new AMD Overdrive video's on youtube

Watch Macci in action with Phenom at 3GHZ.

http://www.youtube.com/watch?v=i_q_9a4zDps

http://www.youtube.com/watch?v=1p-Vi...eature=related

Wow! :eek: :eek: :eek:
I will pray to have this feature for linux
good vids, thanks for sharing.

Look at the screens, this is 3dMark05.

Will it work for my Q6600 and X1800 card ^^ ... HAHA

Jokes aside! Seems good for AMD owners "maybee ill become one :/"

sorry looked like 06 to me...ohh well its still some cool vids showing the overdrive utility.

Edited the first post to remove reference to 06

it is 2006, but the resolution is at 1024 by 768, you can zoom in after the end of 3dmark. its 06 because it shows the CPU score as well.

Well the first movie is edited when he runs the test on 3000mhz prosessor. He is actually using the clip from 7 days test startup after he pulled START.... so i wuld like to see that prosessor on air running that stress test for 7 days ^^.

Anyone looking for a new toaster??

Well AMD have Macci,and nvidia have k|ngp|n. :)

Cool vid.

Quote:

---

Originally Posted by revogamer

it is 2006, but the resolution is at 1024 by 768, you can zoom in after the end of 3dmark. its 06 because it shows the CPU score as well.

---

CpuTest 1 / CpuTest 2 and Deep Freeze is missing during the run of the Test. So it should bei 3dMark05.

theres no way thats 05. 20k 05 with a 3ghz phenom and 3870 crossfire?

05 has a cpu score... if you pause the video when it zooms in on the screen, you could clearly make out 3dmark05 business edition...

omg this might move me to ATi!

Quote:

---

Originally Posted by afireinside

theres no way thats 05. 20k 05 with a 3ghz phenom and 3870 crossfire?

---

Its not at 3GHZ for that score, there is no way auto overdrive will clock that high.

Its probably more like 200mhz OC on the CPU and a small clock on the video cards

Quote:

---

Originally Posted by afireinside

theres no way thats 05. 20k 05 with a 3ghz phenom and 3870 crossfire?

---

So 3GHz Phenom and 2 3870 will beat Yorkfield @ 4 GHz and 2 3870 ? Well i realy wish it would be so.

Maybe its just a lower clocked Phenom.

Its 3dmark05 business edition
you can see it at "02.14s" ;)

yep it is 05, my mistake, one big give away is in firefly forest, in 06 theres a purple and a green firefly, in 05 its only a green one, which is what this has

you can tell an experienced oveclocker had a lot to do with this utility

awesome work macci :up:

http://i22.photobucket.com/albums/b3...itled12121.jpg

i want.....that....tool.....

nvm :wasntme:

Nice videos :)

Lol?

Very smart of AMD to start on 06 and end on 05 in an effort to show the huge performance gains :rolleyes:

Quote:

---

Originally Posted by Faktor

Lol?

Very smart of AMD to start on 06 and end on 05 in an effort to show the huge performance gains :rolleyes:

---

yeah i thought how the hell did the gain so much from a mild OC lol :D

nice observation there and very misleading lol

Really bizarre actually. Since 20K 3DMark06 is possible with 2 3870's...

Some more BS actions done by AMD... When will they use their brain and build a competitive product instead of trying to BS their custumer? Is it that hard... Seriously

Look closer, both are 2005. Besides, do you really think Macci would be so stupid as to put his reputation on the line doing something like that?

You all need to put the pipe down and step back. :rolleyes:

man my eyes suck i can't tell which versions they are
gotta give the man benefit of the doubt heh

actually when it races through the bench it looks like '05 for both ;) :)

Quote:

---

Originally Posted by dinos22

man my eyes suck i can't tell which versions they are
gotta give the man benefit of the doubt heh

actually when it races through the bench it looks like '05 for both ;) :)

---

Yeah I noticed that the second time I watched it.

Yea its 05 for both runs (of course!) =)

Quote:

---

theres no way thats 05. 20k 05 with a 3ghz phenom and 3870 crossfire?

