Overclocking Preview Bulldozer 8150

[psolord]

Agreed. I would totally buy the 8core BD if its OC'ed self was similar to how a stock 2600k performs.

Our friends above were referring to the quad core Zambezi which will probably be half the price of the 2600k.

If the 8core needs to be overclocked to be able to reach 2600k performance, it wouldn't be too good.

 

[aigomorla]

Our friends above were referring to the quad core Zambezi which will probably be half the price of the 2600k.

which is why im hoping for the X6 to drop in price if ANYTHING from bulldozer launch.

And im going to upgrade my friends rig to a X6 and call it a day.

 

[Neutralman]

Our friends above were referring to the quad core Zambezi which will probably be half the price of the 2600k.

If the 8core needs to be overclocked to be able to reach 2600k performance, it wouldn't be too good.

Oooh I see. I'd still be happy if theoced 8core is anywhere close to the 2600k. I've been holding off a new build waiting for bulldozer and amd 7xxx series, so excited.

 

[LOL_Wut_Axel]

What about it? 1.5V on 32nm SB and 4.8ghz is horrible. We have seen people achieve that on our forum with lower volts.

You can't compare them directly, and you probably never will. For K10.5, 1.5V is the max you want to go before you do some very major degradation over time. It's like 1.4V on Sandy Bridge.

AMD has been more tolerant of voltage than Intel for quite some time now, not to mention that while they're on the same process node, IIRC Intel is using high-k while AMD is using SOI.

You can't make an objective observation comparing the voltage Intel and AMD need because they have different ranges and tolerances. While 1.4V is the max you want for Sandy Bridge, for K10.5 that's as easy as cake.

 

[Cookie Monster]

My bad, I didn't know that "extreme overclockers" get this defensive and personal about overclocking with one module and 1/2 the L3 cache disabled.

"Hot Hardware was on hand for the event and claims that the CPUs used for testing were early AMD FX-8150 processors, and that only two of the eight cores were enabled for the record-setting run."

There not being defensive nor personal, but rather annoyed because some people seem to be ignorant enough to not understand what they were trying to achieve. Plus when they do this kind of overclocking, they tend to disable cores, use 1 stick of ram and keep everything else to a bare minimum to attain maximum cpu clock frequencies.

I was not impressed by an 8.3ghz Celeron LN2 that gets destroyed in performance by a 4.0ghz air cooled X6 1100T. I guess the thread title was misleading then because it said nothing about LN2 overclocking for 1 minute.

The whole point of the extreme overclocking is that it has nothing to do with performance. Why do you keep bringing this up? The kind of overclocking they were performing (repeating myself) was pretty much trying to squeeze the highest clock frequency the chip can handle. And its about time too seeing as no other parts including SB part (even at single core with parts disabled) broke the record set by the single core celeron which is by no means an easy feat to achieve.

And basically this event showed that BD chips will be highly overclockable. Whether or not it will be worse or better at it compared to the intel counterparts is something thats remained to be seen. Nothing more, nothing less.

 

[PreferLinux]

i think the issue is mainly user driven here...

One sample doesn't dictate overclocking potential so their is nothing to say about that. The only thing that we can mention is that BD broke a record that hasn't been broken for 10 years orso.

Given this oc wasn't on the final silicon either this unknown oc behavior might even change in a negative or positive way.


We don't even know the default performance. we don't know how BD overclocks can be tweaked.. What about overclocking with turbo? What about overclocking the l3 cache?

Stop looking for things to be negative about something that hasn't been done in years.

Actually, the record they broke was only set last month...
http://hwbot.org/submission/2200362_tapakah_cpu_frequency_celeron_lga775_352_8308.94_mhz

 

[Cookie Monster]

Notice how all of them are celerons? No other modern chip has managed to do better til now.

 

[RussianSensation]

And basically this event showed that BD chips will be highly overclockable. Whether or not it will be worse or better at it compared to the intel counterparts is something thats remained to be seen. Nothing more, nothing less.

But if they used helium and LN2 to overclock CPUs, how does that translate to "BD chips will be highly overclockable"? That only tells us that they overclock well on exotic cooling. It doesn't say anything about how well those chips will overclock on air or water cooling (which is what 99.9% of the market will be using).

Secondly, they only overclocked 1 module w/ cache disabled. That tells us nothing about real world overclocking with full cache enabled and all cores enabled. For instance, if someone overclocked 1 core of 2500k to 10ghz, so what? Who is ever going to be using a 2500k with 1 core? No one. Do people overclock a Fermi GPU with half or 1/3 of its shaders disabled? No. I am just amazed these so called extreme overclockers find it "an achievement" to take an 8 core CPU and overclock only 2 of its cores to break a world record...a world record in frequency (which from Pentium 4 days we know is meaningless).

Analogously, that would be similar to taking an 8 cylinder car engine and then claiming you achieved world breaking fuel economy by only using 2 of its cylinders, yet in the real world that would never happen. So what is the point of such an exercise? Bragging it seems.

