First look at Piledriver, how does it compare.

[wagex]

basicly what hokie is saying in plain english, is that board was a freak incodent that the LLC was the main contributing factor on that board and that board in particular.
and it only helped that board because the voltage regulation sucked to begin with (that board and that board only had crappy voltage regulation)
lamens terms
the LLC making higher overclocks only applies to that board because it sucks

 

[hokiealumnus]

Thanks wagex, that does sum it up nicely.

 

[EarthDog]

basicly what hokie is saying in plain english, is that board was a freak incodent that the LLC was the main contributing factor on that board and that board in particular.
and it only helped that board because the voltage regulation sucked to begin with (that board and that board only had crappy voltage regulation)
lamens terms
the LLC making higher overclocks only applies to that board because it sucks

I think you may have a board problem which is shaping/causing this misconception of LLC and what it REALLY does for you.

Glad we got this resolved...today.

 

[Hits9Nine]

Thanks wagex, that does sum it up nicely.

And we are suppose to believe what you guys say to eh! LOAD OF **** good team support!


Without a doubt, the Gigabyte 990FXA-UD7 Rev. 1.1 is a solid board. If you like the way it looks, or just like Gigabyte, no one could fault you for going this route.

- Jeremy Vaughan (hokiealumnus)

 

[Hits9Nine]

basicly what hokie is saying in plain english, is that board was a freak incodent that the LLC was the main contributing factor on that board and that board in particular.
and it only helped that board because the voltage regulation sucked to begin with (that board and that board only had crappy voltage regulation)
lamens terms
the LLC making higher overclocks only applies to that board because it sucks

And we are suppose to believe what you guys say to eh! LOAD OF **** good team support!


Without a doubt, the Gigabyte 990FXA-UD7 Rev. 1.1 is a solid board. If you like the way it looks, or just like Gigabyte, no one could fault you for going this route.

- Jeremy Vaughan (hokiealumnus)

 

[EarthDog]

Yeah, just speaking the truth, so of course you will get piling on (when you are not listening or reading the support we brought to the table). You are talking about one board to which we all acknolwedged has an issue...other boards do not. I looked at other 990FX reviews and they have not mentioned anything about massive vdroop issues like this board. Perhaps get a different board so you can experience LLC the way it was meant to be and hopefully that will help form a more proper idea on how it is supposed to work on the rest of the boards outside of gigabyte/that board.

 

[wagex]

yea well maybe it was an underlying problem with the board he didnt notice because he wasnt going for a global cpu-z validation.

done trolling yet?

 

[EarthDog]

Maybe. I understand where he is coming from... but what he doesnt seem to see is that its a problem with THAT board ONLY. He mentioned 990FX but linked an article to ONE board which we all acknowledged, and even the reviewer here noted, has a problem. Oh well, as they say, you can lead a horse to water. We cant force him to believe what is true... though we gave it a try.

Id still love to see a P95 stable 5Ghz 1.46v though...If you can prove it, looks like you have a buyer on your hands and can upgrade that board and get a new chip!

 

[hokiealumnus]

Wow, did that post seem like such a good point you had to say it twice? Feel better now?

Obviously you're not getting what we're saying. Try to separate yourself from what appears to be your high level of emotion and look at the situation not as a personal affront but rather people trying to help you understand something. There is no teamwork in this thread; no one enlisted anyone else to come make comments. Contrary to what your defensive posture may tell you, no one is ganging up on you.

If you could step back for a moment and perhaps it will help you realize that, when you are the only person saying one thing and everyone else is saying another, perhaps the fault doesn't lie with everyone else?

I still stand by what I said, no one can blame you for purchasing your board. It is a solid board, they did "fix" the problem that was going to fail rev. 1.0; it was a band-aid fix, but it's not too difficult to enable LLC, so I didn't think it deserved to be panned as an awful board. That doesn't mean all 990FX boards share the same voltage problem with LLC turned off that the UD7 does. It's not your position, which we all agree is a problem on the 990FXA-UD7; it's that you're applying it as a blanket application to the entire line of 990FX motherboards when combined with a Bulldozer CPU.

Before you fly off the handle again, I would encourage you to rethink doing so. Spirited discussion is fine, but when you start going with all caps and cursing, that can make a thread devolve quickly. I'd recommend refraining from being that catalyst.

 

[Hits9Nine]

Maybe. I understand where he is coming from... but what he doesnt seem to see is that its a problem with THAT board ONLY. He mentioned 990FX but linked an article to ONE board which we all acknowledged, and even the reviewer here noted, has a problem. Oh well, as they say, you can lead a horse to water. We cant force him to believe what is true... though we gave it a try.

Id still love to see a P95 stable 5Ghz 1.46v though...If you can prove it, looks like you have a buyer on your hands and can upgrade that board and get a new chip!

also problem with these boards too I have used
990fxa-gd80
970-g45
Asrock fatality 990
Asus Crosshair v

another users comment
"ASUS Crosshair V Formula AM3+ AMD 990FX SATA 6Gb/s USB 3.0 ATX AMD ..... Right now if you turn LLC off, the system suffers a MASSIVE V-droop"

....GOOGLE AM3+ massive vdroop its that easy!!!!