---

I dont think the CPU clock was visible in the 20k run... But feel free to check the ORB for 20k scores, the lowest CPU MHz to break 20k is 3150MHz right now (not saying it could not be done with lower clocks but that should give an idea about the requirements for 20k).

that is a serious stretch to think they run 2 different 3dmark's...

also, that's almost a 50% score jump an unknown overclock that was done using the auto-overclock feature of Overdrive

makes you wonder what the overclock was ;)

Very exciting application!

I'm surprised you ever tested for stability, who does that? :cool:

good to see macci again, but damn, who made that videoclip? it really sucks, :P
the only good parts are when macci is talking and using the tool, all the "somewhere in a secret lab" and the car tuning stuff was painfull :lol:

about the tool, great looks, couldnt be better, great featureset, could hardly be better, the benchmark and stability test is great too... but the tool is slow and buggy... it should reset all msr values when you reboot which it doesnt, this is not only annoying but could be dangerous too.
if it would work faster and only load the different parts of code once you click on the tab of the bench or stability test for example, i think it would be faster and better to work with.

Quote:

---

Originally Posted by macci

Yea its 05 for both runs (of course!) =)


I dont think the CPU clock was visible in the 20k run... But feel free to check the ORB for 20k scores, the lowest CPU MHz to break 20k is 3150MHz right now (not saying it could not be done with lower clocks but that should give an idea about the requirements for 20k).

---

The CPU clock showed 3010mhz I think....

http://thumbnails2.imagebam.com/91/b51e44905615.gif

Heh some old style bench from it :)
Yep can't do nothing in it , all i can is use system stability, general info on hardware and this benchmark.
Nice tool.

This app should be nice. It allows independent core clocking, so if you have one dud core, you can probably scale the other chips higher. So this should take out the equation of having to overclock 4 really good cores at one. But I do wonder if this will cause issues in programs where you have 4 cores running a different clock frequency on a a heavy multithreaded application.

Can I run that stability tester on my Intel system :p:

Quote:

---

Originally Posted by kiwi

Can I run that stability tester on my Intel system :p:

---

if(CPU == Intel)
format drive:c

:D

damn that application looks dead handy..

Quote:

---

Originally Posted by SP33DFR34K

This app should be nice. It allows independent core clocking, so if you have one dud core, you can probably scale the other chips higher. So this should take out the equation of having to overclock 4 really good cores at one. But I do wonder if this will cause issues in programs where you have 4 cores running a different clock frequency on a a heavy multithreaded application.

---

You would get a performance rollercoaster in games when Windows changes the core a thread runs on now and then.

But ye, stability is a whole other issue aswell for any software timings.

Quote:

---

Originally Posted by SP33DFR34K

if(CPU == Intel)
format drive:c

:D

---

if (CPU == Intel)
return theBombHasBeenPlanted;

:D

Quote:

---

Originally Posted by Shintai

You would get a performance rollercoaster in games when Windows changes the core a thread runs on now and then.

But ye, stability is a whole other issue aswell for any software timings.

---

There's a thing called CPU affinity... not that you didn't know, you just played dumb because of your agenda... the same agenda that lets you see the future and conclude that it will be a "performance rollercoaster".

Quote:

---

Originally Posted by DoubleZero

There's a thing called CPU affinity... not that you didn't know, you just played dumb because of your agenda... the same agenda that lets you see the future and conclude that it will be a "performance rollercoaster".

---

And with CPU affinity...how do you run your multithreaded games again? By deselcting cores for...less overall performance? Also you cant control any spawned childprocesses in a game.

Perhaps you should do soem REALLY BASIC RESEARCH before spreeing out BS.

And currently I have been right, and you have been blatently wrong. So history aint exactly helping you there either.

Quote:

---

Originally Posted by Warboy

i want.....that....tool.....

