AMD OC Record Broken, Still Powered by AMD FX-8150

[cdawall]

but its a semi pentium 4 slow un efficient u may call it that way
i.e. desined to run at highest chocks possible without taking in mind the ipc that was the p 4 and now the bulldozer

Well this chip technically is not inefficient, it is not as efficient as the Intel offerings. It is however still a better CPU than the previous generations in certain aspects. IPC on these chips is still 4 the same as the intel offerings and up from 3 on the old chips and 2 on P4 "netburst".

It isn't a Bulldozer thread per se since we're talking about a 2-core, 2-threads CPU from AMD (which doesn't exist under the "Bulldozer" lineup).

No that little box still says "Bulldozer" the core is still a bulldozer core. This isn't any different than a rebadged Sargus core which is really just a Regor core tuned down.

And also don't forget this is on LN2 and not LHE I think 9GHz is close

You would be surprised how little gains they get from LN2 to LHe we will see larger gains moving from stepping to stepping as fab techniques improve. The same way the last batches of Phenom 1 clocked better and the last batches of Phenom II clocked better.

 

[cadaveca]

It's definitely skill. And there is a a lot more knowledge and experience involved than the uninitiated is like to think, starting with knowing how not to kill your hardware from any number of things from condensation to too much voltage.

I do not agree 100%. Current motherboard technology takes alot of the guesswork out of it.

You can kill chips on air easy enough. The biggest thing in the past was that you needed to know how to OCP mod most boards, and had to know how to recognize coldbug, and how to get even the memory contorller to work right, by adjusting drive strengths and such. That means that to get the best results, you needed not to have real knowledge, other than knowing what parts to buy. And there are literally only a few guys out there that gave out this info in the first place, and the rest simply did what they were told works. If you are not friends with someone in the know...you're wasting your time.

Insulation and such is a given. But I hardly see that as a skill.

The big thing for me, is that I cannot call something that people do not share with others as a skill. It's a SECRET...and there are many SECRETs in the Extreme scene. that closed community, and the secrets that surround it, is what has me think it's a waste of time, marketing-wise.

 

[cdawall]

I do not agree 100%. Current motherboard technology takes alot of the guesswork out of it.

You can kill chips on air easy enough. The biggest thing in the past was that you needed to know how to OCP mod most boards, and had to know how to recognize coldbug, and how to get even the memory contorller to work right, by adjusting drive strengths and such. That means that to get the best results, you needed not to have real knowledge, other than knowing what parts to buy. And there are literally only a few guys out there that gave out this info in the first place, and the rest simply did what they were told works. If you are not friends with someone in the know...you're wasting your time.

Insulation and such is a given. But I hardly see that as a skill.

The big thing for me, is that I cannot call something that people do not share with others as a skill. It's a SECRET...and there are many SECRETs in the Extreme scene. that closed community, and the secrets that surround it, is what has me think it's a waste of time, marketing-wise.

While I have not played with BD yet I can still say there is some skill involved with extreme clocks. There is still a lot that plays into any of these chips regardless of Intel or AMD. Simple things like 45nm AMD chips had a 32nm memory controller that would fry the entire chip if you pumped its volts high and left the overall chip volts low. Same sort of thing applies to these there is only so much voltage you can push in certain spots to get them clocked up. A lot of the guesswork and trail and error might be gone with new highend boards, but there is a reason why any Joe Blow with LN2 cannot hit these clocks. Plenty of people with money buy the best of the best parts and can't break into the top 10 clocks done by guys who know what they are doing.

 

[cadaveca]

I agree, CDA, but the fact that it's always the same 10-15 guys is stupid. ALWAYS. Hasn't really changed at all in the past 10 years. Few new guys, sure.

And to me, figuring this stuff out isn't that hard. Most people, on the other hand, sure...don't even have any idea how to set secondary timings.

I'm not a smart guy, really. if I cna do it, so can near anyone else. Trail and error ain't all that hard, really.

 

[Super XP]

Skill is always apart of Overclocking. We are talking about NB, CPU and Memory OC's, if you can get them at a nice speed with stability, then that is called skill IMO. To achieve this, you take it slow and perform the OC step by step removing all possible limiting factors such as keep memory speed and NB speed at stock, take your CPU step by step as high as you can. Find the max or close to it. Then put the CPU back to stock, and try upping the NB speed until you can at least get close to your CPU speed. When you have that, then keep everythign on stock, and play around with your memory speed and find it's max, then drop it down a little to certify stability.

That said, if you find you run into stability issues, bump the voltage a little and continue.

What I noticed in my past with both AMD and Intel CPU's is if you rush with the OC too fast and you don't take her up slowly, for some reason you end up with a higher vCore and at a lower OC speed.

Sure today's hardware make it easy to OC, along with auto OC options, though I would be against this IMO unless AMD/Intel implimented something in the CPU hardware that will not mess up the true potencial for a good stable OC by moving too fast to achieve it.

