Is an E2160 my bottleneck? or something else?

[Cheezor722]

My computer has been feeling sluggish lately and I'm looking to upgrade it. My price range is $400-$500 dollars (could be stretched a little if it was really worth it). Regular performance in Windows XP just doesn't feel very snappy to me. I'd like things to load faster and programs to run smoother. Its gaming ability is also slowing down a bit. I feel like i'm pushing my system's to its limit when running games at High settings.

Here's my current System:

Abit IP35-E rev 1.4
Intel E2160 (350FSB x 9 = 3.15ghz @ 1.325vCore) w/CoolerMaster TX2
A-DATA 4GB(2 x 2GB) DDR2 800 -- 876mhz 5-5-5-18@1.95V
eVGA 8800gts 512mb (G92) @ 740/1800/1040
Antec NeoPower 550w
Antec P182 w/3 x Scythe Slipstream
Samsung Spinpoint F1 HD103UJ 1TB
Creative Labs Sound Blaster Audigy 2 ZS
Windows XP Pro 32-bit
Samsung Syncmaster 2493hm @ 1920x1200

Games I'm Currently playing: Company of Heroes, Dragon Age Origins, Warhammer 40k: Dawn of War 2, Mass Effect, Empire: Total War, Crysis

Games I want in the future: Mass Effect 2, Starcraft 2, Assassin's Creed 2, Warhammer 40k expansion, Crysis 2, Modern Warfare 2



These are the options I am considering for upgrade:
1) CPU + Mobo + Ram (thinking about i5 750)
2) Graphics Card (thinking about 5850)
3) SSD (dunno?)
4) Suggestions?



I just picked up Windows 7 Home Premium but haven't had time to install it yet. I was hoping that might liven things up a bit.

One other thing to note, I upgraded my Hard drive about 6 months ago from a Seagate 7200.10 320gb -> Samsung Spinpoint F1 HD103UJ 1TB. I used Norton Partiotionmagic to copy an image of the old drive to the new one and then resize the partition to the drive. Ever since then my load and wait time performance seems really crappy, despite the new drive benchmarking higher.

I've done a bunch of research in the past few weeks and have a good idea of the parts out there, but I'd like to know what you think would be the most worthwhile for my upgrade budget.

 

[silent-circuit]

You need a better CPU (I'd recommend a Q9540/9550 if you don't want to change platforms -- look for once used, new prices are robbery at this point) or going i5 / used i7, and you need a new GPU; the 8800 GTS 512MB was capable a year or two ago but it's pretty much bare minimum for running 1920x1200 at this point.

 

[Cheezor722]

Would upgrading to a i5/i7 get me snappier performance on the desktop? Just web surfing or loading applications?

 

[lt_shiro]

I would upgrade the Board+CPU+RAM and SSD and later on get a New PSU and Video card.

 

[Sayth]

I use a core 2 duo 3.0 Ghz Dell optiplex 760 at work and it's no where near as snappy as my new i5 system in my sig. That's with win7 x64 installed on both. Get a good i5 rig and you'll be happy. As for your hard drive, theoretically imaging an exact copy would be fine but with the dozens of times I've had to do it at work, I always recomend full install. Keeps you from porting over all your background proccesses and crap to a new hdd too.

I've found win 7 x64 to be perfectly stable on both my home and office rig so install that and you'll feel a slight improvement too.

 

[silent-circuit]

Without a new GPU you're still going to be disappointed with game performance. The 8800 GTS 512MB just isn't enough for high settings in many titles at 1920x1200 anymore.

 

[defaultluser]

Would upgrading to a i5/i7 get me snappier performance on the desktop? Just web surfing or loading applications?

Let's lay this down for you:

SSD will make your apps start faster, and will speed game load times.

The Core i5 will make multitasking smooth, with 4 cores and Turbo Boost to give you fast single-threaded processing. Your existing system has just two cores and a tiny cache that doesn't translate well to smooth multi-tasking. It will also make sure your video card gets fed frame data more consistently in games.

You need a new GPU to replace your G92 there, it's not enough for 1920x1200 anymore. You can probably just turn-down some settings for now, but you will need a replacement soon.

The problem is, you have aspirations of the high-end, but no girth in your system to back it up. Why did you go 1920x1200 when you knew you wouldn't have the money to continue the vicious upgrade cycle? You're probably not going to get your money's worth out of the upgrade unless you go with an HD 5850. A 5770 would be an upgrade, but not enough of one to justify the purchase.

