4Ghz Core 2 Duo vs M.A.M.E. 0.120 (benchmark results)

[Tiger-Heli]

DiveMaster - Thanks!

Thanks TheManuel!

DDR2-800 is really cheap right now - I can understand you not upgrading, but it wouldn't make sense for me to buy less than that.

Before we start, I have some dumb newb questions coming from the AMD world and being somewhat out of touch:

Can you change both the multiplier and FSB on the C2D's or just the FSB or both but only up to the max.  On AMD, usually you can only mess with the FSB b/c they are multiplier locked.  For example, the E4500 is stock at 11x200 FSB, but Taz mentioned a 9x multiplier.  Your E4300 is stock at 9x200FSB, but you said you had it at 3 Ghz, was that 10x300, or 9x333FSB?  (Errmm - you said you're at 300FSB at 2.7 so that tells me you are running at 9x, but Taz said the E4500 was picked for the 9x multiplier, so I'm guessing maybe you can reduce the multiplier but not increase it?)

Also, can you run the memory bus independently of the CPU FSB or not (or does that vary from mobo to mobo)?

The games I would mainly be interested in would be
SFrush
Cruisin'USA
SWTrilogy (not much chance)
RidgeRacer

and the like.

If you don't mind, I would like to see framerates at Stock clock (1.8 Mhz), 2.4 Ghz, 2.7 Ghz, and 3.0 Ghz, for the above games.

(I thought I would save you some work and just look at the MAME benchmarks site, but CrusnUSA is running at 69% with a 2133 Mhz Core2Solo/Duo and 101% with a 2033 Mhz Core2Solo/Duo both with 0.118, so ....)

I'm just kicking around the upgrade idea now, so no rush on the results, but I'm curious to see what you come up with!!!

Thanks again! 

[TheManuel]

DDR2-800 is really cheap right now - I can understand you not upgrading, but it wouldn't make sense for me to buy less than that.

You are absolutely right on this.  I built my PC in April right around the middle of the price drop curve and they are dirt cheap right now so just go for the fastest you can afford although I would not pay too much extra for boutique memory with low timings.

Can you change both the multiplier and FSB on the C2D's or just the FSB or both but only up to the max.  On AMD, usually you can only mess with the FSB b/c they are multiplier locked.  For example, the E4500 is stock at 11x200 FSB, but Taz mentioned a 9x multiplier.  Your E4300 is stock at 9x200FSB, but you said you had it at 3 Ghz, was that 10x300, or 9x333FSB?  (Errmm - you said you're at 300FSB at 2.7 so that tells me you are running at 9x, but Taz said the E4500 was picked for the 9x multiplier, so I'm guessing maybe you can reduce the multiplier but not increase it?)

Multiplier for Core 2 Duo's is locked if you are increasing so you will be limited by the stock multiplier.  However, you can run a lower multiplier which is what some people do because they want to increase their FSB fo improve communication with the memory (for what it's worth) but they don't want their CPU speed beyond the point where they cannot control stability with their setup.

Also, can you run the memory bus independently of the CPU FSB or not (or does that vary from mobo to mobo)?

In my motherboard, I can run the memory bus faster than the CPU bus but I don't know if I can running it slower.  There is a limited selection of multipliers I can use to set this up in my motherboard.  Maybe someone can pitch in with more info?  I am not much of an overclocker.

The games I would mainly be interested in would be
SFrush
Cruisin'USA
SWTrilogy (not much chance)
RidgeRacer

and the like.

If you don't mind, I would like to see framerates at Stock clock (1.8 Mhz), 2.4 Ghz, 2.7 Ghz, and 3.0 Ghz, for the above games.

(I thought I would save you some work and just look at the MAME benchmarks site, but CrusnUSA is running at 69% with a 2133 Mhz Core2Solo/Duo and 101% with a 2033 Mhz Core2Solo/Duo both with 0.118, so ....)

I will try to run this tonight when I get home or tomorrow if I can't get to it since I will have to find the games first. 
While I'm at it, if you want me to mess around with the memory speed, let me know.  I can run it at 800MHz since this is what you want to buy although I don't expect a difference.

[divemaster127]

You can pick up the adata 2 gig ddr ram 800 on newegg for less than $56.00, its a no name but tons of people on newegg on given 5 stars.  I purchased 12 gig of it & this stuff works great & overclocks very good.  I have a 2 page form that I printed off the internet that is a step by step guide for the gigabyte overclocking of the boards, if you pm me a fax number I can send it over, I cannot remember where I found this guide but its great
dm

[Tiger-Heli]

TheManuel - I'm not too worried about the memory speed - I've never seen it to make much difference.  Thanks again.

DM - I'll remember A-Data.  Thanks!

[taz-nz]

Just to clarify a couple of things (this got way long winded, repeats some of what I've already said but still please read it)

First I got the Core Multiplier on the E4500 wrong it's 11x not 9x my mistake, long day.

The stock Core Multiplier matters as you can't increase this on a Core 2 Duo only decrease it in steps of 1x down to 6x, so the higher the Core Multiplier the higher you can get the CPUs clock speed with and given FSB speed.

So if you want to hit say 3.6ghz and don't want to go any higher than a 400mhz FSB, because your using cheep DDR2-800 and don't want to overclock it , then you need at least a Core Multiplier of 9x to achieve this that CPU speed.

So why would you buy a E4500 with a multiplier of 11x when you could buy a E4300 with a multiplier of 9x would in thoery do the job? Well the answer is that the faster stock clock speed a CPU has the better binning it has, this mean the fastest model or a given series of CPUs will tend to have the highest overclocking potential. Any example of this is the E6750 with a multiplier of 8x @ 500mhz FSB it should be able to hit 4.0ghz but in reality they tend to hit a wall around the 3.8ghz mark with air cooling, as where the E6850 is know to be good up to 4.0-4.2ghz before it hits the wall on air cooling.

