AMD Phenom II 940 "Xtremely" Benchmarked

Page 11 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
SighQ2, If the P2 running at 3.0 and I7 is running at 3.2 does that mean AMD actually made up some clocks? Or am I off base here.
It would really be intersting if they dialed those babies up a little.
 
As far as I'm not married with either Intel or AMD, I could build either a Phenom or Intel Q, but at this moment I just hope my motherboard to be phenom II compatible with a bios upgrade.

Core i7 is sweet but I can't even dream in getting one with such a price tag. In my personal world its either Phenom X4 or Q6600. I could even get some good juice from an old Phenom 9600 with the bug, and fix turned off, as I'm pretty sure It won't apply to my business. Also, the buggy P is quite cheap these days, like 120$, I think its fair.

If the Phenom II is good enough, the price war is the key moment I'm thinking about.
 
The major mobo makers are currently posting their lists of boards with new bioses that will support Phenom II or AM3 cpu's generally.

The first P2 is an AM2+ cpu. (940 pins)

Note that the sok AM3 CPU's (938 pins) will work in sok AM2+ BOARDS w a bios upgrade.

That's because the DENEB has DUAL IMC. integrated mem controller. and will support EITHER DDR 2 or DDR3.

The reverse is not true. You cannot fit an AM2+ cpu into an AM3 socket.

There is a remote possibility that a mobo maker COULD actually create a sok am2+ mobo that supports ddr3 ONLY - not both kinds due to incompatible rams. But don't hold your breath on that one, cos you could be waiting for a cold snap.

We need more details (NDA) about whether the AM3 938 pin cpu running in an am2+ 940 socket will suffer any performance hit - but it is known, it will work.

Supposedly it will work in Sok am2 as well, but that's not recommended and is really going to do a performance hit = stupid idea but it's possible to do it if you are really lost in Siberia and need to get online to use a google map to save your butt :) - whatever

This stuff really confuses people but it is actually simple - you have to realize that the initial release is a socket am2+ cpu. (easy upgrade if you have an am2+ mobo already - especially if you have one with an SB750 southbridge - so you can use ACC oclox utility for easy oclox.

Later, like late spring, the real change happens - to socket AM3 with DDR3 ram. There will be new mobos, etc. for all that. And a whole list of various models releasing through the summer. quads, triples, duals all based on DENEB core.

Including the early release AM2+ models, there are 14 new cpu's planned, all 45nm coming in the next roughly six months. Some have less cache - kinda like what they did a long time ago with the old durons.

See amd zone website forums - thread called 'ez guide to socket am3' for a better description if you are interested in resolving some of the confusion or want more detail.
 


I don't really know.

It's a simple, well done, comparison review. (I think).

That's why I posted it - for basic comparison, and to see what some experts here really think of this review.

There isn't a great deal of formality, and yet there is some, and apparent consistency in the test procedure - so I don't think the tester is trying to snow job or spread a lot of fud or lies or bias or what not. I thought it looked clean enough, with a lot of benches that a lot of people can relate too.

But I wonder if the results compare to what is posted here = the main focus of this thread = and no I do not want to hijack anything.

But honestly, there are not many people commenting on the massive test results posted here. SO I thought a simple review might be interesting BUT I really want to know if there is agreement generally between the tests. If so, then we have hi tek info plus we have simple info so a lot more people can relate to what's happening - that's all.

I mean no disrespect to dattimr - I think it's magnificent that he has gone to all that effort to feed our starving interests. There is so much data here, like wow. This thread could be a reference perhaps for a long time.
 
Thanks Sigh, and hats-off to Dattimr for starting this thread, most definitely been informative.
Hopefully somebody will come up with the 2 CPUs at the same clocks. Then see which one can surpass the other on air, water, and ice.
I'm sure it won't be long.
 
The OCW contains tons of...ummm misinformation, if you will. None of those numbers match previous numbers from either or all of the cpus tested. There isnt enough complete info on all the rigs used to pinpoint exactly how and where it went off track, but it is off nonetheless. I dont think theyre lying, I think they did the pooch, and agin why Im not sure. Maybe some kind of insane NDA thing, but who knows. Every glimpse Ive seen has been close, but , no, you dont get the whole thing, and I believe thats how AMD wants it, and possibly thats whats going on with OCWs findings as well, maybe nothing exact, but a close picture overall. Thats my opinion anywho, and no, Im not discrediting them, or the numbers, its just not optimal, and I dont know why
 


Thx jdj. I typically enjoy your posts, and your individualistic point of view - ummm, that's a compliment, I think.

