AMD Phenom II 940 "Xtremely" Benchmarked

[uguv]

Its not just the memory thats being changed on the AM3s. Do your homework and find out for yourself why theyll be faster. Im sure a few here know why, and Im NOT talking about the DDR3, which will help in a few ways.

OK, so you're just guessing. Just wanted to be sure.

I'm hoping the AM3 Phenom II's can keep up with the Q9550. It might bring another round of price cuts. Plus with more boards using DDR3 it should drive the costs for DDR3 down, making the i7 platform more affordable. If AMD can pull it off it should be a win for everyone.

 

[JAYDEEJOHN]

No, not guessing, itll be at least 4%. The HTL is increased, as well as better revs and DDR3

 

[JAYDEEJOHN]

Figure anywheres from 5% from the HT Link speed increase, plus about 1% for DDR3. Accross the board, 4-6%

 

[gallag]

4-6% just for them improvements, Man, So you think that slightly increasing the ht link and ddr3 will bring more ipc improvements than deneb achieved over agena? Fascinating stuff.

 

[JAYDEEJOHN]

Deneb is around 7% over Agena, Im saying around half that. HT Link speeds scale perfectly in alot of apps, and DDR3 gives a slight boost, almost accross the board, and yes, its mostly just 1%, but more is more. If you think Deneb is only 3% faster than Agena, dont know where youve been getting your info. Or do you think its less?

 

[BadTrip]

typical JDJ, When I was posting links to pre-release benchmarks of I7, it was hype. Now he is pulling facts from thin air and its ok.

 

[gallag]

Deneb is %5 over agena, http://www.xtremesystems.org/forums/showthread.php?t=214784

http://www.xtremesystems.org/forums/showthread.php?t=214784

Do you honestly think that a 200mhz bump in htlink will nett 5%? If so why did amd wast time improving core logic, cache,tlb, branch etc when by your logic all they have to do is up the htlink to 2.6ghz to have a nehelam killer.

 

[JAYDEEJOHN]

Why didnt i7 come out at 4.5Ghz? Heat? Power? Maybe theyll only get so fast? No better speeds in the future? Revs dont really help? Process maturation is a myth? Cmon...
http://www.overclock.net/amd-cpus/434005-phenom-i-phenom-ii-review-comparison.html
http://www.overclock.net/5205435-post35.html
http://www.overclock.net/5205455-post38.html
http://www.overclock.net/5205478-post41.html

 

[JAYDEEJOHN]

If you take out the audio encoding, add a few games, its waaaay more than that. For some reason audio encoding doesnt go that fast on Deneb. Look at the game numbers tho, and its waaay more than 5%

 

[gallag]

Lame rebuttal, Also have you just linked to the comparisons with the strongest numbers to prove your point while forgetting all others? What about the avg? You actually as a amd fan link to superPi?

Think about it, There are many people out there with crazy clocks on the htlink 2.6-2.9 is common, and the numbers do not lie. Just reconsider that you might be wrong.

 

[MosportMike]

It always amazes me how Intel guys want nothing more than AMD to fail due to fear that they may produce something better than what is in their system. Could you imagine how quiet the Intel community would be if AMD ever takes the lead again. The difference is I would never rub in their faces!!

 

[JAYDEEJOHN]

Theres nothing lame about using the exact same setup, just different cpus as seen here http://www.xtremesystems.org/forums/attachment.php?attachmentid=92737&d=1231703017
Ive already seen a clock for clock using the higher HT Link, and thats what Im going on, that and AMD has stated 4% also. Besides which, AMD has to sell these chips, and offer more by doing so. It only makes sense. And whats so lame about the audio encoding only? Im a gamer, If Id wanted to, I couldve just included those nimbers, which are far far higher than a measly 5%

 

[gallag]

Its simple, Just link to a complete set of benchmarks with a avg without removing or adding any thing to prop up your argument, Like I did. Where have you seen a 200mhz ht bump nett 6% ipc? Just show the facts.

 

[gallag]

And I said it was a lame rebuttal because you went on a rant about i7 when we are talking about denebam2 to denebam3 ipc improvements, Keep to the one discussion.