---

Quote:

---

AMD Overdrive 2.10

The AMD OverDrive™ utility provides high-end users the ability to maximize the capability, flexibility, and adjustability of the platforms utilizing AMD processors, sockets, and chipsets. Hardware parameters may be controlled manually, or allow AMD OverDrive to automatically tune the system. Stability tests can be performed, once a system is modified, to verify the system and conduct a performance test to measure the impact of the modification.


supports the following chipsets only:

RS740(770)

RS780(780)

RD790(790X & 790FX)

Download>> http://www.techpowerup.com/downloads...rive_2.10.html

---

:up:

I don't have enough background in multithreading yet, but it should be an issue of concern. Something tells me that in heavy multithreading application, the slower core will bottleneck the other cores.

Thanks for the clip, did you realize they used and Phenom FX (1:17min).
Seems they release an FX for AM2 and Socket F.

Other "source"

http://www.youtube.com/watch?v=1p-Vi...eature=related

But what fun is overclocking with a tool like this. :shrug:

EDIT: Also not so cool if everyone runs his machine on an unacceptable speed vs. energy consumption level.


As i just crippled my AM2 mobo Asus (M2A-VM HDMI) with my first oc approches i preordered the low budget Asus M3A
for 66€. Hope a 9500 will ab available for ~200€ in Q1 08.

waiting for macci LN2 run... :D

Quote:

---

Originally Posted by SP33DFR34K

I don't have enough background in multithreading yet, but it should be an issue of concern. Something tells me that in heavy multithreading application, the slower core will bottleneck the other cores.

---

The best and most simple way to "calculate" the speedup from variable core OCs is to take the arithmetic mean of all frequencies of the cores.Yeah,it sounds simple and "dumb",but that's closest you will get when calculating the OC result in some multicore aware app.In monothread app. it depends on which core you run it,so you can bind it to the fastest core/OC and that way you will insure the best performance.
From available data,K10's perf. seems to be somewhat limited by ram bandwidth and Northbridge(L3) clock as well.The L3 runs at measly 1600Mhz/1800Mhz and we should see some decent OCs on L3 with retail Phenoms.I think tictac posted the way how ram speed is calculated and how Northbridge speed is calculated.

Quote:

---

Originally Posted by informal

The best and most simple way to "calculate" the speedup from variable core OCs is to take the arithmetic mean of all frequencies of the cores.Yeah,it sounds simple and "dumb",but that's closest you will get when calculating the OC result in some multicore aware app.In monothread app. it depends on which core you run it,so you can bind it to the fastest core/OC and that way you will insure the best performance.
From available data,K10's perf. seems to be somewhat limited by ram bandwidth and Northbridge(L3) clock as well.The L3 runs at measly 1600Mhz/1800Mhz and we should see some decent OCs on L3 with retail Phenoms.I think tictac posted the way how ram speed is calculated and how Northbridge speed is calculated.

---

You cant actually calculate it so easily. It also depends how much serialized code is being run. And how the multithreading is done.

If we very roughly say Core0 and 1 is 2Ghz, Core2 and 3 is 3GHz. Then 3 bad rough examples could be:

1. When you game and Windows flips the main thread from core to core now and then. Then you could have 20FPS one minute and 30FPS the other.
2. Say you render, and the render hard assign (in code) affinity in 4 spawned subthreads. Then you would have 2 fast renderings and 2 slow. Still waiting for the slow.
3. Running programs that still do software timings like mentalray etc. Then it would basicly just mess up to say it mildly. But it already does with CnQ or speedstep anyway.

However for something like WCG/Folding etc it would give a permant speedboost.

shintai, thats why every core has its own counter ;)
none of the things your worried about can happen afaik, except for some threads on some cores running faster than others, but if a game or app is programmed well itll know this from the counters and clockspeed.

Quote:

---

Originally Posted by Shintai

And with CPU affinity...how do you run your multithreaded games again? By deselcting cores for...less overall performance? Also you cant control any spawned childprocesses in a game.

Perhaps you should do soem REALLY BASIC RESEARCH before spreeing out BS.

And currently I have been right, and you have been blatently wrong. So history aint exactly helping you there either.

---

It's this a joke or are you being serious? lol

The point of overclocking just one core or two, is to get max performance for games\apps that don't use more then, guess what, one or two... lol
No one is going to have a core at 3.2ghz and the others at 2.0ghz to play games that use 4 cores, or would use that stupid scenario to
criticize this independent core overclocking feature...

