HardOCP Overclock Phenom 9600 Black Edition

[cnumartyr]

Did you use a multimeter?

Also 200W for just 3Ghz vs 200W for 3.6Ghz and on top of that the Q6600 beats Phenom clock for clock. Hmm...

I've checked all of my voltages with a multimeter to varify they were near BIOS setting. However no I haven't checked Amp draw on the CPU with a multimeter.

That said, I've read numerous places on the wattage of a Q6600 at 3.6 GHz so I assume it's around 200W.

And I agree with what you said about the Phenom/Q6600, but you said 200W for a CPU was rediculous.. point/counter-point.

 

[Evilonigiri]

Yeah if you answered yes about the multimeter, I'd ask how you did it 😀

I still think 200W for just the cpu is ridiculous. But now that I think of it, the watt usage is exponential as you clock higher and higher.

 

[sailer]

pathan7- Yes, performance probably would go up if the L2 cache size was larger, in the 1 MB per core range. In fact, it could probably use 2 mb per core. But there probably isn't enough space left with the L3 cache already in place, so it may not be possible, but I can't say for sure.

For a point of comparison, Intel is using 2 MB per core on the Kentsfield design and up to 12 MB on the Yorkfield (I'm not sure if the 12 MB is per core or total, which would mean 3 MB per core). There's got to be good reason for those large L2 caches.

 

[cnumartyr]

pathan7- Yes, performance probably would go up if the L2 cache size was larger, in the 1 MB per core range. In fact, it could probably use 2 mb per core. But there probably isn't enough space left with the L3 cache already in place, so it may not be possible, but I can't say for sure.

For a point of comparison, Intel is using 2 MB per core on the Kentsfield design and up to 12 MB on the Yorkfield (I'm not sure if the 12 MB is per core or total, which would mean 3 MB per core). There's got to be good reason for those large L2 caches.

Yorkfield is 12 MB total.

Wolfdale is 6 MB total.

Honestly.. I can't remember the link off the top of my head, but clock for clock 1MB -> 2 MB Cache was worth about 1-3% and 1 MB -> 4 MB was about 5-7% improvement. (Comparing equally clocked Allendale "Lite" (E21x0 Series), Allendale (E4x00 series), and Conroe (E6x00 series).

 

[sailer]

Cnumartyr- Thanks for the clarification on the L2 cache sizes.

 

[pathan7]

oh thanks guys so seems like when AMD move to 45nm technology then they can implement larger L2 caches, and possibly larger L3 caches too.but if cache can be of so paramount importance why didn't AMD go for larger L2 cahces in the first place (unless they under-estimated their competition). BTW i am no AMD fan boy but i do not like monopoly as it affects the end user in many ways , price being one of them.

 

[cnumartyr]

oh thanks guys so seems like when AMD move to 45nm technology then they can implement larger L2 caches, and possibly larger L3 caches too.but if cache can be of so paramount importance why didn't AMD go for larger L2 cahces in the first place (unless they under-estimated their competition). BTW i am no AMD fan boy but i do not like monopoly as it affects the end user in many ways , price being one of them.

Cache for AMD isn't "as" important because it has it's own memory controller on die. Intel has slower memory bandwidths and as such has more Cache to compensate... Something like that.

 

[Mathos]

And for an additional light of hope for AMD, according to the Dallas morning News, the stock price for AMD actually raised 30 cents in todays paper.

 

[sailer]

I think the reason they didn't go for the larger L2 cache was their decission to build a large L3 cache. The 45 nm die might help for increasing the size of the L2 cache, but I can't say for sure. When designing chips, many things have to be considered, including compromises, and the engineers made what they thought was the best choice. Sometimes only time and use tell whether their choice was a good idea or not.

 

[d4NjvRzf]

Check it out.seems rather interesting (don't know if that site is trustworthy though)
http://www.neoseeker.com/Articles/Hardware/Reviews/phenom_9900/

As I posted earlier in a previous thread, that article demonstrates that Phenom wins mostly in memory benchmarks -- traditionally AMD's strong suit -- while Core wins in the real stuff like games. The situation isn't much different from before Phenom's release.

 

[pathan7]

As I posted earlier in a previous thread, that article demonstrates that Phenom wins mostly in memory benchmarks -- traditionally AMD's strong suit -- while Core wins in the real stuff like games. The situation isn't much different from before Phenom's release.

i wasn't thinking about 9900 beating Q6700 but i was looking at the wider picture: Penom 9600 Black Edition if overclocked to 3.0 GHz looks rather interesting considering the price point it is released at.

 

[d4NjvRzf]

Perhaps so, but unless you intend to use > 2 cores, is it worth the extra $$$ over the 5000+ BE for the per-core performance delta?

 

[coret]

Perhaps so, but is it worth the extra $$$ over the 5000+ BE for the performance delta?

Depends on what you do ... If you do a lot of encoding or play supreme commander, the extra 2 cores are very useful ^^

 

[sailer]

Perhaps so, but is it worth the extra $$$ (say, over 5000+ BE) to get a rather marginal performance increase?

That's a very good question. I'd like to see some benches where an overclocked 5000+ BE was compared to a Phenom BE. As whether or not its worth it, it would also depend on what you do with it. Very few games would take advantage of any quad out there, whether AMD or Intel. At the moment, quads are more useful in business apps, and to people who might use their computer for both business and gaming.

 

[Mathos]

Well, they did a comparison on Toms regarding clock for clock between the phenom and X2 cores, they phenom was either just slightly ahead, or in some things very much ahead.