 

[spathotan]

This is incredible. 19 pages of the same arguments with no end in sight.

Deneb delivered for AMD users. Not the industry.

 

[gallag]

Just like this, Real world apps, No sinthetics , Nothing removed or added to support my argument, I did not link to only the lowest improvements like you do as the average is the important figure. Credit for the CHART hard work goes to petnorth from XS.
, Also, This is nothing to do with wanting to see AMD fail, I would like to see people stop hyping every thing that they do until it can only disappoint.

 

[JAYDEEJOHN]

What i7 rant? I think we are talking about P2 here. And as you requested http://www.xtremesystems.org/forums/attachment.php?attachmentid=92737&d=1231703017 Ill post it again

Its simple, Just link to a complete set of benchmarks with a avg without removing or adding any thing to prop up your argument, Like I did. Where have you seen a 200mhz ht bump nett 6% ipc? Just show the facts.

 

[JAYDEEJOHN]

Have you looked at my link?

 

[dattimr]

4-6% just for them improvements, Man, So you think that slightly increasing the ht link and ddr3 will bring more ipc improvements than deneb achieved over agena? Fascinating stuff.

It's not that "fascinating", however, is not only doable, but also inevitable. K10 can't have its NB clocked lower than its HT link, which means that, since AM3 will start with an HT link of 4000 (effective), its NB will be clocked at 2 GHz, just like the BEs of Agena.

People have no issues whatsoever clocking their PII NB at 2.2 or 2.4 GHz on Xtremesystems, and that usually results on a 6-10% improvement, depending on the benchmark.

DDR3 also shows some nice improvements in many cases, even if coupled with the FSB system, so, chances are that it will help PII at least a little. Link:

http://ixbtlabs.com/articles3/cpu/intel-c2x-qx9770-ddr3-800-p2.html

In case of the Phenom system included on the benchmarks, it's the difference when going DDR2 1066.

Other interesting fact is that games, for instance, work better with K10 "unganged" memory mode. Link:

http://ixbtlabs.com/articles3/cpu/amd-phenom-x4-9850-ganged-unganged-p2.html

The improvement is in the 0-6.5% range. In most reviews I haven't seen any commentary regarding which mode was used, so, I tend to believe it's "ganged", since that's the default for most BIOS versions (actually, I think it's for all, but I'm not sure about it).

Couple that with a second revision of their 45nm tech and there you go.

 

[spathotan]

P2 940 is 9.69% faster than Phenom 9950 according to JDJ's chart. Considering the $85 difference in price, its really a coin flip, it dosent matter either way. They can both overclock about the same amount on air, around 500-700mhz each.

 

[JAYDEEJOHN]

Both of these users have confirmed like results with each other. And whats so hard to believe having a higher HT Link wont scale? At least mostly so? Wouldnt the bclk scale as well on i7?

 

[roofus]

omg why is 6% gain so hard to digest?

 

[dattimr]

http://i472.photobucket.com/albums/rr84/PetNorth/Summary/summary_agena_deneb.jpg

Just like this, Real world apps, No sinthetics , Nothing removed or added to support my argument, I did not link to only the lowest improvements like you do as the average is the important figure. Credit for the CHART hard work goes to petnorth from XS.
, Also, This is nothing to do with wanting to see AMD fail, I would like to see people stop hyping every thing that they do until it can only disappoint.

Just take a look at the same thread you got this chart from and you'll see another one with the NB scaling of PII.

 

[JAYDEEJOHN]

typical JDJ, When I was posting links to pre-release benchmarks of I7, it was hype. Now he is pulling facts from thin air and its ok.

I think Ive provided a few links here, as well as dattimr, so wheres your links of rebuttal? This isnt hype, unless all this hard work is meaningless and wrong according to just you

 

[gallag]

Have you looked at my link?

Yes full of synthetic benches, But if that is how you want to roll, I think most people agree that deneb has about 5% over agena, If you through in a lot of sythetic memory benches some of which inflate the figures drastically i.e mbench write data 29% etc, Surley you can see the flaw here, I will give you games though, If you where just to count the improvements in games alone which are real world then your point is more valid but surely you agree that real world averages are more important?