What would be a good upgrade for this PC?

[Masherbrum]

Since no one has brought it up, the LENGTH of your Video Card will depend on the Case you have.  

[Chalenge]

Also something anyone interested in the biggest bang for buck is this test between the 5870 in Crossfire and the 480 in SLI:

http://www.anandtech.com/bench/Product/163?vs=159

Of special interest is the effect of the highest resolutions (in this case Crysis Warhead) and the affect of gamers quality and enthusiast shaders (its not even close). However if your monitor can only handle lower resolutions its obvious the Crossfire system is the way to go.

Then of course the temperatures and power consumption and noise levels come into play. The SLI  system is much warmer and uses much more power but it is on a par with noise.

@jimson: I know your system wont do more than one card. Thats not a bad thing since you can avoid heat issues and you wont need the bigger PSUs (yours is fine for most situations anyway). I also know there are games out there (RoF for one) that just wont use SLI (not sure about Crossfire). But for AH you just dont need a monster video card.

[guncrasher]

take a look at this site.  cpus, gpus, hd ratings all in one place.  but mainly post the case you are using, some cards may not fit.  if you have something like the tempest, than you can almost put a plane in there too.  they're big.


http://www.videocardbenchmark.net/

semp

[MrRiplEy[H]]

I still say the 9800 GTX is a waste and not enough improvement to warrant spending money over. In fact the performance benefit is just like the older SLI improvements that people were referring to as wasteful. Unfortunately the 9800s have slid right off most comparison sites already (because its not much different from the 8800s).

Anandtech has a video benchmark comparison page you should check out:

http://www.anandtech.com/bench/GPU/88

Bear in mind when using that that CF means crossfire (two cards) and SLI means scaleable link interface (two cards). Except for the fact that you cant use two cards I would suggest two 460s in SLI which outperform a 480. Probably your PSU wouldnt handle it though. I will suggest Evga because they have the best warranty going.

I would stay away from SSDs. The only thing (mostly) an SSD can do for you is program startup time. After Windows has started a program the first time every startup thereafter will be just as fast as an SSD because of the windows cache.

Chalenge is right on the videocard the 9800 was a notorious rip-off from Nvidia, essentially a rebadged 8800 performance wise. It would make little sense to purchase one if you already had a 8800. Better to invest a few bucks more to a current generation model.

Chalenge is kinda right about the SSD part too. Windows will cache programs after they've been run or superfetched, SSD will however affect bootup times and that first time run naturally. If your game needs to load a texture for the first time, it might be the time that the stutter kills your aim. However it's anyone's judgement whether that warrants an investment of 200 bucks or not. I wouldn't.

[columbus]

Also something anyone interested in the biggest bang for buck is this test between the 5870 in Crossfire and the 480 in SLI:

http://www.anandtech.com/bench/Product/163?vs=159

Of special interest is the effect of the highest resolutions (in this case Crysis Warhead) and the affect of gamers quality and enthusiast shaders (its not even close). However if your monitor can only handle lower resolutions its obvious the Crossfire system is the way to go.

Then of course the temperatures and power consumption and noise levels come into play. The SLI  system is much warmer and uses much more power but it is on a par with noise.

@jimson: I know your system wont do more than one card. Thats not a bad thing since you can avoid heat issues and you wont need the bigger PSUs (yours is fine for most situations anyway). I also know there are games out there (RoF for one) that just wont use SLI (not sure about Crossfire). But for AH you just dont need a monster video card.

i dont see the 300.00-400.00 in extra cost worth it

[Getback]

I was trying to stay out of this thread.

32 bit Windows OS's can address 4GB of RAM.  However, all memory addressable buffers (sound card, video card, ethernet....) are mapped starting at 4GB and then it goes down from there.  If you have a 256MB video card, then instantly subtract 256MB from 4GB to find the highest amount of memory the OS will be able to use for applications.

