Upgrading for M2TW

[orangat]

Here's my specs:
AMD 64 3000+ Venice @2.7 GHz
XFX 6600 GT (PCI-E)
DFI Motherboard
1 GB (2x512) Patriot XLBK Memory
Zalman 7700 CU CPU HS/F (installed with AS5)
550W Antec Truepower

Your weak link is definitely the video card.
Upgrading memory and the cpu can wait till the game arrives.

I see you have overclocked your cpu. You should get some ram that can overclock as much as possible too, as it increases memory bandwidth. And I'm pretty sure memory bandwidth is the real limit of how many men you can have in a battlefield without the pc choking.
.......

What makes you say the main memory bandwidth is the limiting factor?

[CBR]

What makes you say the main memory bandwidth is the limiting factor?

Well that goes back to the early days of RTW where several of us with 2400+ CPU's had horrible lag in battles but other people with a newer generation of AMD cpu's of just 2500+ had no problems at all. Other people over at TWC did tests with high end CPU's that showed only a minimal increase of max amount of men before it turned into a slide show too.

Earlier this year I was reading up on hardware as I was planning on getting a new pc and also decided to do some overclocking with those wonderful Opterons. That was when I first noticed the huge increase in memory bandwidth from using dual channel ram(something my old pc didnt have) and ofc the increased bandwidth from overclocking ram too.

Now im no tech geek but I thought that if super CPU/GPU didnt do much, then what other type of hardware limit would there be? A high end CPU doesnt mean much for memory bandwidth so maybe thats why they didnt see any real difference.

My new PC has overclocked ram too and I can run at 12k men without any noticeble lag, which appeared to be higher than what people just using a good cpu and standard ram could do. Of course not very scientific, as I dont have precise tests done on other systems but it seemed to work the way I thought. But I could be wrong of course.


CBR

[Boohugh]

There will almost certainly be newer graphics cards around when M2:TW arrives meaning you'll be able to get the current ones for less, or just splash out on a new one, so waiting is definitely the best course of action imo .

[fatsheep]

First off, thanks for all the suggestions guys!

Second off, I do intend to wait. Once they put a playable demo out to test my system on I can be more sure what I need to upgrade. Also, it would be silly to upgrade now since prices will have gone down quite a bit by september.

Oh btw, about the memory. I have it on a divider so the memory is actually running at only a tad below 240 mhz (200 mhz is default for PC3200 RAM) while the CPU is going at 290 mhz HTT. I also have the memory on somewhat loose timings (3-3-3-8) so I would like to get some memory that will either allow a 1:1 divider and tighten up timings (although I doubt I'll be able to do both, I'd much rather have the 1:1 divider then the slightly tighter timings).

Again thanks for the advice.

[CBR]

Well 1:1 divider is pretty tough at nearly 290 Mhz but this test did have a couple that could do it: http://www.anandtech.com/memory/show...px?i=2676&p=16

But your current 1 GB is very good so its not that important to upgrade, at least its not your first priority if money is the issue.


CBR

[fatsheep]

Ouch, those RAM setups are out of my price range (<$200). I guess I'll just hope a high OCer falls down below $200 by november.

[orangat]

First off, thanks for all the suggestions guys!

Second off, I do intend to wait. Once they put a playable demo out to test my system on I can be more sure what I need to upgrade. Also, it would be silly to upgrade now since prices will have gone down quite a bit by september.

Oh btw, about the memory. I have it on a divider so the memory is actually running at only a tad below 240 mhz (200 mhz is default for PC3200 RAM) while the CPU is going at 290 mhz HTT. I also have the memory on somewhat loose timings (3-3-3-8) so I would like to get some memory that will either allow a 1:1 divider and tighten up timings (although I doubt I'll be able to do both, I'd much rather have the 1:1 divider then the slightly tighter timings).

Again thanks for the advice.

Aggressive ram overlocking does relatively little for A64 systems. Spending to replace your current memory modules is pointless.

Memory dividers do diddly squat for performance. So don't worry about getting it to 1:1.
The reason is obvious to anyone who has a clue about the Athlon64 HT model.

[CBR]

I assume you are talking about the Barton cores which started at 2500+. The big improvement from 2400-2500+ is probably mostly due to the increase of the fsb from 133->166 and bigger l2 cache _not_ due to faster memory speeds. The older pre-Athlonxp behaved in exactly the same way also.

The A64 is not memory bandwidth starved and buying faster memory over the standard is pc3200 is generally pointless. Simply compare 754 vs 939 benches for proof.

With A64 systems, faster memory modules will give big huge improvements in synthetic benchmarks like Sisoft but do ziltch in real world benchmarks.

Oh yeah the Barton core and it was 2600+ actually. What I couldnt understand back then and now was the apparent huge difference in framerate between our older 2400+ and their newer 2600+ cores. Whereas we had horrible input lag, that made MP games at 8k men near impossible to play, they reported no problems at all.

But nonetheless the tests I have done on my pc shows an increase in fps when overclocking ram. Overall my overclocking means I can play with about 2-2.5k more men before the fps is low enough to make it unplayable. Around 30-40% of the boost comes from just the ram.

Now whether its because of memory bandwidth or the lower latency, or whatever I can test in various benchmarks, I dont know, but what I do know is that there is a clear advantage from overclocking ram in a game like RTW. If you never go anyway near that max number of men (before the game turns into a slideshow) then a couple of fps extra wont do much, but that depends on what unit size you play on of course. You are more likely to run into that barrier when playing with huge unit setting.


CBR

[orangat]

Oh yeah the Barton core and it was 2600+ actually. What I couldnt understand back then and now was the apparent huge difference in framerate between our older 2400+ and their newer 2600+ cores. Whereas we had horrible input lag, that made MP games at 8k men near impossible to play, they reported no problems at all.

But nonetheless the tests I have done on my pc shows an increase in fps when overclocking ram. Overall my overclocking means I can play with about 2-2.5k more men before the fps is low enough to make it unplayable. Around 30-40% of the boost comes from just the ram.
.......

Can you link TWC thread on the tests?
What is your specs and what did you do to overclock? ram speed/divider etc.

[orangat]

Well that goes back to the early days of RTW where several of us with 2400+ CPU's had horrible lag in battles but other people with a newer generation of AMD cpu's of just 2500+ had no problems at all. Other people over at TWC did tests with high end CPU's that showed only a minimal increase of max amount of men before it turned into a slide show too.

Earlier this year I was reading up on hardware as I was planning on getting a new pc and also decided to do some overclocking with those wonderful Opterons. That was when I first noticed the huge increase in memory bandwidth from using dual channel ram(something my old pc didnt have) and ofc the increased bandwidth from overclocking ram too.
........

I assume you are talking about the Barton cores which started at 2500+. The big improvement from 2400-2500+ is probably mostly due to the increase of the fsb from 133->166 and bigger l2 cache _not_ due to faster memory speeds. The older pre-Athlonxp behaved in exactly the same way also.

The A64 is not memory bandwidth starved and buying faster memory over the standard is pc3200 is generally pointless. Simply compare 754 vs 939 benches for proof.

With A64 systems, faster memory modules will give big huge improvements in synthetic benchmarks like Sisoft but do ziltch in real world benchmarks.

http://www.anandtech.com/memory/showdoc.aspx?i=2676&p=5
The Ballistix only gives ~3% improvement going from 200->300(pc4800) which is pathetic. Other modules in the roundup have bigger 'improvements' but that is due to their terrible pc3200 performance (since fast memory modules are bad when configured to run at pc3200).