My lazy mind tends to avoid synthetic benchmarks, and so I look to other ways of relating to performance. Being a gamer, I try to get a seat of the pants feel for what's shaking with various hardware. I also consider games to be real world tests, since people don't usually just play benchmark all day. So that's just me, and how I succumb to and preserve my limitations - sad but true. I guess there's no end to research.

And yeh, working through an NDA is a grind. Almost as bad as working through biased reviews - it's so hard to find crystal clear tests; and even then the limits of my knowledge come into play. That's why I seek expert opinions from senior posters. After a while, the truth is absorbed via osmosis, or mere repetitive exposure; although I like to think there's some personal evolution involved, and hope that even that is not merely self-delusional. Hey, where's the bottom line here? :) I suppose we all go with what we got.
 
Its harder to muss up a game bench than a synth bench. We were given a lil more info on the gaming tests, and what they could mean, tho again, it could be fudged both either or one direction(s), but wouldnt show as great a change as a synth test.
The CB numbers are off for all three cpus, why? But theyre consistant, which may mean, if done optimally, it could possibly be apples to apples, but not enough info to prove it one way or another
 
You mention "optimally" several times. I assume you mean balancing the respective platforms as much as possible to level the playing field and avoiding advantage for one or the other? In otherwords, neutralizing the affects of necessarily different components to better focus on the particular components of comparison - ie the cpu's?
 


LOL
 


Pretty much clock per clock probably depending on the app and how well that app uses multiple cores.

And so now AM3 is 938? I read earlier that it was 939 pins. Weird. Of course it will change when AMD introduces tri/quad channel DDR3 later on. I think thats the SG34.

And a 5GHz anything is hard to beat. But then again since its not on air yet I doubt there is any reason to really care as it will be only for those OCing ppl who only do that and get paid. Luck bastards.

I think we may see 5GHz on air if Intels 22nm process can do 3.8GHz on air @ 1.1v. But thats still not until 2011 at the earliest and we have yet to see anything from AMD on their 22nm, and have yet to hear much about IBMs 32nm w/HKMG. I think tehy plan 32nm HKMG in what 2010?
 



Acually with the help of IBM AMD made a sucsessfully 22nm testing and is due at around 2010-2011 with 22nm.
 


Well, thanks for the recognition, although it isn't like I've written a true "article" or survey or something like that, but rather my common practice of always reading and searching anything I can about new tech and then sharing the results with people who might care about them.

Being an admirer of the Intel brand, I was sort of amused before Nehalem's launch. Not so much now, but I would happily own one if I had not to say goodbye to twice the amount of dollars the US guys do just because of a criminal tax that the Brazilian government demands for any imported tech.

Keep in my mind that I pay for a GA-EP45-UD3P what the Americans pay for a GA-EX58-UD5P, so I have a natural interest about new opportunities to get better performance for my money and, being this the case, I keep my "Intel liking" very away from me whenever I'm looking at benchmarks or using the credit card.

So far, my initial thoughts about Phenom II are:

http://www.xbitlabs.com/articles/cpu/display/core2duo-64bit_7.html#sect0

As we have expected, nothing serious has happened. CPUs with Intel Core microarchitecture and EM64T technology work normally in 64-bit modes. No dramatic performance drop has been detected in most benchmarks.

At the same time I would like to point out that it looks like Athlon 64 processors ensure higher performance increase when switching to 64-bit work mode. The average performance improvement we have seen from Athlon 64 FX-62 equaled 16%, while Core 2 Extreme X6800 demonstrated only 10% average performance boost. This way, there is a certain difference: AMD K8 turns out 6% mode efficient in 64-bit mode than Intel Core.

Phenom II should wide this performance gap while in 64-bit mode or at least conserve it (when talking vs Core 2). The article was written prior to Phenom's I launch, but, if anything, I would say Phenom I still has this or a slightly higher advantage, being this a consequence of design or the simple lack of Intel's macrofusion in 64-bit. Nehalem, however, should not have this little reduction in performance while switching between modes.

Please note that all of these CPUs improve or decline in certain areas while going 64-bit, but, as the article stated, is just a matter of averaging the results, so, it's not that "AMD is faster than Intel in 64-bit", but rather than AMD improves itself more than Core 2 does when operating this way.

Phenom II overclocks really well. Let us not talk about 1-second-stable-uber-overclock, but something that you and me can achieve with a half-decent cooling solution. Should the "previewers" and AMD prove themselves right, 3.6-3.8 GHz should be sort of an easy walk with a $25 cooler.

Probably, the stock voltage for Deneb is around 1.5 and, since it's a reasonably different 45nm approach from that of Intel, it should not be something to cry home about. Some say it can be undervolted and remain operating at stock speeds, so, expect at least 3.4 GHz at stock volts (if you are the lucky buyer of a good piece of silicon).