So running a E4500 @ 9x 400Mhz FSB is a safer bet to get to 3.6ghz (a 63% overclock) than an E4300 @ 9x 400 FSB to get to 3.6hgz (a 100% overclock).

When deciding on what FSB to run to achieve your final clock speed you need to think about what else your going to overclock then you increase the FSB, the FSB speed effect the speed of your CPU, Memory and Chipset, and these all need to be taken into account. The lowest FSB you can run is the Final CPU speed you want to hit divided by the stock Mutliplier of you CPU, for a E4500 @3.6ghz that is  3600/11 = 327mhz (1307 FSB), now 327mhz * 2.4 memory multiplier gets you DDR2-784 so now your memory is underclocked at there for you have less memory bandwidth, if you increase the memory multiplier to 2.6 you end up with your memory at DDR2-850 now it's overclocked, so the easy thing to do is find the common ground between your memory speed and the final CPU speed you want of 3600mhz with is 400mhz 2x for the memory multiplier to give DDR2-800, and 9x for the core multiplier to give 3.6ghz.

The one other thing you have to watch out for when selecting you FSB is whether your motherboard and chipset can handle the new FSB speed, now the P35 Chipset is good for well over 500Mhz FSB speed, but this is only on the right motherboard with good chipset cooling, some motherboard become unstable when you increase the FSB speed due to their design, others because the North bridge heatsink can not keep the chipset cool at the higher FSB speed. Also the higher you increas the FSB the more likely it is that you will need to overvolt the FSB and Chipset some to make it stable and the higher speed, that extra voltage equal more chipset heat.

Higher FSB speed give the CPU more bandwidth to talk to the memory and the rest of your system with, but there is limit to what is practical, an extreme setup would be run say a CPU at 6x 600mhz FSB to get to 3.6ghz, this would require a very stable motherboard with very good chipset cooling, and a lot of extra voltage to the chipset & FSB to work, it would give you massive FSB bandwidth but would be very hard to get stable.

Memory Speed & bandwidth why should you care?
Memory bandwidth is the speed at which the system can get information from your RAM, basically more is better as long as you CPU can put it to use, little point in having a 800 FSB CPU with dual cannel DDR3-1600 the CPUs FSB bandwidth is just now wide enough to make use of the huge memory bandwidth with would give you.

Memory bandwidth & FSB are important because of the way CPUs work, without getting into a whole heap of tech-talk, The great the Memory & FSB bandwidth a system has the more able the system is to supply the CPU with information to process, that catch is if you don't have enough memory or FSB bandwidth then the system has trouble keeping up with the speed of the CPU (this is made worse if the CPU is noticeably overclocked). This is why CPU have Cache Memory it acts as a buffer between the CPU and the system memory, hiding any gaps in the flow information from the CPU.

Now what happens if the memory bandwidth can't keep up with the CPUs needs, and the memory access is to random for the cache memory to buffer fully, then the CPU becomes starved or information and can't process anything for a split second, now it's worse than that because of the way a CPU works, what might be a one cycle break in the flow of the data can result in a 20-40 cycle break in the processing of data, so the penalty for not having enough memory or FSB bandwidth can be much great than you would think.

A modern Celeron is nothing but a Pentium 4 Core with 1/2-1/4 of the cache memory and a lower FSB, these two things make the same CPU Core perform much slower than it would in a Pentium 4 at the same CPU Clock speed because it spends so much more time wait for data to process.

This where memory & FSB speed & dual channel kick in.

Theoretical DDR2 Memory bandwidth:

DDR2-800 Single channel   ~6.4GB/s   (400mhz)
DDR2-1000 Single channel   ~8.0GB/s   (500mhz)
DDR2-800 Dual channel   ~12.8GB/s (400mhz)
DDR2-1000 Dual channel   ~16.0GB/s (500mhz)

800 FSB         ~6.4GB/s (200mhz)
1600 FSB         ~12.8GB/s (400mhz)
2000 FSB         ~16.0GB/s (500mhz)

As you can see the drop from DDR2-1000 to DDR2-800 is only 23% but the Drop from Dual channel to single channel is 100%, this is why it's important to run dual channel, you loose way to much memory bandwidth if you don't run dual channel.

How does this effect MAME if you drop from dual channel to single channel, well very little at 500mhz (2000 FSB) DDR2-1000 with a E6850 with 4mb of cache like I'm running, about 2-3% across the board in fact as you can see in the benchmarks below.

The catch is that while MAME is basically CPU limited and doesn't appear to require huge memory bandwidth, these result are on my system and thus not the norm, the effect of dropping to only 6.4Gb/s memory bandwidth with a single DDR2-800 Dimm on say a E4*00 series with only 2mb CPU Cache will probably be much more noticeable that the effect of dropping to 8.0GB/s with 4mb of CPU cache on my system.