If you have a 1GB video card, subtract 1GB from 4GB (regardless of the physical amount of RAM in the computer), which yeilds 3GB.  This means 3GB will be the most system RAM available to the OS for applications.

The reason?  When all 32bits of a binary number are set, it is equal to 4GB (4294967296-1).  This is the largest directly addressable number for a 32 bit system.  However, if Microsoft would properly implement PAE (physical address extensions), then a 32 bit OS could address 64GB of RAM.  UNIX OS's have been doing it for years.

That I sure didn't know.

May I ask, If you have a 1gig video card does that not mean the vc is carrying its own memory? If that be the case, then why would it subtract from the ram? Is it because MS allocates so much to Ram for certain things such as sound and Video and when a VC is added it says, okay, we don't need this in RAM?

[eagl]

getback,

Here is an analogy.

One man has 10 fingers but no toes.  He can only count on his fingers.  If you give him 5 apples and ask him how many he has, he will count up to 5 and answer correctly "5".  If you give him 10 apples, he will answer 10.  If you give him 11 apples however, he still answers "10".  Easy enough, right?  But what happens if in addition to those 10 apples, you give him 4 oranges that he MUST count first, before he starts counting apples?  If he has 4 oranges and 10 apples, but you ask him how many apples he has, he will answer "6".  With 32 bit windows, the apples are system RAM, oranges are video RAM, and whenever you add more oranges, they have to be counted before you start counting the apples.  That's just how it works.

With 64 bit windows, imagine that same dude, but with 1000 hands instead of 2.  He can count a hell of a lot of apples and oranges...  Probably more than we need for the next few years until memory density and bus bandwidth catches up to address space.  Being able to count 400,000,000,000,000,000 apples doesn't help much if it takes you a week to retrieve each one before you can eat it, and that is the reason why although you see 32 bit windows systems shipping with their "max" of 4GB installed, you don't see 64 bit windows systems shipping with the full amount of memory they can theoretically address installed.  The memory isn't that dense and it would take too long to transfer all that information over the current memory and cpu bus anyhow.  Eventually the subsystems will catch up, but it will take a while since bus bandwidth is lagging the fight right now.

BTW I see this as being the next major impetus to improving home LAN speeds in the next few years...  5 years ago, I could transfer the entire contents of the largest HD on the market from one computer to another in under an hour, over a cheap home lan.  Today, transferring an entire 3TB drive from one computer to another over even gigabit ethernet can take a day or longer, depending on the speed of the computers and the quality of the network components end to end.  The bottleneck will be information transfer once again, just like in the days of the 56k modems.  Next up - fiberoptic home lan starter kits.

[Getback]

getback,

Here is an analogy.

One man has 10 fingers but no toes.  He can only count on his fingers.  If you give him 5 apples and ask him how many he has, he will count up to 5 and answer correctly "5".  If you give him 10 apples, he will answer 10.  If you give him 11 apples however, he still answers "10".  Easy enough, right?  But what happens if in addition to those 10 apples, you give him 4 oranges that he MUST count first, before he starts counting apples?  If he has 4 oranges and 10 apples, but you ask him how many apples he has, he will answer "6".  With 32 bit windows, the apples are system RAM, oranges are video RAM, and whenever you add more oranges, they have to be counted before you start counting the apples.  That's just how it works.

With 64 bit windows, imagine that same dude, but with 1000 hands instead of 2.  He can count a hell of a lot of apples and oranges...  Probably more than we need for the next few years until memory density and bus bandwidth catches up to address space.  Being able to count 400,000,000,000,000,000 apples doesn't help much if it takes you a week to retrieve each one before you can eat it, and that is the reason why although you see 32 bit windows systems shipping with their "max" of 4GB installed, you don't see 64 bit windows systems shipping with the full amount of memory they can theoretically address installed.  The memory isn't that dense and it would take too long to transfer all that information over the current memory and cpu bus anyhow.  Eventually the subsystems will catch up, but it will take a while since bus bandwidth is lagging the fight right now.