To me, this is a poor a marketing exercise that gets a couple people excited, and most likely those who are into extreme overclocking. Apple (arguably the most successful marketing/technology company of all time) would never allow for such an amateur marketing event to take place before the launch of their new product. AMD really has needs to take a page from their book and learn how to properly market and launch their products for its intended market segments. Doing an extreme overclocking event to get people excited about BD is a waste of a time. Why wouldn't they show any benchmarks? Talk about power consumption parameters?

I still don't understand why it's a big deal to break 8.3ghz on exotic cooling for anyone who intends to buy BD. Unless you personally use LN2/Helium cooling at home, that overclock is only theoretical in nature because it's worthless in the real world.

Again, this has nothing to do with AMD. If Intel did this with IVB and told me oh look IVB can get to 25ghz on LN2, I'd also find such an exercise a marketing gimmick.

 

[LOL_Wut_Axel]

But if they used helium and LN2 to overclock CPUs, how does that translate to "BD chips will be highly overclockable"? That only tells us that they overclock well on exotic cooling. It doesn't say anything about how well those chips will overclock on air or water cooling (which is what 99.9% of the market will be using).

Secondly, since they only overclocked with 1 module and cache disabled, that again tells us nothing about real world overclocking with full cache enabled and all cores enabled. For instance, if someone overclocked 1 core of 2500k to 10ghz, so what? Who is ever going to be using a 2500k with 1 core? No one.

So again, I only this is as a marketing exercise that gets a couple people excited who are into extreme overclocking. Apple (the most successful marketing/technology company of all time) would never do allow for such an amateur event for their new product. AMD really has no idea how to properly market or launch their products. Doing an extreme overclocking event to get people excited about BD is a waste of a time. I still don't understand why it's a big deal to break 8.3ghz on exotic cooling for anyone who intends to buy BD?

Again, this has nothing to do with AMD per say. If Intel did this with IVB, I'd also find such an exercise a marketing gimmick.

True. However, given that most samples reached 5GHz on air cooling/hybrid water cooling and were apparently stable, at least it's decent in that department, even if it needs 1GHz more to be competitive with SB on single-threaded.

The only real point I see in this chip is folding or perhaps a cheap video encoding or rendering machine, if I'm honest. The FX-8120 doesn't look that bad considering the price and the fact it's unlocked.

 

[Imouto]

@RussianSensation

Is a WR after all. There's a WR even for swallowing boiled eggs in a given time. If you find eating 35 boiled eggs in 1 min useless or retarded the guy with the WR prolly won't give a fck.

This thread isn't about IPC, performance or any other shit. Bulldozer holds CPU clock WR, just deal with it and stay on topic. I'm more tired about the annoying guys ranting the same in every page of every single Bulldozer thread than the lack of information already.

 

[thilanliyan]

Bragging it seems.

There ya go, now you get it.

 

[BlueBlazer]

Notice how all of them are celerons? No other modern chip has managed to do better til now.

And those Celerons are on LN2 only. Perhaps if somone can provide those Celeron overclockers with LHe, then the world record will fall again? 8.3GHz is just a whisker away from 8.4GHz.

 

[Riek]

And those Celerons are on LN2 only. Perhaps if somone can provide those Celeron overclockers with LHe, then the world record will fall again? 8.3GHz is just a whisker away from 8.4GHz.

Thats why the whole top 115 (except for the AMD BD) is all based on p4 derivatives. Its not that they don't try, it is just that they don't get there.

 

[BlueBlazer]

Thats why the whole top 115 (except for the AMD BD) is all based on p4 derivatives.

And these are Netburst designs with long pipelines that are highly suited for high frequencies. It also happens to be coincidental that Bulldozer also have long pipelines for high frequencies.

Its not that they don't try, it is just that they don't get there.

They don't try? Or they don't have LHe? These Celery overclockers are doing it on their own mostly, and do not have financial and hardware support from any big name companies. If you noticed, their hardware components (CPUs and boards) are old. :hmm:

 

[Riek]

And these are Netburst designs with long pipelines that are highly suited for high frequencies. It also happens to be coincidental that Bulldozer also have long pipelines for high frequencies.

They don't try? Or they don't have LHe? These Celery overclockers are doing it on their own mostly, and do not have financial and hardware support from any big name companies. If you noticed, their hardware components (CPUs and boards) are old. :hmm:

You might try to stick a p4 775 celeron into an LGA 1155 board... but i expect worse results though. PS: they run on those mobos because they get the highest with them not because they have no budget.

Bulldozer pipeline length is being exagurated... it is closer to phenom than it is to prescott.. Heck Bulldozer pipelinelength is also closer to Sandybridge than it is to phenom2 or prescott.

 

[Rifter]

OP is misleading

Overclocking with exotic cooling and butchered CPU(cores/cache disabled) has zero meaning towards "normal" overclocking(such as everything enabled, 24/7 stable, at voltages that will not degrade CPU at astronomical rates)

Also saying BD overclocks well fromt this example which is almost guranteed to be hand picked samples is a stretch.

Wake me up when its released and we have real world numbers and a large user base to draw conclusions on its overclocking from..