I think you need to get out more an maybe use Google, as there are a lot of AM3+ users out there stating the large Vdroop issues on all AM3+ boards.

Its not just this board thanks!

This is my last post ever on this site, I will never recommend you to anyone! You guys are too much and obviously don't listen to anyone of the 100's of google searched you can find by just typing

 

[hokiealumnus]

I've gotten out a bit. I also have a Crosshair V. It seems to work just fine. I still turn LLC on to keep idle & loaded voltage the same, but I don't recall ever having that Vdroop problem when reviewing Bulldozer.

Yes, Vdroop is big with LLC off. It is. No one said it isn't. That isn't the cause of the instability. It's the wavering of Vcore with LLC off on the UD7. Yes, all boards suffer from Vdroop. It's part of their design. X set voltage = X loaded voltage. It's on every board, AMD or Intel. Again, you're missing the point.

Anyway, at this point it's like we're talking to a brick wall. We type the words, perhaps you read them, perhaps not; but they definitely aren't getting through. Have fun in your future eneavors. We're sorry to see you go.

 

[EarthDog]

also problem with these boards too I have used
990fxa-gd80
970-g45
Asrock fatality 990
Asus Crosshair v

another users comment
"ASUS Crosshair V Formula AM3+ AMD 990FX SATA 6Gb/s USB 3.0 ATX AMD ..... Right now if you turn LLC off, the system suffers a MASSIVE V-droop"

....GOOGLE AM3+ massive vdroop its that easy!!!!


I think you need to get out more an maybe use Google, as there are a lot of AM3+ users out there stating the large Vdroop issues on all AM3+ boards.

Its not just this board thanks!

This is my last post ever on this site, I will never recommend you to anyone! You guys are too much and obviously don't listen to anyone of the 100's of google searched you can find by just typing

All boards have vdroop. But just the board you mentioned had a real problem where the voltage fluctuated without it enabled. Anyhoo, good luck at other websites.

EDIT: Crap, what hokie said.

 

[Frakk]

Why is vDroop the hot topic here?

I get vDroop underloaded, it's not huge, maybe as much as 100mv as a rough guess, i never bothered to measure it.

I just add some LLC, which not only fixes the vDroop is also increases the volts.

So i can set it at 1.325v for idle which is fine for normal use but if the CPU is under heavy load it fails.
I put a higher LLC and it will actually boost the volts to 1.335v once stressed, just by doing that i'm running lower volts when not needed and once it is the LLC kicks is and boosts the volts for when i need it at high stress automatically
LLC is a good tool, it can be used as a vCore Boost / Turbo thingy.... use it.

 

[BeepBeep2]

I really don't care what you think, or what your getting at. I don't need to prove anything to you! I know what I can do and what I have! Sorry I don't Prime again for 7 more hours. I have a job and family with kids that I have to tend to, and when I have time I spend it very efficiently with my PCs. So setting up my other water cooling system and running Prime95 for 7 more hours just to get you off my back is a big waste of my time since I don't have much time in the first place.
I don't believe you know what your talking about, you will have to prove it... now how about you prove it to me! To me your just some dude on the internet who thinks he knows what he is talking about...see how I can play this game too! No matter what you say to me I wont believe you either, so I think this conversation is done now. BYE

Sir, I talk to people every day that buy tens, twenties and thirty of these chips to overclock simply so they can find the best.

I've started a thread at XtremeSystems.org for our members there to add their information and highest overclocks, as well as temperatures and more, all so we can find the average clocks of the platform as well as what is normal for higher and lower than average.

You outreach higher than average by about 250 MHz, at a voltage used for average chips at 50-100 MHz under the average clock achieved.

Over twenty people replied to that thread, several of which are the people that buy 5,10,15,20 of the CPU at a time to find the best to overclock on liquid nitrogen cooling and achieve world records.

While I do not have such deep pockets as some of my friends, I'm definitely not an idiot.

No, it wouldn't reach the same overclock with LLC off on that board. That's why I said it was a band-aid to patch the problem, not a fix. They never fixed the larger problem in the first place - that their power section was incapable of keeping voltage constant. Instead they introduced the controller that had LLC and left the actual power delivery the same. The LLC chip controls voltage much better than the board does by default, allowing people to use it to smooth out the voltage output.

LLC should be there to control Vdroop (or eliminate it), it should not be there to fix a problem of voltage output fluctuating dramatically.

This statement has caused more trouble in this thread than it was designed to fix.

vdroop is not a problem, vdroop is a safety measure. When a VRM output voltage droops under load, (within a tolerance, of course) it is doing what it was designed to do. When we set up the VRM to offset this vdroop when there is more current draw, ...(by adding/removing resistors or using a variable resistor to alter a circuit, or making a circuit with different paths, which is physical, but the path changes when you enable/disable LLC in software which sets the controller)...you introduce something else into the behavior of the circuit, a transient overshoot, which is a small burst of higher-than-normal voltage when the load is first applied, which can go outside normal component working specifications.