Benchmark performed same as normal just with my memory in single not dual channel. (still 2gb of RAM in system)

ROM NAME   MAME Ver   dual Channel   Single Channel

- 1942      0.121u1      6572.08%   6555.78%

- a51site4   0.121u1      235.35%      234.89%

- airco22b   0.121u1      143.19%      141.29%

- alpinerd   0.121u1      88.50%      87.20%

- biofreak   0.121u1      145.12%      143.02%

- blitz2k   0.121u1      160.68%      160.12%

- blitz99   0.121u1      153.09%      151.04%

- blitz      0.121u1      167.30%      166.52%

- calspeed   0.121u1      204.23%      203.53%

- carnevil   0.121u1      285.30%      283.93%

- crusnusa   0.121u1      276.53%      275.63%

- crusnwld   0.121u1      249.91%      249.36%

- cybrcomm   0.121u1      135.27%      134.07%

- Daytona   0.121u1      212.79%      209.73%

- gauntdl   0.121u1      134.20%      133.14%

- gauntleg   0.121u1      159.81%      151.36%

- gradius4   0.121u1      96.85%      96.45%

- hyprdriv   0.121u1      175.17%      167.02%

- mace      0.121u1      251.07%      248.84%

- offroadc   0.121u1      438.94%      436.40%

- propcycl   0.121u1      137.27%      135.14%

- radikalb   0.121u1      177.72%      176.58%
   
- raveracw   0.121u1      121.21%      119.77%

- ridgerac   0.121u1      151.72%      148.99%

- scud      0.121u1      28.61%      27.87%

- sfrush   0.121u1      211.70%      199.13%

- sidebs2   0.121u1      215.90%      213.46%

- speedup   0.121u1      179.16%      179.30%

- starblad   0.121u1      145.58%      144.87%

- surfplnt   0.121u1      163.14%      161.74%

- tenthdeg   0.121u1      85.30%      83.64%

- timecris   0.121u1      145.30%      148.09%

- wargods   0.121u1      411.74%      411.54%

- wg3dh      0.121u1      375.54%      374.59%

- xevi3dg   0.121u1      371.37%      372.76%

I am running the Rosewill Z-3 coolers I have 3 of these & they work great, the artic cooler you are looking at is fine.  But with the tags on it, it makes it a pain to install these coolers are not good for being used over & over.  That is why I like the rosewill or similar they use a back plate w/screws & nuts.  I am kicking around dumping my 6750 & going to the 6850 just for the extra performance.  I also prefer Gigabyte motherboards I have 3 of the GA-P35-DS3L, these boards are tough & great overclockers.
dm

I had a look at the cooler your using and while it's the right design type for overclocking it's only a mini tower-cooler, with only a 92mm fan and what looks like three heat pipes, I'd look at tyring something bigger like the Coolermaster Hyper 212 or the thermalright 120 utlra before you change to a E6850, I would be suprised if you couldn't get to at least 3.6-3.8ghz out of your e6750 with a larger heatsink and a bit of tweaking, presuming your ram is not what is holding you back.

 

[TheManuel]

How does this effect MAME if you drop from dual channel to single channel, well very little at 500mhz (2000 FSB) DDR2-1000 with a E6850 with 4mb of cache like I'm running, about 2-3% across the board in fact as you can see in the benchmarks below.

The catch is that while MAME is basically CPU limited and doesn't appear to require huge memory bandwidth, these result are on my system and thus not the norm, the effect of dropping to only 6.4Gb/s memory bandwidth with a single DDR2-800 Dimm on say a E4*00 series with only 2mb CPU Cache will probably be much more noticeable that the effect of dropping to 8.0GB/s with 4mb of CPU cache on my system.

tar-nz:
I think you are well intentioned and you've got the theory down in great shape for the mid-terms :-)
However, I think in practice, it can all be mis-leading.  When I tried dual vs single channel on my setup, just out of curiosity, I did not see any significant difference.  I did not run the benmchmarks in a very disciplined manner but there was nothing that jumped out at me.  I will try to run some tomorrow night (I have to do Tiger-Heli's benchmarks tonight :-) ).  Beside, the theoretical bandwith you quote above is just that, theoretical.  Please run some synthetic benchmarks on your system using Sandra or some other software and report back on your bandwidth.  I'm sure you will notice that it falls way short of theoretical.  In fact, for a more eye-opening experiment, run this bechmark on single channel mode as well and see if dual channel gets anywhere close to twice the bandwidh of dual channel.  Oh, and sysnthetic benchmarks are always the most optimistic test you can run and real world performance is usually lower anyway.
Overclocking webistes tend to get too hung-up on memory bandwith but every time I test different setups I don't see any differences worth writing home about so I'd hate for people planning out their systems to spend too much time and money for extra bandwidth when it does not do much in practice.  Looking at your benchmarks, most games improved by less than 2% (dividing dual channel percentage by single channel percentage), some even reduced performance while a few improved by as much as 6% and this is in the extreme case of single vs dual channel.  So in the end you have to ask yourself if 6% in a few isolated games is worth going after. 

Nevertheless, going dual channel probably makes financial sense since most modern motherboards support it and a pair of matched memory sticks is only a few bucks more than a single stick with the same total memory.  Even if you are not sure it's going to make much difference, you just protect yourself for the unknown for a few bucks more.  However, spending a lot of money on super fast memory doesn't make much sense to me.

I still owe you some benchmarks to put some evidence behind my claims.

[Tiger-Heli]

Taz did post both specs and the info was useful - and he said his system was likely less affected by it than most.

I still want to see TheManuel's benchmarks, but I did get a chance to test three of the game's out here on the work, ermmm, test system and wanted to post them here:

This is an Hewlett Packard E6400 running at stock clock of 2.13Ghz, WinXP, 2G of Ram (dual channel).  I used standard (non multi-core optimised MAME 0.121) with total default settings and didn't run anything specific, just looked at F11 (I also had Access, Outlook, and Seamonkey, and maybe Word running in the background):

Crusnusa ran at pretty near 100%, dropped to 90% at times.
Swtrilgy ran at around 34% from what I could tell
Sfrush ran at about 53-57% but I just got past the startup screens and closed it out (won't really work without a joystick).

Not bad at all and looks like if I get to about 3.0 Ghz, those games are likely playable.

