Archived

This topic is now archived and is closed to further replies.

Guest triangle

Upgrade results - Q9550

Recommended Posts

Hi allI have completed rebuilding and testing my machine (original specs in my sig - will change later). I went from a Q6600 B3 (2.4ghz stock) which would not overclock beyond 2.7ghz with good air cooling to a Q9550 (2.83ghz stock) (note to Sam - reseller I have been using for 10 years was v kind and sold me the Q9550 at a highly discounted price so chose that over a Q6600 C0).I also installed an MSI Platinum X48C mainboard - do not go there unless you have alot of patience and at least enough knowledge to get down and dirty with your BIOS. The X48C was barely operable at the start and a flash to the latest official BIOS did nothing - rarely booted past POST, ran the FSB at 50%, wouldn't recognise more than one DIMM, set all DIMM voltages and timings incorrectly etc. The MSI engineers acknowledged the board was highly problematic at release and gave me a beta release they had just completed and I spent a full 16hr day getting the board stable (three of which were dedicated to figuring out how to flash from a DOS bootable USB flash drive ... LOL).Also added 4gb 1333mhz DDR3 RAM as previously only had 800mhz DDR2 RAM.Anyway, the plan was to complete the upgrade, guarantee stability running Prime95 overnight and then start overclocking. Am currently sitting stable on 3.5ghz and there is obviously headroom to go a fair bit further - the 45nms chips suck very little voltage compared to the 65nms and generate much less heat.I benched FSX using FSX Benchmark 0.7 (high config settings) before upgrading and when I had the machine stable at 3.3ghz. The chart is below - I am really only providing this to give any interested people an idea of the FPS differential you might achieve between different CPU speeds. All up, average FPS increased 33%.FYI ...Andrewhttp://forums.avsim.net/user_files/191355.jpg

Share this post


Link to post
Share on other sites
Help AVSIM continue to serve you!
Please donate today!

Andrew,Thanks for posting this as it is always good to see a before and after upgrade benchmark run. If you recall my test runs, I was returned about a 68% FPS increase efficiency for a CPU overclock on the same system. Your results show a nearly 1 for 1 gain (ie. 100% return on CPU clock boost to FPS improvement), which can be accounted for by changing to a new gen motherboard, DDR3 RAM and core efficency improvements of the Q9xx over the Q6xx.Gary

Share this post


Link to post
Share on other sites

Thanks, Gary. Yes, I was pleasantly surprised by the 1:1 performance gain. Am still to fly a proper flight (bit over the computer at the moment after the days of dramas getting everything sorted)but hopefully the gain will translate into a smoother, more reliable flight experience overall.Andrew

Share this post


Link to post
Share on other sites

Don't know what might have been happening, but the Q6600 runs precisely 1:1 with frames v clock. These are the same FSmark runs at the same canned GLobal high settings at 2.4 v 3.6. Average FPS increase it exactly 50%. 2.4 to 3.6 is also exactly 50%.http://forums.avsim.net/user_files/191363.jpg BTW, set your TEXTURE_BANDWIDTH_MULT=40 to 70-ish. That will smooth out a lot of that FPS chop we're seeing in these charts.

Share this post


Link to post
Share on other sites

Very interestingFrom a performance standpoint, with your system(s), can you guys run a payware plane (like LvlD 767) at major airports above 20fps?I'm just curious to see what I may need for an enjoyable flight. Right now I use FSX mainly for GA flying in smaller communities and rural areas. I get great FPS with payware GA but once I switch to something faster than oh say 250kts the scenery gets blurry. Major airports aren't very fun either - 10-15 fps depending on the plane.

Share this post


Link to post
Share on other sites

With my Q66@3.6, from my GA and default airplanes settings I have to give back: 1) All AG except one notch. 2) All traffic. That get's me ~20 ~ clean FPS on approach with the PMDG 744X.The blurry scenery is about the number of cores involved. Tweaking settings won't help that.

Share this post


Link to post
Share on other sites

Impressive results I would be happy to break 20fps. How much of being able to increase it that far can be attributed to the DDR3 or was it mainly just moving to a faster core giving you more head room?