If I had $500 to spend TODAY, I'd get the Core i5 upgrade as a start, and then just add things as you get the money. The longer you can wait on the video card upgrade, the cheaper it will be, and the SSD you can add anytime. But both would be gutless if you tacked them onto your current machine.

i5 750 + Gigabyte mobo with SATA 6 USB 3 $320
http://www.newegg.com/Product/ComboDealDetails.aspx?ItemList=Combo.319613

2x2GB Corsair DDR3 1600 $93
http://www.newegg.com/Product/Produ...0145260&cm_re=4gb_DDR3-_-20-145-260-_-Product

Total = $413

This upgrade is even better because the SATA6 SSDs are just over the horizon, and your board will support them natively. I would probably upgrade your video next, and save the SSD for last.

 

[engiNURD]

I agree with luser. Other things to consider...

If you do get an SSD, use Win7 (TRIM). And as for which SSD, I'd recommend the Kingston (Intel based) or Intel 40GB. Load OS and main apps only onto it, with user data on your F1 (which is still fast, so don't replace it; install secondary games/apps or whatever won't fit on the SSD; Don't fill the SSD, 30GB at most, IMO.)

Coming from an E2160 @ 3.1Ghz, a Q9000 chip would greatly improve multi-tasking and would be cheaper than an i5 build. If you can find another 4GB of RAM on the cheap, create a RAMDisk and use that for your browser cache -- makes browsing a LOT faster, IMO (I typically have 30-40 tabs open).

HD 5850 for $280 @ dell right now.

quick calc: $280 HD5850 + $150 used C2Q9 + $60 used 4GB + $130 SSD = $620 ... $250 more and you have an i7 with 4GB, lol.

 

[silent-circuit]

Eh, a 4890 is roughly as fast as a 5770 (a little faster, in fact) and plenty for 1920x1200 at highest settings with 2-4x AA in most titles. It's a pretty big jump from a G92 based card, based on my own experiences with 8800 GTXs, which are only a little faster than a G92 in most cases. A single 4890 is a little faster than 8800 GTX SLI, assuming good scaling. That's a substantial boost.

That said, I'd never recommend anything less than a 5870 for 1920x1200 if you're buying now. I guess a 5850 if you simply can't get a 5870 in budget. Personally I'd wait and see what Fermi does to graphics card prices shortly here. If the OP has held on to that G92 this long another month or two won't be the end of the world.

 

[Shadowssong]

Eh, a 4890 is roughly as fast as a 5770 (a little faster, in fact) and plenty for 1920x1200 at highest settings with 2-4x AA in most titles. It's a pretty big jump from a G92 based card, based on my own experiences with 8800 GTXs, which are only a little faster than a G92 in most cases. A single 4890 is a little faster than 8800 GTX SLI, assuming good scaling. That's a substantial boost.

That said, I'd never recommend anything less than a 5870 for 1920x1200 if you're buying now. I guess a 5850 if you simply can't get a 5870 in budget. Personally I'd wait and see what Fermi does to graphics card prices shortly here. If the OP has held on to that G92 this long another month or two won't be the end of the world.

The 4890 is plenty faster than the 5770 and nothing less than 5870 for 19x12? lol

 

[silent-circuit]

The 4890 is plenty faster than the 5770 and nothing less than 5870 for 19x12? lol

I said the 4890 is a littler faster than a 5770... which it is. Not sure what you meant by the first part of that statement-question-thing.

I also said I personally wouldn't recommend anything less than a 5870 for 1920x1200 if you are buying now (and taking in to account the OPs desire for high settings) and, to clarify, buying new.

When I buy I tend to buy high end. I went from an 8800 GTX to 8800 GTX Tri-SLI, held on to that for a while, then to 4890 CrossfireX. Were I to replace my current setup, it would probably be with a pair of 5870s, though I'd wait for Fermi to release first and hopefully drive prices down a bit.

I'm not that impressed with current hardware, though; my 4890 x2 setup is generally as fast if not faster than a 5870, DX11 is simply not here yet, and I don't boot my tower often enough to worry over power draw anymore.

 

[obobski]

of those games, ignore Crysis' performance, it won't run "good" on most hardware, period

Mass Effect, DA, DoW, ME2, etc should run fine, not full max, but should be fine
StarCraft 2 is unreleased, nobody can tell you what performance you'll get, period

I'd suggest a few free things first
firstly, if you're saying that performance has gotten worse over time, for example Mass Effect runs worse today than when it was new, that isn't "I need hardware lololol", thats "I need to clean the crap out of my system", so look into Spybot, CCleaner, etc, unload senseless background crap, you know, basic good usage habits

most of the games you're looking at are RPGs or RTSs so massively high framerates aren't required, and minor changes to detail settings can dramatically change performance, so keep that in mind with considered upgrades

if you are going to upgrade, I would agree with silent-circuit, wait for Fermi to release and stir prices up a bit, worst case scenario all it does is pressure AMD to drop their prices, which is still good for you (best case it doesn't cost a ton itself or provides more performance/features itself)

::edit

I said the 4890 is a littler faster than a 5770... which it is. Not sure what you meant by the first part of that statement-question-thing.