I've about decided how I want to tackle this is a stock E2xxx processor and cooler, either a P35 or 965 mobo (probably Gigabyte or Abit, maybe Asus, Asrock, or BioStar (are these okay?) (also on the 965, I assume I want P965 rather than G or Q since I won't have integrated graphics, but are there other disadvantages to a G965 or Q965?) also I saw FoxConn uses solid capacitors also - any opinions on them, one stick of 1G ram, and a cheap vid card (might need a new power supply also - recommmended brands? - I've used Fortron in the past).

I'll probably see how this does stock, might try to overclock and hit close to 3Ghz.  May add a second stick of 1G ram and a better CPU cooler at that time.

Like TheManuel said, I'm really not too concerned if CrusnUSA runs at 91% instead of a possible 93%, but getting it to 91% from about 39% (Last I remember - XP Barton 2800+) is important.

[TheManuel]

The info was useful indeed.
Taz is the man.  I certainly was not meaning to diminish his comments in any way.
He obviously knows this stuff better than I do.

[taz-nz]

The info was useful indeed.
Taz is the man.  I certainly was not meaning to diminish his comments in any way.
He obviously knows this stuff better than I do.

Ha I get where your coming from, there is way  to much hype around bandwidth and getting that last 100 point in this benchmark or that. I probably shouldn't have posted the Theorical numbers knowing good well it's impossible to reach those figures in the real world, in fact the best Intel memory controllers are only about 50% efficient, the intergated memory controller in the 939 Ahtlon64 on the other hand is 90%+ efficient and get much closer to the theorical number than any other system setup.

Memory and system bandwidth play a greater of lesser roll in preformance depend on your the software your running, But there is point where if you decrease the available bandwidth to the CPU enough no amount of Cache memory will stop a noticably drop in prefromance, now it appears we're lucky with MAME and it's memory bandwidth requirements are fairly low, but a drop in memory bandwidth will still effect it's preformance to a greater or lesser degree depend on your over all system setup.

I lot of my knowledge of effects of bandwidth on system preformance come from my P4 days, when I ran a 2.8ghz P4 northwood (512k cache) @ 3.5ghz 250mhz FSB with DDR-500 , that system was faster than a 3.6ghz P4 Prescott (2mb cache) with DDR-400, it was almost as fast as a 3.8ghz P4 Prescott. The extra cache memory in Prescott Core could not make up for the lack of system & memory bandwidth, the P4 Core simply needed more bandwidth to keep it work hard at high clock speeds.  I also spend a lot time disigning Dual CPU Quad Core Xeon server setup for customers at work, these use a Quad Channel memory setup to supply enough memory bandwidth to keep 8 cores working hard. 

Now not all of this knowledge directly applies to Core 2 Duo, but the basic rules still apply bandwidth matters it just depend of your setup how much. I'm just trying to get people to look at their system setups as a whole rather than just focusing on the CPU clock speed, the best preformance always comes from balanced systems.

I hope you get where I'm trying to come from.

I've about decided how I want to tackle this is a stock E2xxx processor and cooler, either a P35 or 965 mobo (probably Gigabyte or Abit, maybe Asus, Asrock, or BioStar (are these okay?) (also on the 965, I assume I want P965 rather than G or Q since I won't have integrated graphics, but are there other disadvantages to a G965 or Q965?) also I saw FoxConn uses solid capacitors also - any opinions on them, one stick of 1G ram, and a cheap vid card (might need a new power supply also - recommmended brands? - I've used Fortron in the past).

Watch yourself with the 965 chipset boards, there was much larger spread in overclockablity on these boards, the P35 chipset handle high FSBs better and the P35 boards tend to designed to handle those high FSBs better, there are some good 965 overclocking boards out there but you'll probably find you can pick up a P35 chipset board for about the same price. As for brand stick with Gigabyte of ASUS, I was a big fan of Abit up until about 2 years ago, and then they just seamed to loss it, Asrock & Biostar both make cheap rubbish stay away from them. Power supply Cooler master makes some ok 460watt PSU at a good price, but you may not need anything that big depending on what other hardware you plan to run.




 

[TheManuel]

A quick post here because I have to run off to a friend's house but did not want to leave Tiger-Heli hanging.
I only managed to run half of the benchmarks:

      1.8GHz      2.7GHz
sfrush      60.20%      84.33%
crusnusa                   99.21%      150.16%
swtrilgy                   25.64%      40.82%
ridgerac                   44.81%      66.17%

Some games scale linearly, some don't.
I ran taz's command line:
mamep.exe -noautoframeskip -frameskip 0 -seconds_to_run 240 -nothrottle -nosleep -video ddraw -skip_gameinfo -effect none -nowaitvsync -noreadconfig -mt -rompath f:\games\mame\roms sfrush

For sfrush, I press P1 button at the calibration screen to get rid of it in both cases.
I did not run swtrilgy to completion for the 1.8GHz (only 187sec) because I am in hurry so it may be somewhat biased.

Tomorrow I will run the other half of the benchmarks and also check the in-game FPS while actually playing the games with the CPU at 3.0GHz since this is ultimately what matters.

Regards.

Taz:
I got a glimpse of your message but will read it thoroughly tomorrow.

[John IV [MameUI64]]

Also just an FWIW note for those that have had their C2D machines for a while, I just opened my case after a year or so and used compressed air to blow out the dust that had accumulated on the heatsink/fan assemblage and it immediately dropped 10+ degrees under load.  So be sure to keep your equipment clean so it can do its job.

[taz-nz]

Also just an FWIW note for those that have had their C2D machines for a while, I just opened my case after a year or so and used compressed air to blow out the dust that had accumulated on the heatsink/fan assemblage and it immediately dropped 10+ degrees under load.  So be sure to keep your equipment clean so it can do its job.