 

[sailer]

I read that review, but it wasn't quite the whole truth and nothing but the truth. First off, they were overclocking the Phenom. Second, they underclocked the X2 6000+. So neither chip was as you would buy one off the shelf. They managed to show the Phenom being better per core then the X2 at a given clock speed, but they didn't prove the Phenom was a better chip as a whole.

Had the X2 been allowed to run at its standard clock speed, or been overclocked, I'd imagine the X2 would have left the Phenom behind in the tests. That's been one of the sore spots for the Phenom. Except for the fact that its a quad and therefore can do four things at once, as opposed to a dual core only doing two things at once, it so far has shown no real advantage. The 9600 BE, if overclocked to 3 ghx, should logically run better than a X2 6000+ at stock speed. But the 6000+ can be overclocked as well, thus still canceling out the Phenom BE's advantage.

 

[coret]

Sailer, while what you say is true the 6000+ doesn't clock much higher than it's stock ... around 3.4GHz max, and that's really pushing it (125w TDP at stock vs phenom's 95w).

Meanwhile the K10 has about 10-15% average clock for clock advantage ... which would (at 3GHz) would put it's cores somewhere around the equivalent of a 3.3-3.45GHz K8 ... except there are 4 of them, not 2.

 

[someguy7]

a K10 at 3ghz uses alot more then 95w. That TDP is for the stock settings. The quad core will use alot more w when its OC''d compared to its stock rating

 

[coret]

As will the K8 ... I was just pointing out that the K8 has a higher starting point/

 

[soloman02]

Xbitlabs did a test of Phenom in real world settings. In games, even a 3Ghz Phenom is beaten by a stock Q6600 (let alone an overclocked one),

http://www.xbitlabs.com/articles/cpu/display/amd-phenom_11.html

 

[gpippas]

Xbitlabs did a test of Phenom in real world settings. In games, even a 3Ghz Phenom is beaten by a stock Q6600 (let alone an overclocked one),

http://www.xbitlabs.com/articles/cpu/display/amd-phenom_11.html

Just read the article and he doesn't half give Phenom a kicking at the end. I have to say that I'm suprised that a 3ghz Phenom still can't beat a Q6600. Thats worrying because like the reviewer said AMD aren't even planning to release a 3ghz cpu.

 

[sailer]

Sailer, while what you say is true the 6000+ doesn't clock much higher than it's stock ... around 3.4GHz max, and that's really pushing it (125w TDP at stock vs phenom's 95w).

Meanwhile the K10 has about 10-15% average clock for clock advantage ... which would (at 3GHz) would put it's cores somewhere around the equivalent of a 3.3-3.45GHz K8 ... except there are 4 of them, not 2.

First, Phenom consumes far more than 95wt when overclocked. Some of figures from overclocks in the range of 3 ghz have been around 235wts. That's a lot of power, which in turn makes a lot of heat and will take a large heatsink to disapate it.

Second, I did point out that Phenom has four cores verses X2's two cores and thus would have an advantage with those programs that can use four cores. But that leaves the practical question; how many programs, particularly games, use those four cores?

I think the Phenom has a lot of promise, but at this point, it has shown no real advantage over the AM2 chips. It is slow, has errata problems, and makes too much heat. These difficiencies are even greater when the 65nm X2 chips are considered, such as the AM2 5000+ BE instead of the 90nm 6000+. AMD needs to address these problems, and until it does, no amount of cheerleading will make the Phenom better than it is. For the moment, in my opinion, the 5000+ BE remains the best chip to buy from AMD.

 

[bfellow]

Phenom 9600 is 95W, 9700 is 125W, 3.0 ghz = defintely above 125W (200W maybe)?

 

[coret]

First, Phenom consumes far more than 95wt when overclocked. Some of figures from overclocks in the range of 3 ghz have been around 235wts. That's a lot of power, which in turn makes a lot of heat and will take a large heatsink to disapate it.

Second, I did point out that Phenom has four cores verses X2's two cores and thus would have an advantage with those programs that can use four cores. But that leaves the practical question; how many programs, particularly games, use those four cores?

I think the Phenom has a lot of promise, but at this point, it has shown no real advantage over the AM2 chips. It is slow, has errata problems, and makes too much heat. These difficiencies are even greater when the 65nm X2 chips are considered, such as the AM2 5000+ BE instead of the 90nm 6000+. AMD needs to address these problems, and until it does, no amount of cheerleading will make the Phenom better than it is. For the moment, in my opinion, the 5000+ BE remains the best chip to buy from AMD.

1) And in my post after that last one, when someone else mentioned this point, I explained that I was pointing out that the 6000/6400+ ramp up quickly in power consumption as you overclock as well, and that at least the phenom 9600 BE has a lower *starting point* on power consumption than the top end K8's.

2) True, not many things a home user will want to do will utilise a quad core at the moment ... but in a year or so, maybe there will be. Also, while they may not use all four cores for a single task, how many people do multiple things at once? I've been known to have a movie playing, encoding a video, browsing the web and playing an MMO all at the same time (multiple monitors 4tw) and it would be useful in those situations.

3) I agree with you, for the vast majority of people, a 5000+BE is probably the best solution from AMD. However, some amount of "cheerleading" as you put it is required to stop some people from saying it's worse than it is (such as with the TLB issue ... show me a review site which has actually been able to reproduce it)

 

[d4NjvRzf]

First, Phenom consumes far more than 95wt when overclocked. Some of figures from overclocks in the range of 3 ghz have been around 235wts. That's a lot of power, which in turn makes a lot of heat and will take a large heatsink to disapate it.

235wts is pretty close to the 2*125=250wts required by a pair of QuadFX processors. This means noise like the QuadFX too.