AMD Phenom II Review Thread

[panfist]

No one who were looking for the fastest had to wait for the reviews or the launch, the MSRP said it all: the X4 940 is a sub $300 chip...

This is xtremesystems.org, so I'm assuming we all know how to overclock. You can get a sub $300 chip from Intel that will beat the 940, hands down.

If you're not overclocking, you're wasting money whether you buy AMD or Intel.

[Mats]

You can get a sub $300 chip from Intel that will beat the 940, hands down.

That's why I said:

Wait two months and see who has the best bang per buck chip then, when prices settled.

What I meant was that the prices doesn't really tell us what they're worth, yet.
If Intel has a better product, then they will cost more. Most likely.

Then again the 940 is easier to overclock, raise the multiplier and you're done.

[Rammsteiner]

Weren't you the one who pointed out to me that this is XtremeSystems.org not xtremebugetsystems? Most folks I know who own AM2+ own 800 and NOT 1066. So should they upgrade RAM?

Come on, there's like no price difference between DDR800 and DDR1066. I just got a DDR1150 kit for what, only 100 Euro

As I pointed out earlier, there ain't a lot wrong with DDR800, but at least get your timings sorted. Those reviews are done with DDR800 tCL5. DDR800 does the job though, but dont start logging clueless numbers if you're using under rated hardware which have great impacts on performance.

And regarding upgrading to DDR1066 after all, it ain't needed and actually wouldnt be worth it (unless you dont care and like buying new hardware). I would save that money for a next upgrade or something to be honest. But seriously, dont go post numbers everywhere while using under rated equipment.

[AbelJemka]

Big mistake from me
I misinterpret an article i read
http://www.behardware.com/articles/7...-gtx-280s.html

[Particle]

Testing with DDR2-800 just isn't proper these days. Even if the difference is small, the fact remains that DDR2-1066 memory is absolutely pedestrian. Anyone buying into the platform now for Phenom II is going to be getting DDR2-1066 unless they are completely clueless. In the enthusiast circles (the people interested in upgrading from a current AM2+ platform), there's also a high likelihood of already having DDR2-1066. Eight freakin' gigabytes of CL5 DDR2-1066 now costs $100 flat, so it isn't like the upgrade is expensive or outlandish. Anyone reviewing the Phenom II is providing information to people who are knowingly buying into the end of a platform. Soon all PII chips will be superseded by AM3 models. Anyone wanting to use DDR3 will have to chuck their current Phenom II anyway. There's no reason this late in the game to showcase the chip outside of its best light--it's proper light. This means not being retarded: Use some cheap-as-dirt 1066 as it best represents the product with average, cheap support hardware.

Even if the difference between DDR2-800 and DDR2-1066 is 0.05% on average, it's the proper type of memory to test with at this point since it's the standard. Even Phenom was 1066 compliant with two sticks.

[eleeter]

Even if the difference between DDR2-800 and DDR2-1066 is 0.05% on average, it's the proper type of memory to test with at this point since it's the standard. Even Phenom was 1066 compliant with two sticks.

But don't you see, Kyle can't get PH2 to work with DDR2-1066.

But for some reason he thinks he is the best reviewer on the web, and his methods far exceed what anyone else is doing.

[Shintai]

Memory configuration in HardOCP review is more than unfair is invalided the test!
Why? because each 280GTX in sli will adressed 256MB in main memory. 3 cards 768MB will be reserved. 790i and probably 780a will reserved 256MB for itself.
- PhII rig has 2GB and 780a : 1GB remaining for applications.
- Core 2 Quad has 4GB and 790i : 3GB remaining for applications.
- Core i7 has 6GB and X58 : more than 5GB remaining for applications.
Just need one to test Far Cry 2 with 1GB versus Far Cry 2 with 2GB

Thats one big pile of jibberish and so far from the truth it hurts.

[alfaunits]

Man, that is such nonsense it's so good to put you on ignore list.

Memory configuration in HardOCP review is more than unfair is invalided the test!
Why? because each 280GTX in sli will adressed 256MB in main memory. 3 cards 768MB will be reserved. 790i and probably 780a will reserved 256MB for itself.
- PhII rig has 2GB and 780a : 1GB remaining for applications.
- Core 2 Quad has 4GB and 790i : 3GB remaining for applications.
- Core i7 has 6GB and X58 : more than 5GB remaining for applications.
Just need one to test Far Cry 2 with 1GB versus Far Cry 2 with 2GB

[alfaunits]

Where did I state that? If I'd ever buy 3 GTX280 I'd go for what gives me the best performance, no matter the costs: i7
What I'm speaking of is price/performance, seems like you failed to notice that.

IIRC, you were commenting on the [H] review, so it made sense to presume we'd take the Tri-SLI GTX 280 into account.