Yeah, those dust bunnies will get you everytime. 

[TheManuel]

I agree.
My case sucks in dust like you woun't believe.
When re-sat my heat sink a couple of weeks ago I cleaned everything well so I am not sure what made the most difference, the re-seating or the cleaning but I dropped a few degrees at idle. 

Also just an FWIW note for those that have had their C2D machines for a while, I just opened my case after a year or so and used compressed air to blow out the dust that had accumulated on the heatsink/fan assemblage and it immediately dropped 10+ degrees under load.  So be sure to keep your equipment clean so it can do its job.

[TheManuel]

Tiger-Heli:

Here are all the benchmarks the CPU allowed me to do:
      1.8GHz   2.4GHz   2.7GHz
sfrush                      60.2%   75.9%   84.3%
crusnusa                   99.2%   133.6%   150.2%
swtrilgy                   25.6%      40.8%
ridgerac                   44.8%   59.6%   66.2%
All of the above was run on Windows XP with the official MAME binary.  This is the generic one with no optimization.

I could not get to 3.0GHz as much as I tried.  Windows simply would not start.  I don't think I am doing anything different from a few weeks ago when I managed to have it running at 3.0GZ for a whole week until one day it just would not load Windows.  I can't imagine what kind of threshold I might have crossed that now 3.0 is off limits with stock voltage.  I want to run at stock voltage since I want the processor to live long so it does not look like I will be able to get to that benchmark but you can extrapolate from the results above.

All of the notes below are for 2.7GHz:

For sfrush, the results might not be too accurate because, as I mentioned before, I had to hit P1 to get rid of the controller calibration screen or it stays there for the duration of the benchmark.  The problem is that, while at the calibration test screen, MAME runs very fast since there is not 3D modelling going on.  So according to exactly when the game registered the cancellation of the screen, I could have more or less average performance on each benchmark.
During the actual game, the FPS fluctuated between 40% and 80%.  Not very playable.

Crusnusa ran 100%, no problem.

I did not bother to re-run the missing swtrilgy benchmark since it is so woefully slow.  During gameplay, the game ranin the low 30%.

Ridgerac ran in the 30%, very slow.

I suppose you can enable framskip for some of these games and get by.

Regards.

[TheManuel]

I lot of my knowledge of effects of bandwidth on system preformance come from my P4 days, when I ran a 2.8ghz P4 northwood (512k cache) @ 3.5ghz 250mhz FSB with DDR-500 , that system was faster than a 3.6ghz P4 Prescott (2mb cache) with DDR-400, it was almost as fast as a 3.8ghz P4 Prescott. The extra cache memory in Prescott Core could not make up for the lack of system & memory bandwidth, the P4 Core simply needed more bandwidth to keep it work hard at high clock speeds.  I also spend a lot time disigning Dual CPU Quad Core Xeon server setup for customers at work, these use a Quad Channel memory setup to supply enough memory bandwidth to keep 8 cores working hard. 

Now not all of this knowledge directly applies to Core 2 Duo, but the basic rules still apply bandwidth matters it just depend of your setup how much. I'm just trying to get people to look at their system setups as a whole rather than just focusing on the CPU clock speed, the best preformance always comes from balanced systems.

I hope you get where I'm trying to come from.

taz:
I had more time to read you post carefully.
I do understand your perspective on it.  As I was researching my system, I read how bandwidth was a big deal on P4's so much more than with C3D's due to their architecture (I don't know the details, though).  But of course, there should be a minimum you have to supply before you cut into peformance as your benchmarks showed.

[divemaster127]

i got my 6850 yesterday i will drop it in today & overclock to see how it does
dm

[taz-nz]

i got my 6850 yesterday i will drop it in today & overclock to see how it does
dm

Cool, good luck with the overclock.

I'll be interest to see what you get out of it, be nice to have someone else at 4ghz to compare system preformance againest.

The more I read about the E8400-E8500 Core 2 Duo's the more I think I'll be changing CPU's at the end of January, 4.5-5.0ghz sounds good to me  , of cause that last figure will probably required a switch to water cooling.

I'm about to shut down my system, install a spare HDD I have laying around and have a go at instlling 64bit linux & then SDLMAME 64bit, so that should keep me busy for a couple of days getting everything right. I'll post benchmarks when I'm done.

[Tiger-Heli]

Watch yourself with the 965 chipset boards, there was much larger spread in overclockablity on these boards, the P35 chipset handle high FSBs better and the P35 boards tend to designed to handle those high FSBs better, there are some good 965 overclocking boards out there but you'll probably find you can pick up a P35 chipset board for about the same price. As for brand stick with Gigabyte of ASUS, I was a big fan of Abit up until about 2 years ago, and then they just seamed to loss it, Asrock & Biostar both make cheap rubbish stay away from them.

Thanks Taz - I can find TheManuel's Abit 965 board open box for about $45, and other P35's for about $60 while the Gigabyte P35's are about $90 (and the Gigabyte 965's are more than that oddly).  I suppose you get what you pay for to some extent, and I obviously don't replace my systems every six months so nickel-and-dimeing on the mobo doesn't seem like a good idea.

Power supply Cooler master makes some ok 460watt PSU at a good price, but you may not need anything that big depending on what other hardware you plan to run.

I believe the current PS is a Fortron 300W (FSP300-60PN) but most of the reviews said it was actually closer to 350W given the specs on the rail ratings - (Thought it was a 350 until I checked the order history - guess it probably needs to go).

