INTEL - Synchronous is not important

[infinite012]

I don't really care for all this though. If you can get a higher FSB by using a lower ratio, do it. But if you can get a higher benchmark/real world performance by using a lower FSB and 1:1 ratio then use that instead. I REALLY DON'T CARE!!

 

[Slappy00]

Quote: "Actually liked Pascal over C+, but guess which one won...."

har har you are the old-guy #1, I took pascal in high-school, that linear stuff is a pia imo.

With DDR 4200 out now this argument should be for the most part moot, just get some good ram and run it sync you cheap-skates

 

[Wingznut]

You'd never catch me arguing with Thugsrook or Mikki about overclocking and benchmarking.

 

[sman789]

did sum comparing and now i have a higher OC i'm happy with
3.0 with 5:4
CPU: 7297 Mem: 9030 HDD: 1608
3.36 with 3:2
CPU: 8284 Mem: 9539 HDD: 1604

mas y mas maybe

 

[Shimmishim]

Originally posted by: Wingznut
You'd never catch me arguing with Thugsrook or Mikki about overclocking and benchmarking.

yeah... but just don't let thugs go off on his... 'i spent less money than you on ram and get better sandra scores' than you spiel...

whenever he talks now.. i hear "blah blah blah blah blah blah"... LOVE YOU THUGS!!!

 

[THUGSROOK]

sorry shim ~ but Technonut is the new king of cheap ram overclocking :beer:

so youll have to listen to his spiel insted

 

[stevejst]

sorry shim ~ but Technonut is the new king of cheap ram overclocking

Well, according to your numbers, you are the king of cheap ram overclocking. I don't think anybody else can claim to have over 10,000 on PCMark2002 with ram on 382 MHz. But with two-three beers everything is possible.

 

[THUGSROOK]

go back into your programing hole and play with your c++ since you know nothing else

 

[stevejst]

go back into your programing hole and play with your c++ since you know nothing else

That would be still more than you do. Short of benchmark fantasizing.

 

[Tom]

Thugsrook-
you have a great system there, but what does it have to do with the subject of this thread ?

The issue is how significant using a memory ratio that reduces memory speed vs cpu speed is to system performance. Comparing one system to another system isn't the best way to find out, there are too many variables. The way to find out is to change memory ratios on the same system.(like sman789 did a few posts back. and Zroc and Prometheus have posted about elsewhere)

In order to get a meaningful comparison of how using a ratio affects your system you would need to be running at 200fsb or above, otherwise the ratios in question aren't available.

 

[stevejst]

It is not hard to understand why some people like to post unbuffered Sandra scores. It is not only that is a hype of the moment but since when you post Sandra picture with the outcome of the bench, it is very easy to fake it. You can test it with any options in module enabled or disabled and claim it is anything you want.
But if you post buffered score (and note that PC in real application will be using buffering, perhaps not perfect but nevertheless) it is a little harder to fool with a picture because there is no way you can get it better than it is just by fooling with options.

 

[THUGSROOK]

Originally posted by: Dead Parrot Sketch
Thugsrook-
In order to get a meaningful comparison of how using a ratio affects your system you would need to be running at 200fsb or above, otherwise the ratios in question aren't available.

ummm the ratios are all available (4:5, 5:4 and 3:2) weather i run 191 or 200fsb (yes i can run 200fsb)

 

[chsh1ca]

I notice you still refuse to run identical tests to thugsrook for comparison purposes Steve. I'll take that to mean you know you will get lower scores and thus be proven wrong, therefore I think this discussion basically takes care of itself...

A mod should really delete the pointless inflammatory posts in this thread...

 

[Tom]

My mistake, then, have you tried or do you have any interest in finding out what kind of penalty your system gets from going to a 5:4 ratio ? Or speculating about it ?

 

[SinfulWeeper]

Well now, I finally read this thread to call BS all due to the title... but decided to read through the whole two pages anyway.
I was going to recommend running more benchmarks, but you already objected to that... .

1. Sandra is sh!t for a program. It only exists so I can get more money by giving high numbers that have no real world time and place. Other than for those poor clueless peeps who see's it in Computer Shopper or PC Mag. Gotta love the program for that one and only strength.

2. I was going to recommend you bring your CPU down to stock speed and run all the benchies that will bring you closer to real life using the 1:1, 5:4, 3:2 dividers. Your Sync vs. Async can and would be determined there. But obviously you feel Sandra has the final word

.

3. I thought I might actually get the chance to be proven wrong. Unfortunately because of laziness on your side, I still can not .

4. Your attitude in this thread... You Bastage, you had my wake up the sleeping beauty before her regular wake-up time :|. Oh well. Listening to your cr@p was well worth it, being she knows a little bit of computers... I'll have to let her laugh by reading this thread also ... BTW: Nice thread

5. Can you please put in more childish posts? I had to write this post in a childish manner because it seems all you can understand.

Have at 'em steve

 

[stevejst]

I notice you still refuse to run identical tests to thugsrook for comparison purposes Steve.

?
Are you talking about his "unbuffered" scores? Uh, sure I would get lower numbers than the ones he posted.
But it is irrelevant for the point of this thread, isn't it?
Here, once again, I said synchronicity is worth for anywhere between 2 and 6 MHz on FSB, depending on the test used.
That is the claim, I see nothing that would be a sort of argument au contraire in his posts except a pissing/insult contest.