And where did I state that it makes sense for C2Q-owners to change to a Phenom II? I didn't, did I?

I did not say you did. That entire part was just my comment.

They are under "Realitätsferne Spiele-Tests", what means they aren't close to reality... This just once again shows how different the results can be.

So, why are they even mentioned?

[Enisey]

The person who can afford 3 way SLI of GTX280 won't even look in the direction of PII or Core2, so the testing on the HardOCP while interesting was kinda pointless (in terms of assessment of PII vs Core2) and wasn't a real life scenario. What would be good to see is a minimum FPS numbers on max settings with 1 or 2 videocards. Such data would be a lot more useful.

[Zucker2k]

The person who can afford 3 way SLI of GTX280 won't even look in the direction of PII or Core2, so the testing on the HardOCP while interesting was kinda pointless (in terms of assessment of PII vs Core2) and wasn't a real life scenario. What would be good to see is a minimum FPS numbers on max settings with 1 or 2 videocards. Such data would be a lot more useful.

QFT.

To see exactly what is going on, this is what needs to be done:

Test all cpus with current games with a x2 card,

Test all cpus with current games at same settings with 2x x2 cards.

[VoodooProphetII]

The setup is a 9950 @ 3.3Ghz
Asus M3A79-T Deluxe
2GB Crucial Ballistix clocked at 900mhz with 4 4 4 12 timings
EVGA GTX280 (The card was a little overclocked at 640mhz core)

Far Cry 2 - 640x480 very high settings. When I set everything at it's lowest I get about 5 FPS more.



Now doesn't [H] review show more a problem with Tri-SLI scaling on a 780a NV chipset mobo, then a problem with the processor? I'm just asking, as I feel normal SLI might have worked a little better in the test looking at the results I got with a Phenom I clocked to 3.3Ghz. I know the HT is higher, but that could have been done with the review as well.

[cegras]

Thats one big pile of jibberish and so far from the truth it hurts.

Out of curiosity, how would the addressing be assigned in each case then?

[Donnie27]

The person who can afford 3 way SLI of GTX280 won't even look in the direction of PII or Core2, so the testing on the HardOCP while interesting was kinda pointless (in terms of assessment of PII vs Core2) and wasn't a real life scenario. What would be good to see is a minimum FPS numbers on max settings with 1 or 2 videocards. Such data would be a lot more useful.

QFT ditto!

In real life or real world most folks have one Nv-260 or one 4870. So we'd need a pretty good Xtreme Systems Video card poll! I have a feeling (and that's moot BTW) reviewers using one high-end card is closer to real world and than someone using two X2's from Nv or AMD.

Guys, why all the Kyle bashing? Where were all the complaints he got for his BS Conroe launch? I wish I had made copies of the text before he revised it. Sure those tilting toward AMD thought he did one of the better reviews.

Memory performance, there is some truth to 2GB vs 4GB. If P.A.E. is enabled it can cause slow downs and make 4GB seem slower than 2GB.

[AbelJemka]

I asked Behardware explanation and i understand my mistake
If i correctly understand now memory assignement for each card and chipset not exceeding 4GB limit so memory is fully usable by applications.

[Shintai]

I asked Behardware explanation and i understand my mistake
If i correctly understand now memory assignement for each card and chipset not exceeding 4GB limit so memory is fully usable by applications.

Well there are 4 options basicly.

First tho remember. They reserve addressspace. They do not use it as physical memory. And a 280GTX card maps 768MB actually.

1. They used a 32bit OS and no remap of PCIe, etc (Ancient days so to say). In that case the PH2 system would have 1½GB like the Core 2 and i7. To be exact you need to calculate all devices etc in the box.

2. They used 64bit OS and no remap. This is still very very very unlikely. Anyway, that would leave 1½GB for PH2, 1½GB for the core 2 and 3½GB or so for the i7.

3. 32Bit OS with PAE and remap. Unlikely tho. But the second least unlikely. In this case it would be 2, 4 and 4GB assuming no shardy artificial limits like in XP (Tho still artificially 4GB limit). i7 would actually have 6GB on a 32bit windows server OS (Enterprise version or datacenter).

4. 64Bit OS and remap. (Most likely. I dont even think i7 and PH2 would support anything but remap.). In that case its 2, 4 and 6GB ofcourse.

Also remember that the games tested are 32bit and most likely not compiled with being large address aware. In short, they can only use 2GB. Even tho a 64bit OS would allow a 32bit application to use all 4GB.

Remap is a feature that was new some 3+ years ago on the desktop. For those even allowing you to disable it.

Here is a funny example if you can disable remap. Used it to tell people that 64bit OS alone aint enough.



But rememebr remap is ON nowadays. You need ancient HW for it basicly.

[AbelJemka]

I have completely understand now

[Donnie27]

QFT ditto!

