NextGen VS. CurrentGen CPU: 3DMark06

[cookerjc]

here comes 3Dmark06:

From the final result, Phenom only lose QX6850 about 5.5%. But the CPU Score is losing 9.11% . If the rival change to QX9650, the gap will even bigger. If we compare to other subitem, that dual core E6850 will boost Graphic system better than Phenom X4. Comparing to Phenom X4 and E6850, All the advantage score are coming from the CPU score. Phenom X4 does bad in HDR and SM subitem tests.

In fact, LOTs of games are are not have the enough dual core or quad core optimize than 3Dmark06. In other test, you may find out even a dual core Core 2 is better than Phenom X4. That means not enough optimize support for our Phenom X4.

http://www.expreview.com/news/hard/2...85d6811_1.html

[Kasparz]

Why you can't run Phenom with normal ram speeds and timings?

[metro.cl]

Why you can't run Phenom with normal ram speeds and timings?

Maybe divider issues since they run Phenom at 3.0GHz and that is definetly overclocking it (like 20%+)

[Kasparz]

huh?
http://www.xtremesystems.org/forums/...4&postcount=80
Why he used 7 tRP value then? Its like umm...downclocking? Is there any problem to tighten tRP?

[zanzabar]

ive been looking at these and what is the chipset performance difference and what about vista were u can actually use the entire quad at teh same time efficiently


but good post space man

[Epsilon84]

I think he mentioned in another thread that CPU-Z was misreporting the timings even after he set 5-5-5-18 in BIOS.

[expreview]

huh?
http://www.xtremesystems.org/forums/...4&postcount=80
Why he used 7 tRP value then? Its like umm...downclocking? Is there any problem to tighten tRP?

CPU-Z problem,All setting is by SPD,means to be 5-5-5-18.

just wait CPU-Z update.

[expreview]

Maybe divider issues since they run Phenom at 3.0GHz and that is definetly overclocking it (like 20%+)

yes. the Phenom was oc on 200MHzX15,and the RAM frequency would be:

3000/7=428.57MHz or 3000/8=375MHz.

[The0men]

It does seem slightly wrong to dissadvantage AMD by running slower Ram than the Intel, its not the winning difference, but still. Same as the Crysis benchmark.

[Epsilon84]

It does seem slightly wrong to dissadvantage AMD by running slower Ram than the Intel, its not the winning difference, but still. Same as the Crysis benchmark.

It's a divider issue @ 3GHz, either the RAM will run faster or slower than the Intel system.

I actually would rather Expreview run the RAM @ 3000/7=428.57MHz, who cares if it's slightly faster than on the Intel system, at least we won't get all the whining and crap about AMD being disadvantaged (not a slur at you, just what I've noticed from the Crysis threads).

Then again, we'll probably get Intel fanboys chiming in instead that Intel is then disadvantaged, even if Intel wins the benchmark. LOL

[JumpingJack]

Maybe divider issues since they run Phenom at 3.0GHz and that is definetly overclocking it (like 20%+)

It is precisely the memory divider issues, any non-even CPU multiplier will down clock the memory at stock system clock speeds (200 MHz). Don't you feel cheated

This is a well known phenomena (pardon the pun) of AMD's memory controller, most AM2 reviews made note of this:

Example: http://www.anandtech.com/cpuchipsets...px?i=2762&p=10

For a multiplier of 15, the fastest a Phenom will clock DDR2-800 will be 375 Mhz not 400 Mhz.... so what you see is what you are gonna get. To get 400 Mhz on the ram, you will need to force the DDR2-800 ram to 428 (i.e. 3000/7 = 428 and thus risk stability) or buy DDR2-1067 and force it do DDR2-800 (i.e. 3000/7 = 428 running psuedo DDR2-856), but then you would probably want to multiply that memory up to to closest to 1067 as you can get, but it too will underclock it's full rating.

Everyone should note -- if you purchase an AMD CPU with a non-even CPU multiplier, the memory will be down clocked from it's best possible... this is just the way the AMD memory controller works.

The short answer -- the memory speed reported above is exactly where you would expect an AMD CPU to run DDR2-800 memory with a CPU multiplier of 15...

Jack

[justapost]

yes. the Phenom was oc on 200MHzX15,and the RAM frequency would be:

3000/7=428.57MHz or 3000/8=375MHz.

Is 3000/6=500MHz with 1066 DDR2 not possible?
Why is the northbridge running at 16000MHz, like on a non split power board?
ES issue?

[RaZz!]

It does seem slightly wrong to dissadvantage AMD by running slower Ram than the Intel, its not the winning difference, but still. Same as the Crysis benchmark.

idd!

i'd like to see another test: downclock the intels to be on par with the phenom @stock, not the other way around.

[Epsilon84]

idd!

i'd like to see another test: downclock the intels to be on par with the phenom @stock, not the other way around.

What difference would that make? Still clock for clock.

[AbelJemka]

I think he wanna see scaling capabilities.

[Epsilon84]

I think he wanna see scaling capabilities.

Oh yeah, according to some people, K10 introduces superlinear scaling... +10% clock = >+10% performance

[Vapor]

Well I'm very impressed with Yorkfield...stock clocks, highest score of the bunch, miserly power draw.

[kiwi]

Don't forget Yorkfield easily oc to 4ghz+ stable on air

[BeardyMan]

Now those are the resulst i was waiting for

But what's teh platform?
mobo wich chipset?

[Epsilon84]

Now those are the resulst i was waiting for

But what's teh platform?
mobo wich chipset?

RD790 dude... open the link.

[BeardyMan]

RD790 dude... open the link.

if i only could

[tictac]

memory speed for k10 no longer calculate with cpu clock. wait for new version of cpuz or you can use everest to show memory speed correctly..

[knightwolf654]

grrr stop teasing!!! i want a phenom!!!!

[GoThr3k]

Why is the northbridge running at 16000MHz, like on a non split power board?
ES issue?

i was thinking same thing

[zakelwe]

It seems a long time ago now that the summer was about to start and the K10 was 40% faster than Intel. I had the impression that it was not going to be 40%, more like 15%-20% and the main question was whether it would be able to overclock enough, not that it would lose in IPC.

Now it seems that for IPC it is about level or slightly behind ( with current testing setups ) and therefore not enough to offset any disadvantage with overclocking. The prices seem OK though. Disappointing it has to be said. AMD would have been better never to mention the 40%.

The main problem I see is that AMD have only got a die shrink coming up to try and rescue the situation, and historically die shrinks have never been that good for them. Meanwhile Intel is going to be throwing an integrated memory controller and their new bus ( whatever it is called this week! ) which should give a large boost, if AMD's K8 is anything to go by over K7. And we know the Intel 45nm process is good now and will be even better by Q3 2008.

AMD really needed to get the lead back at this point for Intel to pinch it back and then AMD to ..etc etc. That's not going to happen it seems.

Regards

Andy