JMOP

 

[Wile E]

It's not skill that makes an OCer these days, it's knowledge.

Back when you had to mod and solder on your board to extreme clock there was skill involved. Not these days. Typing in the proper voltages and frequencies does not require skill at all. A trained monkey can do that. It does, however, require knowledge to choose the proper settings.

 

[Am*]

Invalid claim is invalid. Unless it's running all the cores of the 8150, it's not setting any records for the chip, which has serious trouble being bench-stable past 4.6GHz with all cores active. AMD's marketing department is taking the piss here once again. Also these records really aren't impressive, considering people were hitting 8GHz+ on inefficient Pentium 4 65nm/90nm architecture chips back in the day, without turning off any parts of the CPU.

In other news, performance records still powered by Intel CPUs...

 

[repman244]

Invalid claim is invalid. Unless it's running all the cores of the 8150, it's not setting any records for the chip, which has serious trouble being bench-stable past 4.6GHz with all cores active. AMD's marketing department is taking the piss here once again. Also these records really aren't impressive, considering people were hitting 8GHz+ on inefficient Pentium 4 65nm/90nm architecture chips back in the day, without turning off any parts of the CPU.

In other news, performance records still powered by Intel CPUs...

This isn't really AMD marketing.
If those Pentiums had 2 cores they would disable them. In fact they did disable HT so you are incorrect. http://hwbot.org/submission/592402_theking_cpu_frequency_pentium_4_631_8179.89_mhz

And yes the FX can run all 8 with 8GHz:

http://www.xtremesystems.org/forums...8-AMD-Bulldozer-FX-8150P-BE-8GHz-Super-PI-1M&

And also this is a FREQUENCY record where you are after highest frequency possible by any means necessary. It's not about how well does the CPU perform.
It's the same thing if you are after a memory speed record and you run it with a low budgte GPU and people would say, that doesn't prove anything because you can't do anything with it...


I see that people have serious problems understanding world records. They aren't set with normal stuff.
For example is the world land speed record set by a Renault Clio? Not really...

 

[HTC]

This isn't really AMD marketing.
If those Pentiums had 2 cores they would disable them. In fact they did disable HT so you are incorrect. http://hwbot.org/submission/592402_theking_cpu_frequency_pentium_4_631_8179.89_mhz

And yes the FX can run all 8 with 8GHz:

http://cfile7.uf.tistory.com/original/172E473A4E9661F42C0797

http://www.xtremesystems.org/forums...8-AMD-Bulldozer-FX-8150P-BE-8GHz-Super-PI-1M&

And also this is a FREQUENCY record where you are after highest frequency possible by any means necessary. It's not about how well does the CPU perform.
It's the same thing if you are after a memory speed record and you run it with a low budgte GPU and people would say, that doesn't prove anything because you can't do anything with it...


I see that people have serious problems understanding world records. They aren't set with normal stuff.
For example is the world land speed record set by a Renault Clio? Not really...

That's quite impressive, IMO.

For those that complain about how bad superpi is, look @ this:

And its more simillarz results-ussually big improvement in photoshop, 3DCAD, practice compression decompression, microsoft office working, megatasking
I thought, if we are at XS, we are not stupid live only in superpi benchmarks etc, but we know also something about real performance of new products....

Taken from here (modified a bit because it's a big post): http://www.xtremesystems.org/forums...-Super-PI-1M&p=4974762&viewfull=1#post4974762

 

[n-ster]

That's quite impressive, IMO.

For those that complain about how bad superpi is, look @ this:



Taken from here (modified a bit because it's a big post): http://www.xtremesystems.org/forums...-Super-PI-1M&p=4974762&viewfull=1#post4974762

An i7 920 seems to be as good as the bulldozer clock for clock. Also note that these comparisons are hardly fair, as many i5 2500K or i7 2600K are being rung at 4.8Ghz +, while Bulldozer is, AFAIK, ran at 4.4Ghz+. Even if it is 4.6Ghz +, the Bulldozer does not reach LGA 1155 in OCing on Aircooling, yet in the benchmarks the Bulldozer is clocked higher

Now bench again with typical OCs like i7 920 @ 4Ghz, and the result will be much better for intel, who usually leave a much greater room for OC

 

[Super XP]

That's quite impressive, IMO.

For those that complain about how bad superpi is, look @ this:



Taken from here (modified a bit because it's a big post): http://www.xtremesystems.org/forums...-Super-PI-1M&p=4974762&viewfull=1#post4974762

Well this explains 3 issues one being Bulldozer requires minor modifications for better overall performance in both single and multi threaded. Two being current benchmarks are not properly optimised for Bulldozer's design structure. IMO it's like trying to ram a 454 big block into a super tiny electric car. Does not work this way. The third being 4P/4T greatly outperforms 2P/4T by a lot especially with the software not being properly optimized.

Looking at all these latest findings, AMD's upcoming B3 revision is going to really be interesting.