Lets use caddi_daddi's quest for 4.5 GHz on his 955 BE for example.
He needed 1.644v in CPU-Z at load (and lower temps, but leave that out) to reach 4.5 GHz in Prime95 for 2 hours.
Without LLC enabled on a high setting on his board, his idle voltage would have needed to be about 1.775v.
See the problem (and idiocy) here ?
Well, when he enabled LLC, and set it to "Ultra High" or "Extreme" like he did, he only needed an idle voltage of about 1.62v, making it a bit safer.
However, for a split second, maybe even 1/250th of a second, when he applied a 100% load to his CPU, chances are, that his voltage spiked unmeasureably upwards of 1.8v.

See the problem here, too?

If you anyone is having trouble understanding why, then there is a good read here:
http://www.thetechrepository.com/showthread.php?t=126

 

[hokiealumnus]

My statement? I was saying Vdroop isn't the problem, it's the lack of consistent voltage delivery. If you look as his link to OCC, they showed an OCCT graph with voltage all over the place when LLC was off. It should droop and stay very near the same voltage while under load. That is normal, expected and designed Vdroop. Having Vcore fluctuating all over the place once drooped is not.

I understand how Vdroop works and why it exists; I had hoped that was evident at this point.

 

[BeepBeep2]

Okay, just making sure, but it started a large discussion about the "problem" of vdroop and how so many other boards have this "problem" and like wagex commented "the LLC making higher overclocks only applies to that board because it sucks" which you all agreed summed it up correctly with is absolutely wrong...

EDIT:
This?


This image is a little misleading due to the polling rate of the software and inaccuracy of software voltage monitoring, programs only read in steps of .012, ie 1.428, 1.440, 1.452, 1.464v because that is the resolution of the sensors used in the CPU/motherboards. If it is right in between, too it can jump back and forth.

Looks like it starts the tests at 1.464v, droops to 1.452v, then continues between 1.440v and 1.452v while dropping to 1.428v a few times.

I think OCCT might be polling 2 or 3 times and taking an average of those times before it moves on to a new data point too, which kinda skews things a bit.

Either way, if you look at the CPU usage, it is in somewhat inline with the mountains on the graph too, looks to me like the VRM is just drooping depending on the current draw, and 100% load doesn't always mean there is the same current draw because it is running different tests.

 

[hokiealumnus]

I think you're missing the context of the conversation. He was saying without LLC, it is not possible to unleash the full power of Bulldozer. We were trying (and failing) to explain to him that it's not LLC itself that unleashes anything. That when run at the same loaded voltage, with LLC on or off, the same overclock is achievable.

No one has said Vdroop was a problem except him. Vdroop is as it is supposed to be by design. The problem is the UD7's voltage output when LLC is off does not stay constant when loaded. So it is not a predictable, repeatable pattern of Set voltage - Vdroop = loaded Vcore. It's more like Set voltage - Vdroop = (loaded Vcore +/- 0.06V).

 

[BeepBeep2]

I think you're missing the context of the conversation. He was saying without LLC, it is not possible to unleash the full power of Bulldozer. We were trying (and failing) to explain to him that it's not LLC itself that unleashes anything. That when run at the same loaded voltage, with LLC on or off, the same overclock is achievable.

No one has said Vdroop was a problem except him. Vdroop is as it is supposed to be by design. The problem is the UD7's voltage output when LLC is off does not stay constant when loaded. So it is not a predictable, repeatable pattern of Set voltage - Vdroop = loaded Vcore. It's more like Set voltage - Vdroop = (loaded Vcore +/- 0.06V).

That is because vdroop depends largely on current draw.

Gigabyte uses a digital VRM in UD5 and UD7, so LLC may group the amount of vdroop vs current draw a bit more broadly than it would if the circuit was just responding to the current draw.

See my observation above...and don't get me wrong, I understand what the context of the conversation is but the way things were worded seemed to have caused some confusion (like wagex pointing out that its just the board that sucks which is why LLC helps an OC only for that board)

Noise and ripple or the idea of fluctuating voltage at a constant load I don't think is what we are seeing here, all because the load is not constant. OCCT doesn't just do one calculation over and over and over again, but does many different ones which change the amount of CPU stress and/or current draw even though it might always say 100% load.

 

[officedemon]

oh dear, ummmm hows Piledriver looking guys?

will the new chips be any better than FX 4170? guessing 10-15%?

 

[mjw21a]

Hmmmm, looks like a nice little upgrade for me.... I do kind of miss my old Thuban..... Just waiting for the fully fledged piledriver chips to be released so I can upgrade.

I think the important thing to remember is that Bulldozer is the very first ground up redesign AMD has done since the original Athlon. Things will continue to improve incrementally from here on though I suspect AMD will never get the performance crown back in terms of pure x86 performance. I do expect them to pull ahead with software that uses iGPU acceleration further down the track though how long that advantage will last I don't know. I've been doing a little reading on GPU acceleration recently and apparently it doesn't scale terribly well. Likewise raw x86 performance is expected to exceed that of GPU performance in the next 5-6 years as development appears to be progressing faster with that than with GPU's..... Sorry for the lack of links guys, I can't even remember where I read that..... Probably on the inquirer or the register a few months back