In real life or real world most folks have one Nv-260 or one 4870. So we'd need a pretty good Xtreme Systems Video card poll! I have a feeling (and that's moot BTW) reviewers using one high-end card is closer to real world and than someone using two X2's from Nv or AMD.

Guys, why all the Kyle bashing? Where were all the complaints he got for his BS Conroe launch? I wish I had made copies of the text before he revised it. Sure those tilting toward AMD thought he did one of the better reviews.

Memory performance, there is some truth to 2GB vs 4GB. If P.A.E. is enabled it can cause slow downs and make 4GB seem slower than 2GB.

[Shintai]

Memory performance, there is some truth to 2GB vs 4GB. If P.A.E. is enabled it can cause slow downs and make 4GB seem slower than 2GB.

Uhm. No. And PAE is on in XP SP2 and up. Nomatter how much or little memory. Mainly due to PAE is needed for NX.

[Donnie27]

Uhm. No. And PAE is on in XP SP2 and up. Nomatter how much or little memory. Mainly due to PAE is needed for NX.

But you can turn it off, right? I notice in Vista64bit I didn't see PAE enabled. Don't need it for 64bit? Where is it? System Properties shows it in WinXP32. (also Bootini IIRC)

[Shintai]

But you can turn it off, right? I notice in Vista64bit I didn't see PAE enabled. Don't need it for 64bit? Where is it? System Properties shows it in WinXP32. (also Bootini IIRC)

64bit doesnt use it since they already use 40-48bit addressing or so. PAE is simply 36bit hardware addressing.

PAE is on by default, hence not in the boot.ini file.

There is also no slowdown with PAE. Unless you use AWE applications like SQl. But the benefit is bigger than the penalty.

[Miss Banana]

Guys, why all the Kyle bashing?

Does that need any kind of explanation?

Using slow speed ram in a review for the platform you dislike, 2 GB instead of 4 GB ram, and then acting like an immature lame fanboy in the entire review is just not going to get you any oscars.

It does not even matter whether or not people who buy AMD would upgrade memory, neither is it relevant that some programs Kyle used to bench do or do not use the full 4 GB.

A person that reviews should at least make some sort of effort to appear to be giving objective info about a product. Not all reviews are perfect, but some of Kyle's reviews are so idiotic that it makes me wonder whether his entire site is ment as some sort of joke or parody on biased reviews.
Sadly he is dead serious.

[savantu]

Does that need any kind of explanation?

Using slow speed ram in a review for the platform you dislike, 2 GB instead of 4 GB ram, and then acting like an immature lame fanboy in the entire review is just not going to get you any oscars.

It does not even matter whether or not people who buy AMD would upgrade memory, neither is it relevant that some programs Kyle used to bench do or do not use the full 4 GB.

A person that reviews should at least make some sort of effort to appear to be giving objective info about a product. Not all reviews are perfect, but some of Kyle's reviews are so idiotic that it makes me wonder whether his entire site is ment as some sort of joke or parody on biased reviews.
Sadly he is dead serious.

You haven't' presented the slightest evidence that his numbers are off or that the benchmarks he used were affected by the memory difference.The burden of proof is on you.

FYI , reviews aren't made exclusively for XtremeSystems users ; I'm sure most of the people who visit review sites don't know what RAM timings are.People will buy a prebuilt system that probably has the cheapest mobo and cheapest RAM , made on a ship between ASIA and Europe/US.They saw the CPU sticker or asked for a system built around a specific CPU.

In the real world nobody cares about DDR3 2000 CL 6 , 1.8V Vcore , QPI voltage , etc.

[Calmatory]

Come on, there's like no price difference between DDR800 and DDR1066. I just got a DDR1150 kit for what, only 100 Euro

As I pointed out earlier, there ain't a lot wrong with DDR800, but at least get your timings sorted. Those reviews are done with DDR800 tCL5. DDR800 does the job though, but dont start logging clueless numbers if you're using under rated hardware which have great impacts on performance.

And regarding upgrading to DDR1066 after all, it ain't needed and actually wouldnt be worth it (unless you dont care and like buying new hardware). I would save that money for a next upgrade or something to be honest. But seriously, dont go post numbers everywhere while using under rated equipment.

In which cases does DDR800 -> DDR1066 have great impacts on performance? And what are those great impacts? 1-2% difference?

Oh wait, here in XS, if CPU_A is 3 % faster than CPU_B, then CPU_B is being "completely destroyed" by CPU_A. Wtf people, get the cheaper one and be happy.

[eleeter]

Not all reviews are perfect, but some of Kyle's reviews are so idiotic that it makes me wonder whether his entire site is meant as some sort of joke or parody on biased reviews.
Sadly he is dead serious.

LOL

So true.