Official S754 vs. S939 A64 Thread

**LilGator** · 08-09-2004, 10:36 AM

Since we have so many decent threads in here, that have been completely overrun with this topic, I think it needs it's own thread.

To get started, we need some volunteers that have similar systems to compare

What we need:

A) S754 CH 1MB with a 9800 Pro or XT
A) S939 NC 512K with a 9800 Pro or XT

B) S754 CH 1MB with a X800 Pro or XT
B) S939 NC 512K with a X800 Pro or XT

C) S754 CH 1MB with a 6800 GT or Ultra
C) S939 NC 512K with a 6800 GT or Ultra

Benchmarks to be run:

- 3DMark2001 game tests (give compare link)
- Super_Pi 1M
- Pi_Fast
- 3DM2003 game tests and CPU test
- Aquamark game tests and CPU test

Please post some other CPU related benches / gaming benches to stress cache and memory bandwidth (dual-channel).

So we don't have to depend on the multies of the chips that will be compared, let's run each bench at these speeds:

8x200 - 1.60GHz
8x240 (166 Mem Divider) - 1.92GHz
8x267 (150 Mem Divider) - 2.14GHz
8x300 (133 Mem Divider) - 2.40GHz

And finally 8xMMM - (MMM will be the slowest maximum memory speed between the two systems being compared. If one system's memory can do 240MHz, and the other 260MHz, a test should be run on both setups at 8x240 (no memory divider).)

Memory timings will be kept the same. If the memory being compared is different, try and use an equivalent timing. IE: BH-5 that does 2-2-2 vs. CH-5 that does 2-3-2. Both setups should run 2-3-2. If we are comparing EB vs. BH-5, run the EB at 2.5-2-2, and the BH-5 at 2-2-2 to try and minimize any differences there.

WindowsXP - DX9 - Latest official drivers (CAT 4.7 for ATi) (61.77 for nVidia). No OS/Driver tweaks (detail settings all on high)

Video cards will be run at equal stock clocks. 9800Pro vs. 9800XT, run at 412/365 is fine if both cards are capable, 380/340 if not. Comparing an x800Pro vs. an XT shouldn't be done (unless the Pro is modded to 16-pipes, in which case run both cards at stock Pro clocks). Same goes for 6800 GT/Ultra, run both at stock GT clocks.

I'll try and compile the results after benchmarks have been posted and put them in this thread with individual tests differences and an overall % of which of the two chips were faster.

Let's also try and keep the S754 vs. S939 debate in here, so we don't keep running other threads off-topic

(I know, i'm guilty of this)

**chunkylover77** · 08-09-2004, 12:29 PM

I have a 3400with a 6800 ultra. I'm willing to bench for the cause.

**JNav89GT** · 08-09-2004, 04:38 PM

I have 3800+@2940mhz and 6800gt
unfortuneately on abit I am limited to run ram at 333 and less than 245htt.

Also have FX on Asus A8V but crappy 980opro

**LilGator** · 08-09-2004, 04:45 PM

Well, we almost have a match then. Can you both run 8x200, with stock GT clocks ? Should work out fine. All the 4 speeds i have listed up there use a divider to keep the memory at 200MHz so that will be fine. The last one is just to see what impact raising the memory has.

Can you both set up a fresh XP install with DX9/61.77 Det's and run those benches ? What memory do each of you have, and what stock timings can we use for DDR400 ?

BTW, i didn't mention this, but lets use a default LDT of 2X to eliminate any differences there...

**HKPolice** · 08-10-2004, 01:34 AM

No need to compare, clawhammer wins:

http://hardocp.com/article.html?art=NjQ3LDY=

Even in Doom3, 3400+ edges out a 3500+

http://anandtech.com/cpuchipsets/sho...spx?i=2149&p=7

/thread

**iboomalot** · 08-10-2004, 05:05 AM

not always a winner

http://www.anandtech.com/cpuchipsets...px?i=2065&p=15

http://www.pcper.com/article.php?aid=46

3600 is a 3500+ was a typo

http://hardocp.com/article.html?art=NjQ3LDU=

I voted both are equal-- because on any given day or depending on mobo,fsb,ram,video card, timings,clocks,etc etc can tip the scales one over the other.

**JNav89GT** · 08-10-2004, 06:53 AM

Originally posted by LilGator
Well, we almost have a match then. Can you both run 8x200, with stock GT clocks ? Should work out fine. All the 4 speeds i have listed up there use a divider to keep the memory at 200MHz so that will be fine. The last one is just to see what impact raising the memory has.

Can you both set up a fresh XP install with DX9/61.77 Det's and run those benches ? What memory do each of you have, and what stock timings can we use for DDR400 ?

BTW, i didn't mention this, but lets use a default LDT of 2X to eliminate any differences there...

