AMD Phenom II 940 "Xtremely" Benchmarked

Page 40 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
It has been proved with facts and figures that a nb increase of 200mhz will not nett 4-6% ipc, You guys just keep the circular arguments going and ignore this fact. Think about it, If you got 4-6%ipc from a 200mhz bump on the nb that would be near as efficient as raising actual core clocks ffs. This is the last I will say on this as I realise no matter what facts and proof are provided it will not be enough so wee will revisit this thread in due time.
 


Hey, pal! Man, how I do enjoy this setup... and I bought it just to be a temporary step until a Core 2 rig + 4870/GTX 260. I'd be happy to use it during all the lifetime of Windows 7, this I assure you. It's just that, having the compatibility - and $ - to upgrade to an interesting option, I think it's a decent time to do it.

http://www.forumpcs.com.br/viewtopic.php?t=238565&postdays=0&postorder=asc&start=10

There are some benches on this Brazilian review, however, they are in Portuguese (but I guess you can use that translator from Google, which link I just don't happen to have right now). My friend (one of the posters, Alexandre Zielbert) runs a very famous hardware store around here and also does many reviews for that site. The last one was a very nice piece of information about Nehalem. Lately, I have talked *a lot* to him about Phenom and Phenom II, and he said the difference between AM2 and AM2+ is negligible on 90% of the time, unless you happen to use Crossfire, for instance.

Interesting enough, he always said that "Phenom just feels faster", although he uses - and enjoys - a budget Core 2 Duo rig at home.
 


Well, I'll try to find you more links with a direct comparison between NBs running at 1.8 and 2.0, however, 4-6% is simply doable if coupled with DDR3. Core 2 already shows this improvement when going for DDR3 1333, although it is recognized as not being sensitive to memory performance, so, why is it so difficult to believe in a non-earth-shattering 4-6% when a CPU will have a brand new memory controller running at higher frequencies than the one before it?

Personally, I think 4-6% simply "can't be felt", anyway. I do not think it's a reason to wait for AM3, but one to be aware of if you already like what you see on Phenom II.
 
Besides, keep in my mind that we do not know if the NB will be clocked at 2.0 GHz, since that is simply the minimum value if the HT link will run at 2000 MHz. Many are talking about the NB being clocked at 2.2 GHz on Xtremesystems. Given all their earlier "guesses", I have no reason whatsoever to doubt.
 


I would hope a CFX system would be faster with a AM2+ than a AM2 since it should be able to push more data. Much like how a Core i7 blows everything away with CF/SLI since it can push much more data through.

I wouldn't be suprised if with the newer CPUs we might see better scaling with CF/SLI that make it a "viable" solution compared to what it has been, which is not worth the money.

As for feeling faster, its all preference. I could say the same about my Q6600 setup. And I had a bigger jump than your friend from a budget (which kinda kills it comparison wise) Core 2 Duo to a Phenom. I had a Pentium 4 Northwood EE @ 3.4GHz that was 6 years old. SO to me my Q6600 system feels 100x faster.

But as I said its all preference. Just like my friend. He had a old 1.8GHz single core Athlon XP. For Christmas he got some money and wanted a new CPU/mobo/RAM. Funniest thing is that at Newegg the cheapes combo was a Intel combo. He got a Core 2 Duo E7300, 2GB of DDR2 with a Foxconn mobo. To him that system feels much much faster than his old one (it was a S939 I think).

But overall I don't think there will be an amazing jump in performance or any for that matter unless its in certain programs (ones that utilize memory bandwidth such as server programs or maybe multi GPU gaming) which can't be taken as the overall performance increase.

Much like Core i7. Overall its a good performance jump if you do everything. But only in multi card gaming does it shine for gaming. Single card gaming it gives all it can give but one GPU is too slow so it doesn't shine as much as it does in other areas.

There is a lot you have to consider when it comes to CPU performance too. What I find interesting is back in the ols Athlon X2 days, synthetic benchmarks were shoved down our througts. Now some people don't want to use them at all if it makes C2Q/Ci7 look better.

I personally never look at them. I look at the gaming performance and other things like video/compiling performance or things that matter to me really. But still. I always find it strange.
 
I agree, Jimmy. With everything. It's all about preference. But, given he's a knowledged user of both systems, I guess it's a little more than just the "placebo" effect. My opinion is that i7 should give the same "feeling", perhaps because of the IMC (plus the superior performance across the board, of course). I'll ask him next time I go to his store.

As for the syntethic benchmarks, some give a nice perspective of the chip, but that doesn't usually translate to gains in every application. Memory related ones may translate to the "feeling of responsiveness", but the debate would be eternal, I guess.
 

You are not listening to what I am saying, I am not saying that ddr3 will not gain whatever percent and I am not saying that am3 will have no benefits, AM3 could very well be 4-6% faster but I am saying that a nb bump of 200mhz will not nett 4-6% that is all.
 


Maybe, depending on the benchmark. On average it might be a little lower, but I'm pretty sure things like archivers and video-encoders should feel it at least by 4-6%. Maybe, as I said.
 


I think what you're saying is fine. You're predicting a 4-6% increase and you have indicated that by saying, "I'm pretty sure" and "Maybe". My problem with Jay's post is that he just flat out states that there will be a 4-6% increase. He doesn't qualify it at all, just states it as fact.

