customers are willing to pay.

[FlanK3r]

Now Im busy, so not much comments from me today...So thanks Stilt for all information and I would kindly ask you, if you cna post some srenshots of results (air or liquid)? Did you tried with 837 superpi 1/32M with LN2?

[FlanK3r]

Hm, first test with review CPU. Only quick for today. 4800+ MHz R11.5 is benchable with 1.37x V (load), after I tried quick validation with 1.415V over 5300 MHz so easy. 5400 BSOD, more juice please . If Stilt has right, Im still in safe temps zone .

[d_mouse]

That looks like it might make a good cheap upgrade from my 6300, any idea when these go on sale?

[tbone8ty]

Hm, first test with review CPU. Only quick for today. 4800+ MHz R11.5 is benchable with 1.37x V (load), after I tried quick validation with 1.415V over 5300 MHz so easy. 5400 BSOD, more juice please . If Stilt has right, Im still in safe temps zone .

that voltage!!! what? you got a golden chip

[FlanK3r]

Im not sure now, its beginning and I have FX-9370 with similar voltage scaling. This 9370 is Ok up to 4900 MHZ with very low voltage, after dont more scaling. Now Im tetsing stability in AOD for 30 minutes, 4700 MHz (4800 MHz HW error), 1.392 V

[Daveburt714]

Dave, my personal preference is: FX-8370E > FX-8370 > FX-8350 > FX-8320 > FX-9370 > FX-9590.
The FX-8320E would probably go right after the FX-8370 models, however I cannot comment about it since I have not tested one personally.

Go for the FX-8370E if you have air or AIO cooling.
Thats the best model to do some tinkering with too

Thanks for the advice Stilt...

Between your preferences and what Flank3r has shown from the 8370E it's making my 9370 look weak!

My water cooling is quite a bit better than an AIO, so I'll be on the lookout for an 8370...
As much as I want something new to play with, this new revision looks really good, so I'll hold off on the reduced price 9590.
I've been so desperate for a new toy that I've pulled out an old X2 550 and my Crosshair 4 just to have something to play with.

It was an extra nice chip (in it's day) and was fully unlockable. Even though I've been loose with the V's, I still haven't been able to get it over 4.1 stable in 2X mode, so it's not filling my need for speed...

Thanks Again,
Dave

[FlanK3r]

there is my FX-9370: scaling better with lower voltage, but over 49xx MHz its hard get more without juice. Look into the album 9370 (Cinebenchs results with voltage)
https://www.flickr.com/photos/206989...7646803492191/

30 minutes OC stability test FX-8370E done: 4716 MHz with 1.4 V, now Im testing 4762 MHz with same voltage...

edit: 4762 MHz test pass

[FlanK3r]

Still Im in tetsing phase. I have bad feeling, this chip has not good all modules because crash in AOD with HW failure, but at CU0 and Superpi much higher or valdiation its without problems. We will see tomorow...There is validation for Friday night . nothing special, but not bad too.

http://valid.canardpc.com/2twssm

[cab]

Honestly I think they're all the same chip. Towards the end of production they have unused stock they want to get rid of, and they have both high-leakage and low-leakage wafers. I bet xxxxEs are OEM stock.

[The Stilt]

I've been using these settings for two days now.
I have been playing some Arma 3, encoding some Blu-ray episodes to MKV with X264, doing some general browsing, watching Youtube and all the other regular stuff... without a single issue.

The only thing I cannot do is the burn-in tests such as Prime95 or Linpack.
The power dissipation they result is just too much for the cooling I am using.
Everything is fine until the temperature (tCase) reaches 64 degree celsius, at that temperature either the weakest CU produce errors or the system simply crashes.
During X264 or Blender the temperature peaks at 60-62 degree celsius so there is plenty room to spare
During Arma 3 or Hitman Absolution the temperature go no higher than 56 degree or so.

I'm running on Cryorig R1 Ultimate, with no other mods than an additional 80mm fan blowing directly to the VRM heatsink.
The VRM is operating at fully stock & efficient settings, only load-line has been adjusted to 0.000325ohms (-75%).

Must admit that I am quite impressed about the Cryorig.
They came basically from no-where and yet they immediately deliver a heatsink that is able to do 0.188C/W with 1300rpm fans attached :O
The chip draws 202W (DCR) during X264 for example.

The NB behaves quite differently compared to the older revision.
At stock voltage it can't do even 2400/2400MHz (NCLK/MEMCLK), and increasing the voltage by 50mV does absolutely nothing about that.
However when you ram up the voltage up to 100mV (= 1.275V) it starts flying. 2700/2400MHz is not a problem.

The problem appears to be the first L3 subcache (CU0) which requires the additional voltage but also prevents it from going higher than 2700MHz stable.