BTW I see this as being the next major impetus to improving home LAN speeds in the next few years...  5 years ago, I could transfer the entire contents of the largest HD on the market from one computer to another in under an hour, over a cheap home lan.  Today, transferring an entire 3TB drive from one computer to another over even gigabit ethernet can take a day or longer, depending on the speed of the computers and the quality of the network components end to end.  The bottleneck will be information transfer once again, just like in the days of the 56k modems.  Next up - fiberoptic home lan starter kits.

Welp, that's what I think I thought. Doesn't sound logical though.

[SectorNine50]

I don't remember where I read that but do remember reading it at several sources as I was preparing to OC my CPU.  It makes sense when you think about synching the CPU and RAM in that each core operates with relative indepencence.  As I'm sure you're aware there are times where a single core is taking almost the entire load.  I see this regularily on my system monitor.  My CPU is an E6750 2.66 Ghz Core2Duo OC'd to 3.2 Ghz on a 1600 FSB linked and synched to 4 Gb DDR2 800 Kinston HyperX (4-4-4-12).  Unfortunately I'm running 32 bit XP Pro and only see 3.25 Gb of my RAM (eVGA 8800 GTS G92 512 Mb GPU) but it does everything I need it to do so I'm happy.

Okay, so I just did a little reading myself and it cleared up a LOT:

The reason the FSB has a multiplier of 4 is because it sends data 4 times a data cycle.  If you imagine a sine curve (what a clock cycle looks like on a graph), the FSB will send a signal at the top, bottom, and every time it crosses the x-axis, which equates to a total of 4.

This means that even though a processor may be multi-core'd, the FSB is still considered "Quad Pumped" (aka QDR or Quad Data Rate) because when the data is sent from the bus is independent of the number of cores in operation.

So basically, the reason we want to match the system speed (FSB/4) to the real RAM speed (Mhz/2) is because in a Dual Channel memory system, the RAM will effectively pick up on all 4 data cycles of the FSB (each channel will pick up 2 of the 4).

FSB sends out 4 bits (QUAD DATA RATE) of data in one cycle (200Mhz x 4 = 800)
-DDR (DUAL DATA RATE) channel one picks up 2 bits of data in one cycle (200Mhz x 2 = 400)
-DDR channel two picks up the other 2 bits of data in one cycle (200Mhz x 2 = 400)

And if I'm understanding correctly, in a single channel memory situation, you want your true RAM speed to match the FSB speed/2 because:

FSB sends out 4 bits of data in one cycle (200Mhz x 4 = 800Mhz)
-Single DDR channel picks up 2 bits in one cycle (400Mhz x 2 = 800Mhz)
-Single DDR channel picks up 2 bits in another cycle (400Mhz x2 = 800Mhz)

So effectively, by the time the FSB has finished it's cycle, the memory has gone through 2 cycles to compensate for only being able to pick up 2 bits at a time.  Kind of interesting stuff, isn't it?

However, now that the memory controllers built into all the new Intel chips, Core i3', i5, i7, and i9, (which AMD has been doing for some time under the name of Hyper Transport) my understanding is that this will all go out the window, and ultimately faster RAM means better performance.

[Skuzzy]

That I sure didn't know.

May I ask, If you have a 1gig video card does that not mean the vc is carrying its own memory? If that be the case, then why would it subtract from the ram? Is it because MS allocates so much to Ram for certain things such as sound and Video and when a VC is added it says, okay, we don't need this in RAM?

I am not sure I understood eagls analogy.  I suck at them.

Ok, the problem is how does the CPU directly address the RAM on the video card?  The RAM on the video card has to be memory mapped so the CPU can address it directly.  This is how a lot of performance is gained for video cards.  Same with any hardware buffer.  The memory on the local device has to be given a physical memory address so the CPU can read/write data from/to it.