Share this post


Link to post
Share on other sites

None attributable to DDR3 - this board has outstanding issues and I have the DDR3 (1333mhz rated) running at 820mhz while I wait for the next BIOS update. I managed the same on DDR2 800mhz on the same board. The overclock is a direct function of the CPU (and it's low volts and low heat), the mainboard's detailed overclocking ability and the CPU cooling solution. This was the easiest overclock ever for me because of these three factors alone ...Andrew

Share this post


Link to post
Share on other sites

So don't buy an MSI board then? Well thats very good information to know.

Share this post


Link to post
Share on other sites

I didn't say that (but I understand why you would) .. this board is awesome but it needed alot of hands on work to get it working and, yes, the BIOS needs to mature. So don't go there unless you are prepared to work and wait a month or two for all the BIOS issues to get ironed out by new releases. I am - this board has enormous potential and I am both patient and very stubborn :-)Andrew

Share this post


Link to post
Share on other sites

However O/Cing should never be a big deal, For instance all I did on my plain-jane p35 was: 1) Increase the FSB to 400Mhz and 2) Increase the CPU's voltage. 3) Go Fly. That got a CPU to 3.6 that's been stable for ages. Mobo version and ram type/speed are all performance neutrals. There's no need for advanced O/Cing capabilities other than an auto function that works.

Share this post


Link to post
Share on other sites

>With my Q66@3.6, from my GA and default airplanes settings I>have to give back: >>1) All AG except one notch. >>2) All traffic. >>That get's me ~20 ~ clean FPS on approach with the PMDG 744X.>>The blurry scenery is about the number of cores involved.>Tweaking settings won't help that. You're saying you have to turn OFF AI traffic?!?!And most the autogen? Yuck....

Share this post


Link to post
Share on other sites

>Don't know what might have been happening, but the Q6600 runs>precisely 1:1 with frames v clock. These are the same FSmark>runs at the same canned GLobal high settings at 2.4 v 3.6.>Average FPS increase it exactly 50%. 2.4 to 3.6 is also>exactly 50%.>>>http://forums.avsim.net/user_files/191363.jpg> >>>>>>>BTW, set your TEXTURE_BANDWIDTH_MULT=40 to 70-ish. That will>smooth out a lot of that FPS chop we're seeing in these>charts.>>>>Hi There,Do you agree that the upgrade from your Q6600 did little to the average frame-rates ? From the diagrams it does look that way.Clearly the larger L2 cache on the 9550, really does nothing for FSX and I assume the FSB with both the 6600 and the 9550 were very similar.As for DDR3, it is not likely you will see any difference with such low FSB speeds. The extra bandwidth 'potential' are rather wasted at at the low FSB's of today's chips. Of course once the Nehalems are out, things may well change and DDR3 may come into its own.The really interesting question of course is twofold: (1) Would being able to push a Quad to over 4Ghz, continue the linear increase in frame-rates seen between 2.4 and 3.4.....or does diminishing returns set in ? When I clocked a core 2 duo, the biggest increase in FPS was found between 2 and 2.8 Ghz. Beyond that, each increase in speed brought about an every declining increase in FPS. You will find that increasing from 3.4 to say 3.8 will probably show no significant increase in performance.This should work the same in reverse. Try dropping your speed to 3.2...any change in FPS....thought not. All the increase is found up to around 2.8 to 3Gz...nothing much over that except more power consumption and heat....and lower life expectancy.(2) Would an existing E8500/8400 (and indeed the impending G0 steppings of these / E8600) running at 4.2-4.8 Ghz (very likely before the end of the year !) match the Quad at either 3.4 or (theoretically) 4.2Ghz. There is little evidence to suggest that Quads actually give higher frame-rates, and any improvement in "smoothness" or "Blurries", may arguably, be a 'placebo' effect linked to the cognitive dissonance that kicks-in when you buy something expensive with great expectations of it....!Just my two cents....Cheers