I also said I personally wouldn't recommend anything less than a 5870 for 1920x1200 if you are buying now (and taking in to account the OPs desire for high settings) and, to clarify, buying new.

When I buy I tend to buy high end. I went from an 8800 GTX to 8800 GTX Tri-SLI, held on to that for a while, then to 4890 CrossfireX. Were I to replace my current setup, it would probably be with a pair of 5870s, though I'd wait for Fermi to release first and hopefully drive prices down a bit.

I'm not that impressed with current hardware, though; my 4890 x2 setup is generally as fast if not faster than a 5870, DX11 is simply not here yet, and I don't boot my tower often enough to worry over power draw anymore.

I also basically agree with everything said here
buying in the "upper end" is usually a better choice with graphics hardware, you end up upgrading less, and if you've got some sense about it, you can save some money over the long run

 

[engiNURD]

Eh, I guess it would be a nice boost (HD4890 over GTS260), but still not worth it for 1900x1200, IMO.

... I also basically agree with everything said here
buying in the "upper end" is usually a better choice with graphics hardware, you end up upgrading less, and if you've got some sense about it, you can save some money over the long run

... and upper end is basically needed for such high res gaming. Though, as you said, he's not playing FPS games mainly, so maybe not.

 

[Cheezor722]

Wow. Thanks for all the information and advice. Lots of insightful stuff here.

I didn't realize my system was so old and obsolete. lol. It seemed to be running ok to me with games around where I wanted them. But now that you guys mention it i suppose I've been cutting back game settings slowly over time. It's just an old tin can now

So here's what I gather from reading everyone's info:
1) Don't get an SSD. This should be prioritized last, since the rest of my system is so obsolete now.
2) Get a 5850. It will take a performance hit if added to my current rig. Also, wait for Fermi to come out for prices to drop. But I play games that are not graphics/fps intensive, RTS and RPGs, so maybe a super GPU is not necessarily imperative?
3) I have a lot of choices for a processor. If i don't upgrade to a new platform, I should just pick up a 775 quad core. This option does seem appealing, but i feel like i'd be buying obsolete technology. I'm leaning more toward the move to i5/i7.
4) I should just spend more money!

I'm really in turmoil over whether to spend more money or not. I'm now a believer that I should buy mainly on the high mainstream end as the components will last longer before having to upgrade (as obobski mentioned). It sounds to me if I have to wait for Fermi that I should just upgrade my cpu/mobo/ram.
defaultluser pointed these out.

i5 750 + Gigabyte mobo with SATA 6 USB 3 $320
http://www.newegg.com/Product/ComboD...t=Combo.319613

2x2GB Corsair DDR3 1600 $93
http://www.newegg.com/Product/Produc...-260-_-Product

Total = $413


I've also been looking into the MSI GD65 and the Asus P7P55D. Any other options I should be considering? Also, any other options for DDR3 Ram? (I'm totally lost as to the difference between companies). Lastly, i5 750 vs i7 860? Does Phenom II make sense?

I've read a lot of the articles from anandtech, tomshardware, hardocp, and spcr so I have a general idea of how things are laid out. But with all the tradeoffs between price, performance, and other underlying technologies I get a headache thinking about it.

@defaultluser You mentioned a tiny cache affects multi-tasking. Do you know of any articles I can read that explain this phenomenon? I'm really curious about it.

 

[defaultluser]

@defaultluser You mentioned a tiny cache affects multi-tasking. Do you know of any articles I can read that explain this phenomenon? I'm really curious about it.

I'll tell you what, there is no article that will explain it better than this simulator:

http://www.ecs.umass.edu/ece/koren/architecture/CacheTr/cacheSim.html

To use this, select your settings and press the following every time you change something

Update Configuration
Generate Memory References: shows the in-order memory accesses for each task.
Sechedule Tasks: shows the true order in which simulated tasks will access the cache in a multi-threaded environment.

Then you can step through the run, and see each piece of memory get placed into cache. If the data is in cache, it will be registered as a hit and read. If it is not in-cache, it will be registered as a miss, and the data will be loaded from main memory into cache.

An experiment: with just 1 task, you'll notice that after the first run, there will never be a cache miss, since all the data Task A needs is in the cache. If you change the number of tasks from 1 to 8, you'll put more load on that small cache. Can you see how, if you increase the number of tasks or increase their number of memory accesses, the data in cache tends to get kicked-out every time another tasks takes control? This means that when Task A finally gets a chance again, it has to read everything back from main memory (very slow). Try increasing to 16 tasks or more, and watch the problem get worse.