I can setup fresh os in about a week or two. Right now too busy to reinstall. It's a fairly fresh os though. I have dx9c on there. Why don't we run something like 10x200 and 4x ldt(or I'll go 3x if you have nf150). My ram is cas2 2:2:5 in that system now, but I can swap in some other ram if you need me too, or backoff on the timings. I think it will have to be on honor system that no OS/LOD/vid card tweaks will be used. I don't tweak os for gaming/benching. Maybe turn off cheese graphics, that's all.

I can tell you one thing. I setup a 3200+ newcastle system for a friend(he couldn't find a 1mb model :O) and it's definately slower in 3d. I'm at 2.5ghz in testing on that system at 228fsb and 1:1 ram with some mushkin CH-5 at 2.5:3:3:8. Think it barely makes 12k in 2k3 w/GT oc'd to 400/1100 or so. I think it's like 22,5k 2001marks. Not sure how some people score so high with GT's in 2k1 b/c even in my 3800+ at 2940 I can't clear 25k whereas my 9800XT would do 25-26k fairly easy. Maybe I have something setup wrong :O

**LilGator** · 08-10-2004, 10:17 AM

OK, should be fine. chunkylover77, can you set your system up the same way ?

Since you both can use 10x multi's, go for it. Do 200MHz 2-2-2-5 with the memory. Try a 200x10 run, 220x10, and 240x10.

Since JNav has a GT, run both cards at stock GT clocks and use the same driver. Then bench away and post results

**chunkylover77** · 08-10-2004, 10:38 AM

I have to throw a fresh os on it anyway. I can do it this weekend.

**~~StormPC~~** · 08-10-2004, 10:41 AM

Originally posted by iboomalot
not always a winner

Look at the title of the first bench barchart. The whole point of the bench was to prove the 3500+ is faster than the 3400+. When you hand select the tests you can make any CPU come out on top. When the title of the barchart is "Percentage Improvement of 3500+ Over 3400+" WTF do you think the bars are going to say? They'd be idiots to include the benches that have the 3500+ losing! Geeeeeeezzzzzzzzzzz!

Of course you voted they are equal. Should that be a shocking revelation? I don't think you surprised anyone with your vote. Can you guess what I voted???

**chunkylover77** · 08-10-2004, 11:20 AM

Actually guys I have the ZNF3-150 board so will that make a difference in the benches? I dont have a 250 board.

**LilGator** · 08-10-2004, 12:32 PM

It shouldn't, no. Just both of you run 3X LDT. (If you can run 3X at 240HTT, if not, both run 2X LDT at 240HTT)

Does anyone have any benchmark recommendations besides the ones listed ? I'm trying to stick with the tried and true stuff, not WebMark2004

**iboomalot** · 08-10-2004, 01:15 PM

Gator I personally think the benches you picked out are great.

pure processor tests, couple GPU/CPU mixed tests and one pure GPU test. Very nice picks.

Storm your piss poor attitude and a few others are costing AMD sales and loosing people to INTEL, Congrates good job. Don't believe me see for yourself in this thread.

http://www.xtremesystems.org/forums/...threadid=40360

poor kid is scared to even think about buying a 3500+ due to you and handfull of others say how crappy a 3500+ is.

**pkrew** · 08-10-2004, 01:31 PM

That's streching things just a little don't you think?

**iboomalot** · 08-10-2004, 01:38 PM

I didn't ask for Storm's smart ass comments and due to his comments and attitude on two processors that are very close and I can't think why someone would say that a 939 doesn't scale well is absurd and incorrect but of course thats Storm's posistion and its plain to see. He is part of a problem not a solution.

God forbid if in tests the 3400+ should loose by a single point, he might kill himself

wonder what his opinion will be when all the 1mb are sold and the only thing avil are NC cores?? or when the CH are discontinued and all the fast processors are 939s.

difference between myself and Storm I can prove they are very close clock for clock. He just can't accept the fact the 939s NC are nice and fast even without the extra cache.

**~~StormPC~~** · 08-10-2004, 01:55 PM

Do you always get what you ask for? Me neither.

So you can prove the 3500+ is faster than a 3400+? I'm waiting. This should be good!

That must be why over 60% of the people poled believe ClawHammer is faster and less than 15% believe in the Easter Bunny as you do.

I guess you are just smarter than us, eh?

The smart thing for AMD to do would have been to sell the ClawHammers for more and the Newcastle S939s for less. That way they would sell way more of the new stuff and they could completely stop production on the old stuff. Instead they desided to gouge us on the S939 hoping we'd be too stupid to realize the ClawHammers are by far more powerful and a better value. But then, all AMD CPUs are a better value than Intel's chips AND AMD is not notorious for making great marketing decisions, just for making good CPUs for less (most of the time)!

