PCIe And CrossFire Scaling: Does Nvidia's NF200 Fix P55?

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

ta152h

Distinguished
Apr 1, 2009
1,207
2
19,285
[citation][nom]Crashman[/nom]Sir, to clarify further the X58 platform required more voltage to reach 4.00 GHz and consumed far more power in this comparison. In fact, the 870 test CPU operates at up to 4.3x GHz using air cooling with HT enabled while the 920 comes up short of 4.1 GHz.[/citation]

First, don't flatter yourself and assume I was looking for a conversation with you. I was pointing out the flaws in your methodology, which I stand by.

You obviously have a bum chip, as most of the initial reviews of the Lynnfield showed it needed considerably more voltage at the same clock speed.

Also, you seem to be changing your own argument. You are talking about the platform, or the processor???? The processor needs less voltage. Individuals will vary, but the bulk of the Bloomfield PROCESSORS use less voltage and overclocked very slightly higher (assuming C0).

The reasons for this should be obvious, since you don't have the PCIe on the processor, and it's only less thing that can limit clock speed, and generate heat.

The x58 chipset itself uses a lot of power, but that's why your review was incomplete. If you're going to get the lobotomized P55 platform, and then try to make it work as well as the x58 (which you also diminished by using dual channel memory, without so much as mentioning in it passing), you run into the issue of why not getting the better platform to begin with. Did you still end up saving money? Did you end up using less power after adding the chip? You failed to answer these two very important questions.

The results for the chip were still informative, and I am honestly surprised it did as well as it did, but you failed to tie it into the bigger picture - does this setup make sense for someone looking to buy. If it's still cheaper, and it still uses a decent amount less power is it worth being slightly slower (you also left us hanging with how much slower since you didn't test the Bloomfield with three memory modules, which is how they will typically be used, unless you are trying to make a point that the platform isn't the best one).

I'm done discussing this with you and reading this line. You can either correct your mistakes, and see them, or you can assume I'm a fool and asking for irrelevant information. It makes no difference to me.
 

youssef 2010

Distinguished
Jan 1, 2009
1,263
0
19,360
@ta152h I don't think it would make any difference if he used triple channel or dual channel as there's no game that will use more than 2 GB of ram (Dual channel is what games are optimized four as no one will waste time,effort & money to optimize a game for a far less common platform).

oh,BTW,Why are you so aggressive in your comments?
 
G

Guest

Guest
@ youssef 2010 - software don't need to optimize a game for dual/triple channel - this is not like processor where muli core optimization are required.

tripple channel is handled by the memory controller and transparent to the software - your game don't need special code path to get that extra perf, and you don't need to access more than 2gb to get that extra perf.

think of it as a hardware raid 0 where your program or even copy file get speed up right away - you don't need a file that's large enough to span across each hard drive to get the benefit
 

Crashman

Polypheme
Former Staff
[citation][nom]ta152h[/nom]First, don't flatter yourself and assume I was looking for a conversation with you.[/citation]

This is a place for conversation. If you don't want conversation, leave.
 

youssef 2010

Distinguished
Jan 1, 2009
1,263
0
19,360
[citation][nom]youssef 2010[/nom]did any games show meaningful games by going from dual to triple channel(just post the link if you encountered any)[/citation]

there's a typo as I meant gains not games
 

dragunover

Distinguished
Jun 30, 2008
112
0
18,680
"Non-gamers will make non-gaming arguments, such as the future availability of hexacore processors for X58-based platforms,"

Here's your fix.
*Competitive gamers will make gaming arguments, such as the future availability of sexacore processors for X58-based platform*
 
G

Guest

Guest
I agree with Ta152h.

It does not matter for the results, but it's plain stupid not to use triple-channel memory for the X58-setup. Simply, because most if not all of those X58 buyers always buy three memory modules with it. They want triplechannel; how lame the performance improvement is, does not count.

Why don't you show more the performance potential with thriple-channel. Like you show the useless potential of cpu's, running games in 800x600 with low settings to show "the superiority of one chip above the other" ?

The first results have some value in a away: that extra bandwidth can actually help in the days to come. Game benchmarks at 800x600 will always be useless; nobody plays at those stone-age resolutions!

The review is not that bad either, it shows that nf200 helps CF/SLI alot with highend gaming setups. That you don't need a X58 for that purpose.
But to me X58 have more advantages, beside full x16 ports.
 

Crashman

Polypheme
Former Staff
[citation][nom]youssef 2010[/nom]No one can do everything alone.I guess Nvidia learned that from the Core2 days,when intel's chipsets were only licensed for crossfire.[/citation]

It wasn't a licensing issue, it was a driver issue. ATI enabled CrossFire on ALL chipsets (no license needed by the chipset manufacturer) while Nvidia disabled it on any chipset that didn't carry the SLI logo (even its own). I even modified an early nForce 4 non-SLI chipset to enable the SLI tag for the driver.
 

andy5174

Distinguished
Mar 3, 2009
2,452
0
19,860
[citation][nom]Jediron[/nom]I agree with Ta152h.It does not matter for the results, but it's plain stupid not to use triple-channel memory for the X58-setup. Simply, because most if not all of those X58 buyers always buy three memory modules with it. They want triplechannel; how lame the performance improvement is, does not count.Why don't you show more the performance potential with thriple-channel. Like you show the useless potential of cpu's, running games in 800x600 with low settings to show "the superiority of one chip above the other" ?The first results have some value in a away: that extra bandwidth can actually help in the days to come. Game benchmarks at 800x600 will always be useless; nobody plays at those stone-age resolutions!The review is not that bad either, it shows that nf200 helps CF/SLI alot with highend gaming setups. That you don't need a X58 for that purpose.But to me X58 have more advantages, beside full x16 ports.[/citation]
Game benchmarks done at low resolution is NOT useless. Their purposes are to show how well a CPU would perform with graphic bottleneck eliminated.

 

Crashman

Polypheme
Former Staff
[citation][nom]andy5174[/nom]Game benchmarks done at low resolution is NOT useless. Their purposes are to show how well a CPU would perform with graphic bottleneck eliminated.[/citation]

You're right. By focusing on realistic gaming resolutions, the article has failed to deliver that aspect of system performance. An article that focuses on PCI-Express performance should always have identical hardware (including dual-channel mode for all motherboards) and, when another component is added (such as the NF200 bridge) should also include lower resolutions to show whether the added hardware is drawing excessive CPU resources.
 
Status
Not open for further replies.