In a 32 bit operating system, the highest physical address the CPU can address is 4GB.  So, during the initilization of all the devices in the computer, any memory that needs to be mapped so the CPU can access it directly, is placed starting at the 4GB boundary and proceeds downward.

BIOS ROMS, on add-on cards, also have to be memory mapped so the CPU can run the BIOS code.  However, most of the BIOS roms will unmap themselves after they run, unless they need to BIOS code to be present at all times, then they stay loaded at the memory address they were assigned.

[Getback]

I am not sure I understood eagls analogy.  I suck at them.

Ok, the problem is how does the CPU directly address the RAM on the video card?  The RAM on the video card has to be memory mapped so the CPU can address it directly.  This is how a lot of performance is gained for video cards.  Same with any hardware buffer.  The memory on the local device has to be given a physical memory address so the CPU can read/write data from/to it.

In a 32 bit operating system, the highest physical address the CPU can address is 4GB.  So, during the initilization of all the devices in the computer, any memory that needs to be mapped so the CPU can access it directly, is placed starting at the 4GB boundary and proceeds downward.

BIOS ROMS, on add-on cards, also have to be memory mapped so the CPU can run the BIOS code.  However, most of the BIOS roms will unmap themselves after they run, unless they need to BIOS code to be present at all times, then they stay loaded at the memory address they were assigned.

Okay, I was off a notch on my thinking. Its starting to sink in though.

[Chalenge]

i dont see the 300.00-400.00 in extra cost worth it

Well look at it this way. If you buy a $200 card you will be happy for about a year or a little more. Then you will either upgrade with a second card just like it or a brand new better card. Sometimes when you want SLI you cant find the exact same card so you have to weigh buying two exact same cards or buy a brand new better card. Right now the 2## cards are cheaper because of the Fermis (4## series) but the Fermis are much better scale-wise than any previous series. You might think that $450 for a 480 is a lot of money and you would be right. You could also spend $700 on the latest 5970 for about the same performance. I would bet that a 480 or 5970 either one will not have to be replaced for at least two years and you could very easily push that further (especially with an Evga warranty should any card go bad).

To put an ugly rumor to bed... the 480s I am running now (not the original releases) run at the same temperature and noise level as 8800 GTs and my systems boot faster now too.

[SectorNine50]

Now I find something that completely shows the opposite about memory "sync-ing..."

http://www.anandtech.com/show/2146/6


EDIT:
And it seems this is why!

http://www.techspot.com/vb/topic9046.html

It makes sense.  If you go over sync speed, other components can access the RAM while the processor is not.  Therefor, you get a speed boost!

No performance loss for going over sync speed, but it's how much performance you gain per dollar that is in question.

[cattb]

Well look at it this way. If you buy a $200 card you will be happy for about a year or a little more. Then you will either upgrade with a second card just like it or a brand new better card. Sometimes when you want SLI you cant find the exact same card so you have to weigh buying two exact same cards or buy a brand new better card. Right now the 2## cards are cheaper because of the Fermis (4## series) but the Fermis are much better scale-wise than any previous series. You might think that $450 for a 480 is a lot of money and you would be right. You could also spend $700 on the latest 5970 for about the same performance. I would bet that a 480 or 5970 either one will not have to be replaced for at least two years and you could very easily push that further (especially with an Evga warranty should any card go bad).

To put an ugly rumor to bed... the 480s I am running now (not the original releases) run at the same temperature and noise level as 8800 GTs and my systems boot faster now too.

I have a late model 8800gt and it is the same chipset as the 9800, it is just about the same card. I would not upgrade to a 9800.
Video card prices are finally coming down, I would expect lower prices yet and christmas is coming. I also read demand is down for PC's and etc. Demand is suppose to stay soft for ahile, this is good for us the consumer. (If what I read is true)

[TequilaChaser]

You guys are forgetting the point here.. he doesn't want the hassle of overclocking, and all of that.