**LilGator** · 08-10-2004, 01:57 PM

The poll seems to think otherwise

60% +

I actually voted that they are about the same...but as far as preference, I'd take the Clawhammer anyday.

I left the poll kind of vague, which is faster, overall i'd say they are about the same. But in the type of things I do, gaming etc...the ClawHammer is faster. In Sysmark, or WebMark or whatever, maybe they are the same, that all comes as part of being faster or not.

But hey Storm, the way it is now, us smart people get the faster chips cheaper

And as far as marketing goes, AMD is lightyears ahead of iNTEL at this moment

**~~StormPC~~** · 08-10-2004, 02:17 PM

Make no mistake, the only reason AMD is doing better is the superiority of their products. Their marketing is still questionable at best.

It's encouraging to see so many people who actually understand the truth (60%+) here at Xtremesystems.org. It shows me that my decision to spend time here is a good one, despite the existance of a few "not so with it" types who refuse to acknowledge reality.

Well done everybody!

**iboomalot** · 08-10-2004, 02:52 PM

**~~StormPC~~** · 08-10-2004, 04:22 PM

If your point was that reviews are worthless or you have way too much free time on your hands, you made your point.

Sorry buddy, but the members of XS (most anyway from the look of the pole results) know better than to trust reviews, especially from the three sites you pulled bargraphs from. I've never seen anyone work so hard to prove a lie.

**iboomalot** · 08-10-2004, 04:25 PM

Storm you are far from unbias and Iam 100% sure they have more fair and accurate testing proceedures than you do.

as far as free time it only took me 10 mins to get those links. Plenty of FACTS and PROOF to show my standings on this.

Good day. oh and nice sig LOL

**trans am** · 08-10-2004, 04:38 PM

I would be willing to do the test. I fit the criteria for the 3200 claw testbed.

**~~StormPC~~** · 08-10-2004, 04:41 PM

Originally posted by iboomalot
Storm you are far from unbias and Iam 100% sure they have more fair and accurate testing proceedures than you do.

as far as free time it only took me 10 mins to get those links. Plenty of FACTS and PROOF to show my standings on this.

Good day. oh and nice sig LOL

We know your motivation for trying to prove that the Newcastles are as fast as ClawHammers. You have one. I on the other hand have many computers. What is the motivation for my "bias"?

Sorry but you're grasping at straws. Newcastles are weak compared to ClawHammers and everyone who has tried both already knows this, hence the results. I'm finished trying to make you see reason. To do so further would be like having a stink fight with a pig. You just can't win.

**pik-ard v1.1** · 08-10-2004, 05:08 PM

i voted them equal, because yes, im sure anyone here can tell the difference(without seeing the numbers) of 300-700 points difference in 3dmark2k1...

lets just assume stormPC didnt just discount three reviews with one sentence belittling iboomalot for finding said reviews...

ok, we all know on s754 a ch will beat a nc at equal speeds, ram timings, etc.

we all know DC is good for ram, and more cache is good for cpu's. s754 supplies one, s939 supplies the other (non aFX, for the sake of this discussion). now, yes, in benchmarks, there will be a difference, be it s754 > s939 (as stormPC claims), or s754 < s939 (as the reviews claim), the benchmarks will give one higher numbers then the other.

now, yes, for some people, what the benchmark says is the ultimate dictator for what is bought. for people like oppainter, macci, pkrew(maybe??), and other people who bench, to bench.

but, the majority of us, buy pc parts based on how well it will play games, price/preformance ratio, etc, with benchmarks as a secondary reason, or a tiebreaker for what to buy.

what i personally hope i can find the money to do, is buy a somewhat cheaper s939 cpu, and a motherboard that will last, just to get a new compy up and running, then later on, upgrade to one of the higher speed s939 proccys (which arent comming out for s754... upgrade path = kaput)

stormPC, im sure its wonderful to be able to use whatever system you want, but not everyone can go from a 3400+ to a 3700+ to a FX-53 in a matter of weeks/month or two. we have to get one thing, and get it to last. that may be what your forgetting. i want to be able to build a pc with a solid upgrade path, so i dont have to shell out tons of money when i want to get something faster... which granted, the a64 4000+ wont be cheap, but atleast i wont have to spend the money on the motherboard as well, which probably saves me about $100 at the least. (thats alot to some of us)

**ocmyface** · 08-10-2004, 06:21 PM

how long do yall think it will be when the clawhammers sell out? how long is production planned to continue if it even is?
also is C0 the best revision for someone with 2x512 pc3700EB?

thanks and sorry for the quick thread highjack, maybe yall can think of something to argue about thats related to my post

View Poll Results: Which is faster ?

Thread: Official S754 vs. S939 A64 Thread

Thread Tools

Search Thread

Rate This Thread

Display

Official S754 vs. S939 A64 Thread

Bookmarks

Bookmarks

Posting Permissions