Phenom II will do far better than Phenom I in games and just reasonably better in most other tasks. AMD simply improved a lot the cache subsystem, but they can't do sorcery and, with just a die shrink, have the performance jump of a whole new architecture.

AM3 and DDR3 will bring better performance across the board, probably 2-5% in games, some more in things like WinRar, perhaps the same in video-encoding, and little to anything else.

However, if they manage to clock the northbridge at 2.6 with later models (like most people who got samples are suggesting), there can be another general increase of 5% in everything (supposing it's not because of the NB clocks that AMD is claiming the performance jump of AM3).

The same increase can be expected in AM2+ while "tunning" the NB (which has been described by the previewers as having "a lot of headroom").

Also, it's being said that Deneb is full HT3 compatible, but that current chipsets are not. Supposing that is correct, perhaps another 5% performance increase can be expected in well multithreaded apps with the arrival of RD8xx, or perhaps not.

Remember also that, later on, AM3 will get SB800, which is said to be SATA3 compatible (and maybe, just maybe, RD8xx will get PCI Express 3). Such chipsets are said to support Fusion, so, personally speaking, I wouldn't get me anything less than an AM3 board and CPU.

General numbers: Phenom II might be 5-20% faster than Phenom I clock-per-clock, but that greatly depends on the product you are comparing it against and the "operational conditions" while using the CPU.

Forget about video-encoding and image-editing: think only blue when doing this (and i + 7).

For anything else: go for the price. But that applies if talking about Phenom II and i7. Core 2 is awesome, but should only be seriously considered (in my humble opinion) if going Dual-Core or for an upgrade right now, since LGA775 is dead and, when things get a little more optimized for Nehalem (as applications becoming NUMA aware), Phenoms will improve along while Core will be left behind (unless we talk SSE4).
 
The major thing here is, just as i7 buyers are becoming aware of, theres so much tweaking Deneb can do that P1 couldnt do very well at, and wasnt worth the effort, and performance climbs from there, besides having thebetter starting point.
I agree with your assessment, and if priced right, it will bring some decent performance for those that dont want to spend more for i7. I also agree, that since both companies are going for 64bit, and will have the same numa approach, itll actually help AMD more than Intel. My reasons are this. Intel has always been the trend setter, and since its older arch was not native etc, those trends tilted towards Intels arch, and away from AMD. Now that has changed, and whats good for the goose etc etc, and it will benefit AMD like never before
 
I agree. My only complain about Deneb's initial data (and any future AMD CPU, probably) is the lack of SSE4 and, according to an article of Anandtech which talked about AMD's proposed SSE5, it will never grace the green CPUs.
 
Yea, its hurts when youre not the trend setter, one way or another. By not even trying to implement it isnt good at all, tho theyve played this game before, and maybe the benefits will be countered in different ways? Who knows? Do you have that link to AT?
 
The "Men in Blue" always find a way. ;D Trying to find some new information about Deneb, but the rumormill has been spinning slow lately.
 
The Q6600 has long reigned as the value choice, however the Q6600 is more of a high end office CPU now, but overclocking yields some gains . This is logical because the Q6600 is an outdated CPU because it's been around for so long and it feels like a lifetime. I'm not slating the Q6600 because it's been a legendary CPU and deserves that status, plus it's Intel so you know it will last for a longtime and not break on the day of purchase.

The Phenom has been in the Q6600's shadow, despite AMD calling them 'Black editions', but what disappoints me about AMD's 45nm Phenom II next generation Quad CPU is the performance isn't improved enough for them to challenge Intel. The old Q6600 is still giving AMD a rough time and unlike AMD the Q6600 has acceptable temperatures.

Intel has better capabilities when producing a die shrink too, AMD nearly always fail on first attempt. AMD release a CPU > It breaks > AMD comes back with an updated model to address the issues, while Intel is destroying them.

I wouldn't recommend AMD this Christmas or anytime to be honest, don't settle for second best BUY INTEl. If you can afford the extra costs you can't go wrong with Core I7, which destroys even Intels award winning Core2Quad series and you can imagine where that leaves AMD! Core I7 is just so dam fast and consider this is Intels first native Quad attempt :) compare to AMD's first attempt....
 


In spite of the Q6600 truth you just stated, you're wrong in so many ways about AMD and a CPU which we are yet to see that I can't even answer accordingly in the 30 more minutes I must spend here at work.
 
Core I7 is just so dam fast and consider this is Intels first native Quad attempt :) compare to AMD's first attempt....


No one wants to talk about this at all, but its true i7 is Intels first native
Quad core, compared to AMD's first & second native quad core, the i7 is
pretty damn good for intels first shot. AND NO AMOUNT OF SPIN WILL
CHANGE THAT!!!
 

Latest posts