[Tiger-Heli]

Thanks TheManuel -

CruisnUSA was likely a poor choice, at one time it was a real dog in MAME, but I tried with my Athlon XP 2800 (Barton) and got 58-60% in game with probably MAME 0.113 (not sure), almost but not quite playable.  SF Rush ran about 19-21% in game, but it took about 5 minutes to get past the two 15-second countdowns, totally unplayable.  I didn't have the other two games available on that system.

During the actual game, the FPS fluctuated between 40% and 80%.  Not very playable.

But a lot better than I'm getting.

Ridgerac ran in the 30%, very slow.

Hmmmmn, interesting that it's that much slower in-game than the benchmark indicates - not an encouraging sign.

I suppose you can enable framskip for some of these games and get by.

Actually, I don't think you can - I don't remember the actual page, but I think it was on Aaron Giles website when people were complaining about slow speeds in Cruisin.  Aaron said that frameskipping helped in 2D games (like say Tiger-Heli or Striker1945 when the game had to calculate and draw the whole frame.  I crusin, where you are drawing in 3D, Aaron said the game still runs all the calculations and then just doesn't draw them - so you don't get a speed increase and just get choppier video.  (That's probably not 100% technically accurate, but that's the basic concept).

What you can probably do is go into the cheat menu and UNDERclock (I used to know why it was under and not over) the game CPU's and increase performance, but that sometimes messes up the audio.  (Or do like Taz and buy a 5Ghz CPU - just kidding )

Again - I really appreciate you taking the time to run those tests for me!!!!

[Popcorrin]

What you can probably do is go into the cheat menu and UNDERclock (I used to know why it was under and not over) the game CPU's and increase performance, but that sometimes messes up the audio.  (Or do like Taz and buy a 5Ghz CPU - just kidding )

You underclock it so that mame is in a sense emulating a slower cpu, but like you said, if you underclock it too much you get problems with the audio.

[ozzi22]

Is there diffrent results with windows xp 64 and vista 64?  Which is faster?

[divemaster127]

I'm running vista 64, im sure xp 64 is faster but the drivers were never developed much for it.  Also the arcadevga does not support xp 64
dm

[jdjuggler]

Are there any mame benchmarks to suggests that a quad core processor is faster/slow/same as a core 2 duo processor?  And, when I say quad, I don't mean the $500+ processors out there.  I referring to something that could be purchased for not much more than a core 2 duo... in the $200-300 range.
Thanks,
JD

[taz-nz]

Are there any mame benchmarks to suggests that a quad core processor is faster/slow/same as a core 2 duo processor?  And, when I say quad, I don't mean the $500+ processors out there.  I referring to something that could be purchased for not much more than a core 2 duo... in the $200-300 range.
Thanks,
JD

No hs done a full on test of a quad core to date that I know of, I did some testing on a Q6600 at stock speeds, and was less than impressed with the results when compared to a E6850 at stock speeds.  Basically those MAME drivers that are multithread gain hugely switching from single to dual core, but only a little more preformance is gained with the switch to quad core, this small preformance gain from the third and four cores is counter by the dual core higher stock clock speed (2.4ghz for the quad, 3.0 ghz for the dual) add to this the fact the dual core can be overclocked a lot more than the quad core, and the dual core becomes the clear winner currently.

Now this could change with the release next year of the Q9000 series 45nm quad cores and the possibitly on everyday 4ghz+ overclocker quad cores, but in saying that there is the new E8000 series dual core lauching at the same time offer up the possiblity of 4.5-5.0ghz overclocks.

In the end MAME cares more about clock speed than just about anything else and further multithread optimisations are problably going to be slow to come, add to that the fact that only about 2% of all non-MAME software actual make full use of the extra two cores in a quad, I'm sticking with dual core for now and plan to swap my e6850 for a core 2 duo E8500 in the new year.

[joeH]

Will the difference in L2 cache affect performance?  For example, e2180 (1M L2 cache, 10x multiplier) vs. e4500 (2M L2 cache, 11x multiplier)?   Both can easily hit the 3.0ghz range on stock cooling, however, currently there is about a $40 price difference between the two.

[Jdurg]

I'm no expert, but I think that a larger cache means that the CPU has more data waiting for it, and therefore is able to inherently run faster.  (Since it doesn't have to wait for the data to come to it since it's already sitting there in the cache).  Whenever I've bought CPUs for my day-to-day use or for overclocking, if given a choice I've taken the one with the larger cache.

Please feel free to correct me, everybody, if I'm wrong.

[joeH]

Sorry...should have been more specific. I realize the more cache the better. The question was directly towards MAME specifically.   (This would be a machine for a cab...not a day-2-day machine).

[TheManuel]

Take a look at this article at Tom's Hardware addressing this very issues.
The most they managed to get by doubling the cache size was 8% on WinRAR.  Some applications did not gain anything.

I don't know how exactly this would play out in MAME but assume the best you can possibly get is 8% and think whether it is worth the extra $40.  Probably not.  This might get you a few extra frames on those CHD games but nothing worth writing home about.  Then again, you could get nothing for your $40.

Maybe the MAMEWorld forums would be a better place to ask the question but there aren't many people over there willing to help anymore from what I've seen coming back to the hobby.  I think there is too many jaded people there.

[joeH]

Thanks for the post.  I doubt the Mame forums will be much help.  This was just going to be a stop-gap processor until the 45nm cores came out, so the 2180 may be the better buy in the short run.

[Jdurg]

Since filing my taxes, I'm getting a nice present from Uncle Sam.  That means I'll be getting myself a laptop since it's inherently more useful to me than replacing my desktop.  (I can sit in my living room and watch TV while surfing the web with my laptop, and I can also move around the house and do other things while still having my computer on.  Gives me some mobility).