And i don't even remember that i ever been wrong and you right, let alone lately... you're a joke.

Quote:

---

we very roughly say Core0 and 1 is 2Ghz, Core2 and 3 is 3GHz. Then 3 bad rough examples could be:

1. When you game and Windows flips the main thread from core to core now and then. Then you could have 20FPS one minute and 30FPS the other.

---

So playing a game and never going over the 20fps the 2.0ghz would give you it's better how?

Quote:

---

2. Say you render, and the render hard assign (in code) affinity in 4 spawned subthreads. Then you would have 2 fast renderings and 2 slow. Still waiting for the slow.

---

And this is bad how? With all the cores at the same speed wouldn't you still have to wait for the "slowest"? lol

Quote:

---

3. Running programs that still do software timings like mentalray etc. Then it would basicly just mess up to say it mildly. But it already does with CnQ or speedstep anyway.

---

Meh

Quote:

---

Originally Posted by saaya

shintai, thats why every core has its own counter ;)
none of the things your worried about can happen afaik, except for some threads on some cores running faster than others, but if a game or app is programmed well itll know this from the counters and clockspeed.

---

You cant code an app or game for it. Not even the OS scheduler knows it.

Maybe it could come in Windows Seven. But its not in Vista or anything below.

Quote:

---

Originally Posted by DoubleZero

The point of overclocking just one core or two, is to get max performance for games\apps that don't use more then, guess what, one or two... lol
No one is going to have a core at 3.2ghz and the others at 2.0ghz to play games that use 4 cores, or would use that stupid scenario to
criticize this independent core overclocking feature...

---

Anyone who's OC'ed multi core chips knows that one core will be the weak link. Technically, the other cores might have higher OC potential, if it wasn't for their twin brother holding them back.

What AMD is doing is giving us a way around that by letting us OC to the highest level on each core. So say, one core might max out at 3.4Ghz, and the other 3.3Ghz. You'll get the benefit of 100Mhz that you wouldn't normally get on a chip that had only one multiplier for all 4 cores.

Question is, what anomaly's might appear when you do asynchronous over clocking? The result might be performance that seems uneven, like an egg rolling end over end. Or the result might be that the faster core will end up waiting on the slower core, so your effective speed will still be only as good as the slowest core, unless you manually set affinity. No one will know until this ends up in the hands of John Q Public.

Quote:

---

Originally Posted by Rock&Roll

Anyone who's OC'ed multi core chips knows the one core will be the weak link. Technically, the other cores might have higher OC potential, if it wasn't for their twin brother holding them back.

What AMD is doing is giving us a way around that by letting us OC to the highest level on each core. So say, one core might max out at 3.4Ghz, and the other 3.3Ghz. You'll get the benefit of 100Mhz that you wouldn't normally get on a chip that had only one multiplier for all 4 cores.

Question is, what anomaly's might appear when you do asynchronous over clocking? The result might be performance that seems uneven, like an egg rolling end over end. Or the result might be that the faster core will end up waiting on the slower core, so your effective speed will still be only as good as the slowest core, unless you manually set affinity. No one will know until this ends up in the hands of John Q Public.

---

Correcto. But what I'm talking about is in heavy multithreaded application which uses all the cores possible. The workload would be distributed to all the cores evenly, and if one of the cores is slower, it'll take longer to finish executing the instruction. Having this one instruction a bit slower may not seem like a big deal at first, but if the other instructions in the core are dependent on each other to finish a certain clock cycle, this could become problematic. My best guess of how AMD got around this was probably the shared L3 cache and using it as storage for all 4 instructions. Then fetching the finished executed result from there. The other solution is likely to be stalling for the slower core, which would reduce the purpose of indepedent core overclocking. Fixing this solution by far is not an easy task.

Quote:

---

Originally Posted by macci

Yea its 05 for both runs (of course!) =)


I dont think the CPU clock was visible in the 20k run... But feel free to check the ORB for 20k scores, the lowest CPU MHz to break 20k is 3150MHz right now (not saying it could not be done with lower clocks but that should give an idea about the requirements for 20k).