It is also amazing that a number of would be "experts" are doing the same, i.e. participating in pissing/insult contest while at the same time claiming reason, knowledge, and civility they don't have apparently. BTW what makes you think I'd be jumping on the tasks you would demand? If that entertains you, keep on posting in the thread for what is worth. I'll be reading if I want to.

@Sinfulweeper, I am sorry but I have no interest in reading or debating (?) what you have to say.

 

[Tom]

Originally posted by: chsh1ca
I notice you still refuse to run identical tests to thugsrook for comparison purposes Steve. I'll take that to mean you know you will get lower scores and thus be proven wrong, therefore I think this discussion basically takes care of itself..

How would comparing 2 systems with different cpus, different memory, different motherboards, different memory settings, one with GAT one without, different vid cards, etc., prove anything about the relative speed of synchronous vs asynchronous ?

 

[stevejst]

Exactly. My CPU runs stable 3.36 GHz and nothing more than that. That is where I cannot use GAT. So yes, my system is weaker then his, and yes that has very little to do with synchronicity. Even when I run stable say 3.3 GHz, at 275 FSB, my memory is only at 220x2, with 2-3-3-7, which is not that big of a leap from what he has. If I want to run 280 FSB which is the max stable I can run it then I have to relax timings further, to 2-4-4-7.
So the comparison makes very little sense.

 

[SinfulWeeper]

Syncronuous 1:1 using 210/210
Sandra: 5052/5091

Asyncronuous 5:4 using 236/189
Sandra: 5089/5063

You are still failing to realize how much async is hurting you.

In your 210/210... your getting the same results even though you are adding 312MHz.
You think your slightly higher scores have any improvement? They do not at all... Your talking a mere .007% increase on the first set of numbers. Not to mention a decrease in the second one.
You want to know how much 312MHz is a improvement. When comparing my PII333MHz vs. my Celeron 300MHz(a) @ 495MHz, the Celeron runs loops around that PII in everything. Given you are not doubling your speed as in that case. But you should see at least a 2-5% improvement if sync is not important.

Do us all a favor. You are obviously a rich person being some college teacher or something like that. Go out, buy some PC4200 memory, run synchronously and put you money where your mouth is. If you can not afford it for whatever reason that fails to come to my mind (other than lying about your employment), do us all a favor and switch to AMD since it would become apparant that's that's all you can afford.

Out of all my experiences with P4's, they are memory bandwidth hungry. Starve them of that bandwidth and you need a much higher clock for the same performance.
But if you can not afford good memory (you should be easily able to though), go ahead and run async. You need those very high clock cycles to compete with a P4 in sync.

 

[stevejst]

Thanks for the advice.

 

[Slappy00]

This is one of the more entertaining threads I've read here for a long time....

Sinful has a point: Alot of the anomosity here is that some ppl like myself paid good money (woe to all the Mushkin ppl out there who eat ramen for a month after buying some RAM ) and would like our purchase(s) to have some shred of validity. Now I for one dont really care just because I can run Sync up to 533DDR and hellified async, so either way I win, but alot of others what some crushing scores for their purchase.



As an aside: whats the longest user posted thread you guys have seen here? (not to be off-topic or anything, so dont answer if you dont care)

 

[WobbleWobble]

Originally posted by: WarCon
Just curious about modern C++, are the SSE (2) instructions separate or does the compiler just implement them?

By default, they're off (at least with the compilers I've used). In the compiler's options, you can enable optimization for things like K7, P3, P4, MMX, SSE and that stuff.

I've never really used GCC much, but I'm sure some extra flags would do it

 

[stevejst]

Yes, you can use flags to include the use of certian subsets of instructions though frankly I never used it myself. But you better know what you need else you'll get compatibility problems.

 

[oldfart]

Ah....one of my favorite subjects. I was away this week and dont have the time to read through all of this, but a few basic points about 1:1 vs a ratio:

1) Dont limit your CPU speed just to run 1:1 ratio. CPU MHz will gain you much more performance than mem speed will. CPU speed 1st, mem speed 2nd.

2) Dont waste your time with synthetic mem benches like SiSoft, Aida, etc. They dont translate into real world performance. This includes unbuffered tests. I've seen plenty of 20 - 30% gain SiSoft benches that show less than 1% gain or even a loss in real world performance.

3) A lot depends on the memory you use. The so called "PC3700 - 4000" ram being sold runs with very relaxed timings. A setup running 250 FSB 1:1 DDR 500 3-4-4-8 is no faster (very likely slower) than a system running 250 FSB 5:4 DDR 400 2-2-2-6. High DDR speed with relaxed timings is ~= to lower DDR speed with tight timings. If you buy the relaxed timing stuff, you cant use tight timings even if you run a ratio. You are stuck.

4) That tRas to 11 thing does not apply to Intel setups.

 

[Howard]

Originally posted by: DAPUNISHER
Ad Hominen attacks won't win you any credibility in these forums troll Now, you've danced enough, eat the scraps we thrown you fool!

Calm down. I hope you don't derive pleasure from this.