Anyway, with the refund I got i'm looking at getting a new 45nm Core 2 Duo system and just trying to decide if I want the T9500 or the T9300 system.  I'm just not sure if the 100 MHz is worth the extra $225.  From what I've seen in this thread, I wouldn't think so.

So i hope to get this system once my refund arrives and get a chance to see how it holds up.  (It will be a 64-bit system so hopefully the "how-to" for 64-bit compiling will be available at that point).

[headkaze]

Just a heads up Mame Compiler 64 has been released in the Software forum. This should make it easier to compile custom 64bit versions of Mame with optimizations for your CPU. So it might be worthwhile taking a look and doing some more tests for Mame64 if you don't know how to compile 64bt Mame already.

[Jdurg]

Just a heads up Mame Compiler 64 has been released in the Software forum. This should make it easier to compile custom 64bit versions of Mame with optimizations for your CPU. So it might be worthwhile taking a look and doing some more tests for Mame64 if you don't know how to compile 64bt Mame already.

Fantastic!!!  I'm just waiting for UPS to arrive with my laptop and then I'll be looking to compile a 64 bit version of MAME.  This is great news!   

[taz-nz]

Well after way to many months waiting, I scored myself an Core 2 Duo E8500 engineering sample (It helps to know the right people sometimes) So out with the old 65nm E6850 and in with the new 45nm E8500 (3.16ghz 1333fsb stock). Now I've had this CPU a little over 24 hours as I type this and it's been in the my system less than 20, so I still need to tweak a few things, but 4.4ghz (4410mhz) seams to be my limit right now, while it will boot windows and run everyday apps at 4.5ghz, it locks after only a few minutes at 100% load, Air-cooling just isn't enough to keep it cool.  4.4ghz @ 490mhz FSB looks to be stable for the mean time more testing is need, but I may just splash out on a new case and water cooling kit and get it over and done with.

So now I bring you Core 2 Duo E8500 @ 4.4ghz vs MAME 0.123u1.  (benchmark run on MAMEUI 64bit v123.1 in Vista Ulimate 64bit)

ROM NAME   
- 1942      7114.33% (4.5ghz CPU drag race)
- a51site4   225.85%
- airco22b      148.66%
- alpinerd      87.86% (still can't get this sucker to run at 100%)
- biofreak                   154.86%
- blitz2k      172.00%
- blitz99      164.36%
- blitz      176.95%
- calspeed   216.95%
- carnevil   301.32%
- crusnusa   289.38%
- crusnwld   258.61%
- cybrcomm   138.90%
- Daytona   (colours look better, but crashes within seconds unable to benchmark)
- gauntdl   143.39% (was noticably slower this time round, not sure why just yet)
- gauntleg   166.17%
- gradius4   122.94%   (actual used MAMEUI 32bit for this one, still in Vista 64bit)
- hyprdriv   189.75%
- mace      268.49%
- mk4                       186.26% (this is actual too high, but in game is over 100% all the times.)
- offroadc   468.22%
- propcycl   142.66%
- radikalb   178.03%   
- raveracw   118.34%
- ridgerac   159.87%
- scud      30.77% (Ha, back over 30% again, but this still needs alot of work)
- sfrush   219.12%
- sidebs2   178.88%
- speedup   182.32%
- starblad   156.96%
- surfplnt   170.44%
- tenthdeg   97.93% (so close to that magic 100%, but dips to the 70's in some areas game)
- timecris   149.15%
- wargods   406.11%
- wg3dh      387.59%
- xevi3dg   398.29%

Some scores are down on where I would have expected them, but that could just be the Dev's at work or my install of Vista 64 maybe a little broken after the dozen or so lockups testing overclocking settings, I'll probably reinstall and benchmark one of the older builds to confirm which it is.

To anyone looking at building a MAME Monster Machine, I suggest picking up an Core 2 Duo E8400, should be a easy 4ghz overclock with little added voltage needed, just make sure you upgrade the heatsink as they still get hot went you push them hard and the new stock HSF is tiny compared with the older Core 2 Duo's. (I can confirm crazy tempature readies are the order of the day with the E8#00 series CPUs, one core never reads below 44 deg C well the other drops to 29 deg C, over 44 deg C the move together like you would expect.)

Just a heads up Mame Compiler 64 has been released in the Software forum. This should make it easier to compile custom 64bit versions of Mame with optimizations for your CPU. So it might be worthwhile taking a look and doing some more tests for Mame64 if you don't know how to compile 64bt Mame already.

I've download this and I give it a go in the near future, more benchmarks will follow once I get everything tweaked a little better.

If anyone knows of any ROMs or Drivers I should test please let me know, I'm down to only half dozen Roms that are putting up any resistance now.

[releasedtruth]

Very impressize performance Taz, you are really doing us an awesome service here. I've been looking at the Quad core for experimental purposes, see what's possible there, even if it's not in the same league it'd be interesting to see how the architecture works at least.

[headkaze]

Those results are pretty amazing taz! I got myself an E8400 recently but havn't had a chance to overclock it yet.

[divemaster127]

I have a overclocked e6850 & the chd's overall run very good but gauntlet legends still needs a computer with about 15% more legs than mine has...maybe one of the newer cpus would work
dm

[BASS!]

Got my e8400 on the way, I'll be loading gauntlet first. I'll let you know in a week when its all up and running.

[taz-nz]

Very impressize performance Taz, you are really doing us an awesome service here. I've been looking at the Quad core for experimental purposes, see what's possible there, even if it's not in the same league it'd be interesting to see how the architecture works at least.