Another question. I'm hardly an expert with Phenom 2 but I did recall reading that for best overclocking AMD suggests lowering the HT link speed to 1ghz. If what everyone has been saying on the past few pages is any indication, wouldn't this cripple the performance of the processor? Sorry if this is a dumb question.
 


I agree with you.

3dMark Vantage was only something I jumped into because it was loaded and easily available.

I have now done the Phoronix-test-suite "universe-cli" test with the NB at 1.8Ghz and at 2.4Ghz:

You can see the results here:
http://www.xtremesystems.org/forums/showpost.php?p=3593897&postcount=6

I may try to get the full "universe" test with the games done in the future. (Probably after the next ATI Linux driver is available.) I might also do a 2.0Ghz NB run because apparently that seems to be an important frequency in this thread; but then I might not bother because any "enthusiast" wouldn't be running with a NB that slow... they'll be using 2.4Ghz or higher.

OH AND SORRY: I've not been able to play in this forum's games lately. I've been too busy actually doing things. But I thought I'd give you a few more results.
 
...OK, I've just got one Phemom II 940. On first impressions - I'm happy 😀 ;
I've tried just a quick clock with multiplier form 15 to 18 and 1.35V to 1.4V and the freq gone 3.6Ghz - very easy clock. I'll go for further clock later. temps are as follows: 32C idle, 48C full load with prime (cool&quite:disable)
I'm going for another bunch of test now 😀
 


And let's not forget that QFX debacle, either. I'm surprised there wasn't some sort of class-action lawsuit about that one. Oh wait - BaronMatrix forgot to buy his so I guess there was only 1 customer, and you need at least 2 for a "class" action 😗
 


That's exactly what it is: just a prediction of a forum member who tries to have some insight by reading a lot. We may be in for a threat or for lots of compatibility and performance issues, like their transition to DDR2. Given their recent record, I'd bet on the former, but, as we both said, there's no way to really know until it hits the shelves. With regards to the HT link, the loss of performance is minimal, unless we talk server/highly multithreaded apps. My friend has ran some tests on a 690G AM2 mobo (my A6VMX, to be more specific) with the HT link set at 1000. The loss of performance ranges from 0 to 10%. The late was in a syntethic bench, if I remember correctly, but no "true" apps I can think of had a loss of more than 3-5%. But things like Crossfire/SLI would be quite crippled, that's for sure.
 


Oh, that's nice, pal! ;D Would you mind posting a few tests for our curious minds?
 
Games, and encoding. Real world apps would be great
The reason AMD suggests the HT at 1000 is because thats for using LN2 and not for day to day. Toms ignored this info, and failed to get numbers Ive seen with people using far less than LN2
There isnt just 1 thing thatll bring the perf up for AM3, and the 200 increase is the proposed one, tho not written in stone. Remember, most reviewers dont know what theyre doing regarding AMD, so just take some time and focus more on Anands, as he did it himself, and he wont let incompetance stand in the way, regardless of time needed. His and TRs were the best Ive seen, and reflects the knowledge and versatitlity of their respective sites.

Anyways, all this means is, AM3 will edge closer to Yorkfield, where the user really wont be able to tell the difference between the 2 cpus, and thats supposed to be good. If you expected P2 to kill Intel, then go ahead and be disappointed, it didnt happen, but if youre in the market for buying a new setup, and want some future compatibility, its hard not to look AMDs way this time around
 


Word, Playa.
 
Keith thanks for the link - it does make a bit of difference with the higher NB clocks.

I suspect L1 / L2 / L3 cache latency improvements as the design matures will then also have quite an effect.

Not a bit improvement in IPC compared to the core2 line ... but an improvement not to be sneezed at.

Upping the NB speed on the PII would seem to have the most effect for server type operations ... not so much for gaming though.

Bummer all servers seem to run on "rock solid" but slower RAM.

Can you run a few game comparison? Perhaps 3DMark too?
 


I already ran 3DMark and had too many people wanting something else. Besides I'm mainly doing my benchmarking on Linux.

As for game benchmarks... I'm waiting for the 9.1 ATI Linux drivers which should arrive in the next day or so. Then I can run all the Phoronix universe-x applications which includes a dozen or so games and other graphical applications like lightsmark and unigine.

Actually I could run them right now... but knowing that a new driver is going to be out in a day or so makes me hesitate. (Because I'm actually running these benchmarks for my edification not for other people... if others can gain benefit that is a bonus.) While waiting I installed a newer linux kernel and a newer version of phoronix.

I may or may not do the NB 1.8GHz versus 2.4Ghz tests again. I've already proven (to myself at least) that it makes a difference.

All of these memory and NB tweaks that some people claim "don't make a difference" add up when you do them all together. If someone told you that you could do system tweaks and gain 3% to 40% on some of your benchmarks then you would definitely consider them important. (Although I know that there are still some people that want to believe that neither tweak is important. But these people are easy to ignore.)
 


It's not just DDR3, it's not just the HT link, it's also the 880G motherboards. That alone makes me want to wait a month or two.

Rumors about work in the press had me worried, but I can afford the upgrades after all, especially with the price cuts. A Phenon II 940 for the original price of a 920. I don't feel that anyone lost out buying early but I can understand how they'd be a bit miffed.