I'm not going to "double time" you, this is a cherry picked part (yet likely not the best).
I had several target characteristics for the screening, however the silicon itself is the very same as available in retail.

[I.nfraR.ed]

I'm running my 8350 like this 24/7 and it never failed on me for the tasks I do, but it does not pass Prime95. One of the CUs is weaker and stops short after starting P95.
It's a 8.2+GHz valid chip, but nothing close to your chip.
However it never failed on whatever game and bench I throw at it. I also compile under linux x64 on all cores and it passes that test as well.

However, I'm with a full-cover block for the mobo and a custom water-cooling setup with a triple PA140.3 rad with low rpm silent fans. One of the best (if not the best) triple radiators.

[FlanK3r]

My sample is second best from my FX colection. In validation. In all stress is similar as my 8200 MHz 8350s and as 9590. Maybe slightly better, but not better than 9590 ES. This one is still +200 MHz better. But its diferent story.

There is yesterday validation at CU2 (other cores are at lower p-state)
http://valid.canardpc.com/kfb2jp

[IronBalls]

Hey Stilt....

As usual you offer some of the best AMD info on XS!
I'm no dummy, and I know there aren't any really good options on the AMD side of things.

My current chip is a 9370 that's stable @ 4.8Ghz/1.488v. MicroCenter has lowered the 9590 to $220...
I'm jonesing for a new AMD chip to play with, and since you mentioned that the 8370 has some (minor) revisions, do you think I'd be better off to wait until they have an 8370, or just go ahead and purchase the 9590 @ $220?

I'm a decent OC'er, but not a hardcore (HWBot/sub zero) guy. Either way, I know it's kind of silly and I'll be paying close to $1 per Mhz but I really need something new to play with!

If I could get 5ghz stable (on my serious) custom water, do you think I'd have a better chance with the 8370 or the 9590?
Maybe the 8370E would be the best option considering my cooling?

I know it's all luck of the draw bro... Just wondered what you would choose in my situation...

My cooling specs are slightly improved from whats in my sig, but it's pretty close.

Thanks Bud!
Dave

It's only a question of luck. I don't think you are better of with the 8370E for overclocking than any other sharing the same number of cores.
I bought a FX 8320 and it ended up being a good overclocker. I only need 1.4625v for it to be rock stable @ 5ghz for 8hrs OCCT with high summer temps.
So my advice would be, go for the least expensive and cross your fingers

[FlanK3r]

Ironballs, tell more about your chip, this could be monster. What is your max Cinebench run?

[FlanK3r]

My review of FX-8370E, great OC
FX-8370E, good 95 W Vishera







more in review

[IronBalls]

Ironballs, tell more about your chip, this could be monster. What is your max Cinebench run?

Well my chip scales really well until 5 ghz. I can get 5.1 ghz 1 hour OCCT stable @ 1.5125v but everything above will give me less stability. Dont know if it is the chip, my psu or something else " causing this wall " . Even with 1.575v I cant run Cinebench higher than 5.3 ghz. Perhaps with better temps ( I mean in winter with room temps of around 18-20 C ) or with more powerful psu.

[IronBalls]

Max validation :

http://valid.x86.fr/mpzrd1

5418 Mhz @ 1.55v

Higher voltage results in an instant freeze. With a few tweaking, lowering NB, I could probably get 5.5 ghz validation. Room temperature is 23-24C. Do not know if FX scales good with lower temps ...

[Daveburt714]

I'm running my 8350 like this 24/7 and it never failed on me for the tasks I do, but it does not pass Prime95. One of the CUs is weaker and stops short after starting P95.
It's a 8.2+GHz valid chip, but nothing close to your chip.
However it never failed on whatever game and bench I throw at it. I also compile under linux x64 on all cores and it passes that test as well.

However, I'm with a full-cover block for the mobo and a custom water-cooling setup with a triple PA140.3 rad with low rpm silent fans. One of the best (if not the best) triple radiators.

Wow, those are impressive clocks using an FX chip and a CH4F board Ivan...
I've sold a bunch of my test gear, including an ASRock-EX9, so now my test rig is limited to a CH4F.
Good to know it's capable of hitting nice clocks with FX chips!

It may seem a little "shady", but my local MicroCenter has the 8320E (and the 9590) in stock.
I'm seriously considering picking up the 8320E, testing it, and if it doesn't clock well returning it for the 9590...
I've spent tons of money there, and they have a 14 day return policy.

Normally, I don't return stuff, but I really need a new AMD chip to tinker with!
If the 8320E doesn't clock that well, they might have an 8370E in stock before the 14 days are up...

I know it's a pretty cra*py thing to do, but I really need a new FX chip to play with, and they've gotten ALOT of my money over the years!

I may have 2'nd thoughts and change my mind... But right now I'm seriously considering picking up an 8320E tomorrow...