Just understand that cache is much smaller than main memory, and every location in cache maps to MANY locations in main memory. And if both Task A and Task B want data that maps to the same cache location, they will overwrite the other's data each time you switch tasks.

A larger cache means less data is obliterated by multiple independent tasks, because each cache location refers to a smaller number of locations in main memory. There is more space for the tasks to co-exist instead of overwriting the other's data. Experiment: double the size of the cache to 32, and watch as the tasks stop overwriting each other as much. Even with 8 tasks running, a number of cache lines will still contain Task A data at the end of the run, which means when the task runs again, it won't have to go entirely to main memory for data. Thus, task switching is faster and more fluid on a system with more / better cache.

A lovely article on cache mapping if you're interested (but fairly advanced)

http://www.bowdoin.edu/~allen/courses/cs220/lab7/notes.html

 

[obobski]

... and upper end is basically needed for such high res gaming. Though, as you said, he's not playing FPS games mainly, so maybe not.

yeah, he still needs some horsepower, but if you're getting 25 FPS or 75 FPS in StarCraft, it doesn't matter, its an RTS

and most RPGs still feel ok at 20-25 FPS, even first person/real-time ones (like Fallout 3, Mass Effect is a bit slower paced)

 

[Cheezor722]

I'll tell you what, there is no article that will explain it better than this simulator:

http://www.ecs.umass.edu/ece/koren/architecture/CacheTr/cacheSim.html

To use this, select your settings and press the following every time you change something

Update Configuration
Generate Memory References: shows the in-order memory accesses for each task.
Sechedule Tasks: shows the true order in which simulated tasks will access the cache in a multi-threaded environment.

Then you can step through the run, and see each piece of memory get placed into cache. If the data is in cache, it will be registered as a hit and read. If it is not in-cache, it will be registered as a miss, and the data will be loaded from main memory into cache.

An experiment: with just 1 task, you'll notice that after the first run, there will never be a cache miss, since all the data Task A needs is in the cache. If you change the number of tasks from 1 to 8, you'll put more load on that small cache. Can you see how, if you increase the number of tasks or increase their number of memory accesses, the data in cache tends to get kicked-out every time another tasks takes control? This means that when Task A finally gets a chance again, it has to read everything back from main memory (very slow). Try increasing to 16 tasks or more, and watch the problem get worse.

Just understand that cache is much smaller than main memory, and every location in cache maps to MANY locations in main memory. And if both Task A and Task B want data that maps to the same cache location, they will overwrite the other's data each time you switch tasks.

A larger cache means less data is obliterated by multiple independent tasks, because each cache location refers to a smaller number of locations in main memory. There is more space for the tasks to co-exist instead of overwriting the other's data. Experiment: double the size of the cache to 32, and watch as the tasks stop overwriting each other as much. Even with 8 tasks running, a number of cache lines will still contain Task A data at the end of the run, which means when the task runs again, it won't have to go entirely to main memory for data. Thus, task switching is faster and more fluid on a system with more / better cache.

A lovely article on cache mapping if you're interested (but fairly advanced)

http://www.bowdoin.edu/~allen/courses/cs220/lab7/notes.html

Thanks for the info defaultluser. It was interesting playing with the simulator watching the hitrate quickly decline.

In reference to your second article, are any of those memory mapping methods the current way intel implements their cache? Or are these old methods from the 70's and 80's?

 

[obobski]

Thanks for the info defaultluser. It was interesting playing with the simulator watching the hitrate quickly decline.

In reference to your second article, are any of those memory mapping methods the current way intel implements their cache? Or are these old methods from the 70's and 80's?

while its all very informative and such
none of it matters for you

why, you might ask?

abstraction.

 

[defaultluser]

Thanks for the info defaultluser. It was interesting playing with the simulator watching the hitrate quickly decline.

In reference to your second article, are any of those memory mapping methods the current way intel implements their cache? Or are these old methods from the 70's and 80's?

Most chip makers use set-associative mapping for their chips today. For smaller caches, fully-associative could also be beneficial, especially when you want the highest cache usage and highest hit-rate possible (think L1 cache).

Set-associative cache is the best choice for most mid-sized cache workloads because:

1. The control logic is cheap, and is also fast. This is in-comparison to fully-associative cache (the control logic for full-associative cache balloons quickly as the size increases).
2. The cache is more likely to be completely used, unlike the case of a direct-mapped system. The direct-mapped system means that each place in memory can refer to just one cache line, whereas in the set-associative cache, each memory address can be mapped to X cache lines, where X is the "set-associativity" value.

But I think they might use direct-mapped cache for the larger L3 cache on modern processors. As the cache size grows, so does the latency, so you want as little logical overhead as-possible. But this is not the kind of thing they put inside of glossy marketing brochures - you'll have to read their technical documents to know for sure.