Sorry if this has been mentioned already im getting tired of all the BD threads so didnt read all responses which im sure are the same as the other 10 threads on it.

 

[BlueBlazer]

You might try to stick a p4 775 celeron into an LGA 1155 board... but i expect worse results though.

You can't stick that LGA775 Celeron into a LGA1155 board! They are incompatible in every way (including the memory and I/O interface)!

PS: they run on those mobos because they get the highest with them not because they have no budget.

They are "small time" extreme overclockers and small/limited budgets (mostly their own). They do not have access wide range of CPUs and motherboards (to cherry pick one out), since these are old and obselete (EOL'ed).

 

[BlueBlazer]

Sorry if this has been mentioned already im getting tired of all the BD threads so didnt read all responses which im sure are the same as the other 10 threads on it.

There another thread with identical subjects/discussion as this one also. Perhaps the mods could merge these threads together?

 

[Gundark]

@BlueBlazer - i don't understand, what's your point? Why so negative? In this thread we comment on BD overclocking, and it is impresive, not only becouse of WR, but becouse of many speculations of BD small headroom. I also belived that BD won't see 5GHz at all. Since I'm proven wrong, i see that only in positive manner.

 

[Idontcare]

I remember when PhII was about to launch, AMD did the same thing - hosted an extreme OC'ing event, clocked the PhenomII to silly high levels under silly conditions (single core, LHe, crazy voltages, etc) - and then, as now, there was much "rabble, rabble" within the community.

And then, as now, the question of relevance was at the forefront of the arguments.

But we did come to see that the overclocking potential or headroom intrinsic in PhII come to be an advantage even for air OC'ing, chips reaching near 4GHz on air, a little past 4GHz on water.

So what are we to take away from history repeating itself here? In terms of both our community "rabble rabble" response as well as in terms of Bulldozer itself?

 

[BlueBlazer]

@BlueBlazer - i don't understand, what's your point? Why so negative? In this thread we comment on BD overclocking, and it is impresive, not only becouse of WR, but becouse of many speculations of BD small headroom. I also belived that BD won't see 5GHz at all. Since I'm proven wrong, i see that only in positive manner.

Where did I comment about the "small headroom" part (as in low headroom for overclocking) in this thread? :hmm: In my last few post here, I was talking about how close old Netburst Celerons (8.3GHz) are to the World Record and that if that the World Record (8.4GHz) could fall again if they had access to LHe.

 

[Gundark]

You didn't comment about small headroom, and i didn't say that you did. I ask something else. If someone isn't satisfied with expected BD performance that's another matter. This is not performance thread ( as i see it ).

 

[BlueBlazer]

I remember when PhII was about to launch, AMD did the same thing - hosted an extreme OC'ing event, clocked the PhenomII to silly high levels under silly conditions (single core, LHe, crazy voltages, etc) - and then, as now, there was much "rabble, rabble" within the community.

More or less the same thing, however in the last events they did show a bit of benchmarks (like this example), but none here today.

And then, as now, the question of relevance was at the forefront of the arguments.

But we did come to see that the overclocking potential or headroom intrinsic in PhII come to be an advantage even for air OC'ing, chips reaching near 4GHz on air, a little past 4GHz on water.

So what are we to take away from history repeating itself here? In terms of both our community "rabble rabble" response as well as in terms of Bulldozer itself?

It could be deduced that it would clock higher than 4GHz, possibly reaching around 5GHz with good chips (similar to Sandy Bridge). Sooner or later though, people will be wondering/asking how it will perform at those frequencies. :hmm:

 

[Gundark]

@Idontcare - what was the max OC on PhII?
Edit: about 7GHz, right?

 

[Ancalagon44]

So what are we to take away from history repeating itself here? In terms of both our community "rabble rabble" response as well as in terms of Bulldozer itself?

Frankly its quite worrying.

Lets put it this way - everything we have seen indicates that, if there are improvements to IPC, these improvements are minor. We are talking a 0-10% IPC gain over PhII. Most of what we are seeing is the 2 cores 1 module stuff - basically allowing higher core counts for a smaller transistor budget increase than normal. And, we are also seeing optimizations such as longer pipeline length that seem to be designed to allow clock speed to scale higher than PhII.

Lets look at these individually.

1. Clock speed scales better, but from what we've seen, 4GHz is still considered high and normally only achieved via turbo mode or overclocking. 5Ghz is reportedly possible on air but only in ideal conditions, and probably allowing for a lot of extra voltage.
2. Higher core counts is nice, but doesnt really benefit us gamers all that much. Time will tell whether games that benefit from a 4 core hyperthreaded CPU would be faster running on a physical 8 core BD.
3. Middling improvement to IPC is the most worrying to me. Remember, Phenom and Phenom II were both supposed to have a middling improvement to IPC over their predecessors, and while that was true, it still wasnt enough to leapfrog over Intel. Here we have a new BD CPU where it seems we will be lucky if IPC gain is there at all. Coupled with what we know about the clockspeed, which although higher is not high to beat SB CPUs let alone IB CPUs, and you have what I would call a worrying scenario.