Share this post


Link to post
Share on other sites

** Do you agree that the upgrade from your Q6600 did little to the average frame-rates ? From the diagrams it does look that way.My upgrade was from a P4@ 3.2. The core2 quad was a miracle! ** Clearly the larger L2 cache on the 9550, really does nothing for FSX and I assume the FSB with both the 6600 and the 9550 were very similar.Cache is a marketing tool. FSB speed is irrelevant to performance. ** As for DDR3, it is not likely you will see any difference with such low FSB speeds. The extra bandwidth 'potential' are rather wasted at at the low FSB's of today's chips. Of course once the Nehalems are out, things may well change and DDR3 may come into its own.The Nehalem will have a 133Mhz FSB. The current platforms have FSB speeds that are MaSSiveLy beyond any performance requirement. The current architecture uses the FSB to drive a CPU clock. Notice the FSB speeds have been increasing ONLY to lure a lay audience and to thwart a knowledgeable user by lowering the CPU's multiplier. **The really interesting question of course is twofold:** (1) Would being able to push a Quad to over 4Ghz, continue the linear increase in frame-rates seen between 2.4 and 3.4.....or does diminishing returns set in ? When I clocked a core 2 duo, the biggest increase in FPS was found between 2 and 2.8 Ghz. Beyond that, each increase in speed brought about an every declining increase in FPS. You will find that increasing from 3.4 to say 3.8 will probably show no significant increase in performance.This should work the same in reverse. Try dropping your speed to 3.2...any change in FPS....thought not. All the increase is found up to around 2.8 to 3Gz...nothing much over that except more power consumption and heat....and lower life expectancy.See pic. These are runs at 2.4, 3.3 and 3.6Ghz on a Q6600. The center run is at 3.3Ghz. The ramp is liner. I expect it would continue. However remember, FPS is Single Core related. Multi core does NoT influence FPS in FSX.http://forums.avsim.net/user_files/191394.jpg** (2) Would an existing E8500/8400 (and indeed the impending G0 steppings of these / E8600) running at 4.2-4.8 Ghz (very likely before the end of the year !) match the Quad at either 3.4 or (theoretically) 4.2Ghz. There is little evidence to suggest that Quads actually give higher frame-rates, and any improvement in "smoothness" or "Blurries", may arguably, be a 'placebo' effect linked to the cognitive dissonance that kicks-in when you buy something expensive with great expectations of it....!I expect any higher clock on Core 0 will increase FPS. It will NoT matter how many other cores are onboard. However it bears repeating that FPS are NoT the holy grail of FS. Smoothness is the key. THIS is where additional cores can help with scenery loading chores. ** Just my two cents....Me too.

Share this post


Link to post
Share on other sites

I understood from your thread that you had upgraded from a Q6600 to a Q9550 ? I must have misunderstood.....ie. the result from both Quads are not significantly different.I was making the point that it didn't seem to have any benefit - precisely for the reasons you allude to ! Fast RAM, DDR3, 1666 Bus Speeds etc...really don't benefit gaming. so, buying a 9550 Quad is not financial sense if an overclocked Q6600 does just as well !As for the purported differences between 3.2 and 3.4...I don't think that a 5 FPS difference is greater than the inherent variability in FSX frame-rates. I would argue, that statistically, there is therefore, no real difference between 3.2 and 3.4.I also hypothesize that increasing above 3.4 would yield even less advantage. Extrapolation outside the limits is very dangerous...diminishing returns will set in quickly. This is why Tom's Hardware guide etc. only ever use a 3.3Ghz Quad to test graphics cards...they know that a faster CPU won't improve matters, at at 3.3 the cards will not be bottle-necked.You state - as fact - that FPS is not influenced by nos. of cores. How do you know this? I have never seen a benchmark test - good or bad - that uses a Core 2 Duo at over 4Ghz speed. How do you know that there won't be a 'Stepped' effect ? Also, have you ever played FSX at a huge resolution such as 2560x1600 ? It may have an effect when GPU starts to become limiting.You see, this is why I posed these "questions" - everyone seems to have the answer, but none has proved it.I do concur that there is more to "performance" than speed and would imagine (imagine - being the operative word) that a Quad would be better for texture loading etc.Anyhow, I intend to acquire a Quad and a very fast Core Duo at around 4.5 or 4.8Ghz...and do some testing.It will be interesting to see if a very fast Core Duo will increase frame-rates much above a Duo running at 3.3/3.4 - I doubt it will, and the 'qualitative' benefits of a slower Quad, may well be preferential to a small increase in frame-rates.Later.Time will tell.