---

GPU clocks for that run were?...... :)

Anyone else noticed that there will probably be an extreme edition of this soon? He uses multi 14 so his one at least has to be unlocked

Quote:

---

Originally Posted by Shintai

You cant code an app or game for it. Not even the OS scheduler knows it.

Maybe it could come in Windows Seven. But its not in Vista or anything below.

---

of course you can.
remember the TSC problem with dualcore cpus?
one core would skip every second clock and throttle since its barely being used, as a result the TSC would run out of sny and the games stuttered or crashed. once there were patches that synced the TSCs games and apps worked fine even if core2 is idel enough to trigger the clock skipping.

and in case you didnt know this on intel quads this is possible as well, you can clock 2 of the 4 cores at a different clockspeed by changing the multi, and no, games and apps dont crash or produce errors. :P

do you honestly think amd goes all the length to design a cpu where every core can be clocked independantly when doing so actually makes most apps crash and not work properly? :P

Quote:

---

Originally Posted by saaya

of course you can.
remember the TSC problem with dualcore cpus?
one core would skip every second clock and throttle since its barely being used, as a result the TSC would run out of sny and the games stuttered or crashed. once there were patches that synced the TSCs games and apps worked fine even if core2 is idel enough to trigger the clock skipping.

and in case you didnt know this on intel quads this is possible as well, you can clock 2 of the 4 cores at a different clockspeed by changing the multi, and no, games and apps dont crash or produce errors. :P

do you honestly think amd goes all the length to design a cpu where every core can be clocked independantly when doing so actually makes most apps crash and not work properly? :P

---

AMD updated its driver page with latest Phenom drivers :),similar to the ones used with X2s.

Quote:

---

Originally Posted by saaya

of course you can.
remember the TSC problem with dualcore cpus?
one core would skip every second clock and throttle since its barely being used, as a result the TSC would run out of sny and the games stuttered or crashed. once there were patches that synced the TSCs games and apps worked fine even if core2 is idel enough to trigger the clock skipping.

and in case you didnt know this on intel quads this is possible as well, you can clock 2 of the 4 cores at a different clockspeed by changing the multi, and no, games and apps dont crash or produce errors. :P

do you honestly think amd goes all the length to design a cpu where every core can be clocked independantly when doing so actually makes most apps crash and not work properly? :P

---

There is a difference with a hardware TSC bug and being able to dynamicly load cores after performance. There simply aint no such thing.
The affected issue you talk about got nothing to do with throttling. But how the software counter ran and was affected.

Also I bet you that the 2 speed core issue that was with early beta BIOSes didnt get tested much back then. But even then, you still cant make windows automaticly use those faster cores in multithreaded applications. it would still end up with the rollercoaster model.

But again, I didnt say every game and app would crash did I? We talk about the possible speed benefit of it and what could go wrong. Aka some certain apps that still use software TSC.

There is a reason why OEMs ship servers with disabled CnQ and Speedstep too you know.

Quote:

---

Originally Posted by SP33DFR34K

I don't have enough background in multithreading yet, but it should be an issue of concern. Something tells me that in heavy multithreading application, the slower core will bottleneck the other cores.

---

Not necessarily, take Cinebench for instance. It will continue on with it's rendering of an image even if one core has slowed down rendering part of the scene. You can clearly see this in the Cinebench benchmark. Any good multithreading program should be able to do this. Of course, it's easier said then done. :D

Quote:

---

Originally Posted by freeloader

Not necessarily, take Cinebench for instance. It will continue on with it's rendering of an image even if one core has slowed down rendering part of the scene. You can clearly see this in the Cinebench benchmark. Any good multithreading program should be able to do this. Of course, it's easier said then done. :D

---

Yeah, but that is with all the cores running the same speed. We haven't really seen any results from different clock frequency. Anytime you throw multiple clock frequency, things start to get more complicated. I'm sure AMD engineers aren't that stupid not to address this issue. We'll see in a couple of days whether or not this will be an issue. Until then, just speculation.

The day of truth for Phenom is tomorrow. Watch for leaks late tonight!