I had quick go with a stock quad core a few months back, but the result didn't impress me and I haven't had another go since, I may try again with the new Q9000 series once they're readly available, there is a Q9300 engineer sample kicking around work but I've done my impulse buying for the month.
I'd really love to see some MAME benchmark results for Q9000 series quad core at 4ghz (yes people have them running that high).

Those results are pretty amazing taz! I got myself an E8400 recently but havn't had a chance to overclock it yet.

Get to it man, It took me like 10 mins to get 4.0ghz on my E8500, as long as your system is up to it and you've got a good cooler go for, just turn off anything that may get in the way of overclocking, bump the chipset voltages to allow for higher FSB and hit the CPU with 1.4v and crank up the FSB to 445mhz, make any memory setting changes needed to keep your memory speed in check, and you should have yourself a 4ghz Core 2 Duo, that's about all I did. There's a little more to it than that, just ask if you want more details.

I have a overclocked e6850 & the chd's overall run very good but gauntlet legends still needs a computer with about 15% more legs than mine has...maybe one of the newer cpus would work
dm

I'm not going to tell you to dump your E6850 and get a E8400, I think that Z-cooler is still holding your E6850 back some. But the E8400 does look to be an easy 4.0ghz and you can probably keep your current setup including your heatsink, I my new CPU up to 43-46 deg C at 100% load at 4ghz. Every thing after 4.2ghz was hard work, but I think that more down to cooling and luck.

Got my e8400 on the way, I'll be loading gauntlet first. I'll let you know in a week when its all up and running.

Good luck with it, they are a great CPUs, just watch out for those crazy core tempature values, stick to the lastest version of CoreTemp and don't be suprised if one of the Core temps is stuck in the 40s.

[BASS!]

Got my e8400 on the way, I'll be loading gauntlet first. I'll let you know in a week when its all up and running.

Good luck with it, they are a great CPUs, just watch out for those crazy core tempature values, stick to the lastest version of CoreTemp and don't be suprised if one of the Core temps is stuck in the 40s.

Thanks for the awesome thread and all the hard work. I expect to push it to around 3.6 from what I read, and because of the fact I ordered ddr2 800 it may be a bit difficult to overclock to around 4ghz on fsb and voltage/timing ocing alone, I will probably be back for more advice (in another thread) on the ocing of the board and any advice you may have. Thanks again and I'll be lurking this for a while

[taz-nz]

Thanks for the awesome thread and all the hard work. I expect to push it to around 3.6 from what I read, and because of the fact I ordered ddr2 800 it may be a bit difficult to overclock to around 4ghz on fsb and voltage/timing ocing alone, I will probably be back for more advice (in another thread) on the ocing of the board and any advice you may have. Thanks again and I'll be lurking this for a while

Thanks,

The fact that your getting DDR2-800 doesn't mean you can't overclock to 4ghz, It just mean it's more complex and thus more differcult to do. To get a E8400 with it's 9x multiply to 4ghz you need to be running a 445mhz FSB (445*9=4005mhz), since the mininum memory ratio you can run on a P35 chipset board is 1:1 (2.0 in bios) your memory will end up running a DDR2-890 mininum, so you'll need to overclock your ram too, bumping your Memory voltage up a bit say to 2.0v will probably be all you need, but if voltage alone doesn't work for your ram then you can loosen the memory timing some, say it stock 4-4-4-12 then you might try 5-5-5-15 instead, A little bit of testing and tweaking should get you there.

Don't try and overclock everything at once, it's too hard to workout what's causing a problem.

Start with the ram just increase the Memory Multiplier from the stock 2.4 to say 2.66 with everything else stock and test the PC, with the stock 333mhz FSB this will give you a RAM speed of DDR2-886 (333*2.6=886), this is close enough to your required RAM speed of DDR2-890 to test the system for stability with the only the RAM overclocked.

Once you happy you got the Memory timing and voltage right you can move onto the FSB, drop the CPU multiplier to 6x, and the Memory Multiply to 2.0, then pump up the FSB to 445mhz and increase the chipset & FSB voltages as need to stablize the system, The CPU will be underclocked at this point at only 2.67ghz (445*6=2670mhz) and you've already set the memory timings to work at this speed, so your only dealing with the chipset stability.

Once you happy with the system with the FSB and RAM overclocked, increase the CPU Multiplier one step at a time from 6x until you get to 9x, test the system for stability at each Multiplier step, and increase the CPU voltage as need to get the system stable before moving on to the next Mutliplier, Until you get the CPU Multiplier to 9x, you should now be running a 445mhz FSB with the CPU multiplier at 9x and you Memory Multiplier at 2.0, giving you a 4.0ghz CPU with a 1780fsb and DDR2-890 ram.

If done right and with a little luck it will be nice and stable at these settings, now you need to torture test your CPU with something like Orthos, to take it to 100% load and hold it there for hours, I suggest at least 10 hours for you final test with the PC house as it will be in operation. Keep an eye on the CPU tempature and make sure it stays within safe limits anything up to 65 degs C is ok, over that is getting hot, and if it gets to 80 Deg C it will more than likely lockup before it can do any real harm, you really want it in the 20-30s at idle and in the 40s to low 50s at full noise. If it pass that it should be ready to use, it you want to further test it warm up the room tempature to that of a hot summers day and test it again for 24 hours, if it pass that you good to go.

Then all your've got to do is have fun.

Make sure your BIOS and Drivers are up to date before you start, I suggest using CPUz to get Live actual CPU speed & voltage information, the lastest CoreTemp is you best bet for watching CPU core temps, Orthos does a great job of stress you CPU & RAM to in limits (use blend test) and Windows task manager will confirm your CPU is at 100% load on all cores.

Just take your time, and test each new setup with orthos as you go.