Share this post


Link to post
Share on other sites

>You state - as fact - that FPS is not influenced by nos. of>cores. How do you know this? I have never seen a benchmark>test - good or bad - that uses a Core 2 Duo at over 4Ghz>speed. How do you know that there won't be a 'Stepped' effect>? Also, have you ever played FSX at a huge resolution such as>2560x1600 ? It may have an effect when GPU starts to become>limiting.>>You see, this is why I posed these "questions" - everyone>seems to have the answer, but none has proved it.You need to look harder in these forums then, as all bar the over 4GHz config have been tested and reported by myself and others in this forum before.>Anyhow, I intend to acquire a Quad and a very fast Core Duo at>around 4.5 or 4.8Ghz...and do some testing.I look forward to your thorough analysis of these configs. :-)Gary

Share this post


Link to post
Share on other sites

Well someone needs to do one - it's the over 4Ghz situation that's sadly missing.As for benchmarking I reckon the difficult part is creating b/marks that are actually replicable and measurable at settings that are relevant to users and high enough for "real" differences due to h/ware can be measured.

Share this post


Link to post
Share on other sites

>As for benchmarking I reckon the difficult part is creating>b/marks that are actually replicable and measurable at>settings that are relevant to users and high enough for "real">differences due to h/ware can be measured.That is precisely why I created FSXMark07, available here in the Avsim library. Consider using it when you get your hardware bits together for comparison. If you think it misses the mark, by all means come up with a better benchmark and share it with the rest of us.Gary

Share this post


Link to post
Share on other sites

I disagree - motherboards that provide poor incremental voltage changes are never as good as those that give you a finer voltage tuning ability. You'll screw more out of your CPU if you have a board that gives you the capacity to modify all your voltages at fine levels. Just depends on whether or not you want to spend the time doing so ... and most people, understandably, do not. This board, however, needs work at the MSI end, but shows great capacity - i can get can alot more out of this CPU than I could with my previous p45 board.A :-)

Share this post


Link to post
Share on other sites

Cool - I've been looking for that FSXMark thingey - there are around 15 hits on google...couldn't find it anywhere.Thnx....found it.....I assume it can be modified to use different aircraft...to a degree I guess....given it uses the default AP system.Cheers I'll take a look....with interest.

Share this post


Link to post
Share on other sites

And, if you have stability issues, increase the MCH and FSB voltage by 0.1 or 0.2 v also....400 is reasonable high for a standard m/b and may need this.

Share this post


Link to post
Share on other sites

If you disable A/gen completeyl, you will get a massive increase in prformance - surely if you're in a 747. why the need for it ? VFR, yes....you're better with a tad of Ai traffic to get some action in the airport and stuff Agen - unless it's properly designed over photscenery it's rubbish anyways.....just my 2 cents...!

Share this post


Link to post
Share on other sites

One thing I notice straightaway....is that traffic is used. Traffic takes a while to "get up to speed"....it may have significantly more effect towards the end of the benchmarking than at the beginning....given that the test is only 5 minutes, it may be having little or no impact...period.Why not try it with zero traffic - the results may well be the same, although given traffic is very CPU-intensive, the results when comparing faster CPU's / more cores, may well differ.Alternatively, could the benchmark begin further out where there is little in the way of traffic, so that by the time it reaches your current 'starting point'...say, 10 or 15 minutes has elapsed and traffic is functioning fully ?

Share this post


Link to post
Share on other sites

Also, I like the graphs you end up with...I think there may be a way of getting some proper statistical numbers from these.It will take one heck of a long time....

Share this post


Link to post
Share on other sites

A FSB running at 400MHZ was the design speed for the P35 (and its rebranded follow-ons the X38/38, et al) - from conception - . PCIe-V2 was also always onboard, only was only enabled with the X38. These chipsets are a marketing evolution designed to satiate a lay audience. Shame on them? Maybe, but after all, Intel has to eat too.

Share this post


Link to post
Share on other sites