Jimson, really it all comes down to how much money you want to spend, and what you want to use it for.

Got $1000?  Get a new motherboard, i7 setup, 12GB RAM, and an ATI5850 and Win7 Home Premium 64bit.

Got a couple hundred bucks?  You can get a nice little upgrade going to an ATI 5830 or 5850.

If all you use the computer for is basic computer stuff (email, www, etc) and Aces High, then don't bother upgrading the Operating System, as that extra 750MB of Ram isn't going to make THAT MUCH of a difference.  If you DO decide to up the Ram further, then Win7 Home Premium 64bit is the way to go.  Don't let the name fool you, the "Ultimate" version really offers little more than Home Premium, and you're likely to find you won't even use those options.


I'll use Aces High as a benchmark here with my system.  I had the following:

• ASRock G43Twins-FullHD Motherboard
• Intel E7400 C2D (2.8GHZ, slightly OC to 3.0GHZ)
• 4GB DDR3 1066 RAM
• 9800GTX+ Video Card
• USB Turtle Beach Headphones.. nothing fancy about the hard drives or anything else
• Win XP Home (32bit)

I was able to handle Aces High at 512 textures sliders about 2/3 of the way to max and no shadows and run a steady 55-60FPS at 1920X1200 resolution.  1024 (hires) textures didn't seem to affect FR too much until I flew into a lot of action, fire, etc in which it would drop down to 25-30 at times.  Higher sliders dropped it down to 40 or so consistently, and shadows REALLY killed things.  This is all with a notch of AA in AH settings.  It actually seemed to perform slightly worse by disabling AA which doesn't make sense, but perhaps it liked to be stressed... and likewise 2 notches was too much for it... but 1 notch was JUST RIGHT.

Then I upgraded to Windows 7 Home Premium 64bit.  Didn't really see much of a performance increase in Aces High.  I would have to say... about the same.


Unless you specifically WANT TO upgrade to Windows 7, I can't really see a reason to recommend it unless you want to up the ram to 8GB or more... but I don't believe even that will make a significant increase in AH performance.

In a 32 bit operating system, the highest physical address the CPU can address is 4GB.  So, during the initilization of all the devices in the computer, any memory that needs to be mapped so the CPU can access it directly, is placed starting at the 4GB boundary and proceeds downward.

BIOS ROMS, on add-on cards, also have to be memory mapped so the CPU can run the BIOS code.  

I have watched this thread, and have to disagree with ya Tigger29, in regards to whether Win7 64 bit would be of a great benefit or not for Jimson.....

I would strongly suggest he upgrade to the Win7 Pro 64 bit .... he will benefit from the full use of his 4 gigs of ram plus his use of his video ram..... his computer specs he listed:

EVGA 512-P3-N802-AR GeForce 8800GT Superclocked 512MB 256-bit GDDR3 PCI Express 2.0 x16 HDCP Ready SLI Supported Video Card

GIGABYTE GA-EP35-DS3L LGA 775 Intel P35 ATX Intel Motherboard

E8400 cpu

PC Power & Cooling S75QB 750W ATX12V / EPS12V Power Supply

Pioneer 20X DVD±R DVD Burner Black IDE Model DVR-115DBK  

Creative 70SB073A00000 7.1 Channels PCI Interface Sound Blaster X-Fi XtremeGamer

Seagate Barracuda ES.2 ST3250310NS 250GB 7200 RPM SATA 3.0Gb/s 3.5" Internal Hard Drive

CORSAIR XMS2 4GB (2 x 2GB) 240-Pin DDR2 SDRAM DDR2 800 (PC2 6400) Dual Channel Kit Desktop Memory Model TWIN2X4096-6400C5

would see somewhat of a significant increase on the Win7 64 bit OS verses what he is seeing on the WinXP 32 bit OS....


I have witnessed 3 different builds , that showed wonderful increases just by upgrading the OS and using the existing PC components.... I built a PC with all 3rd or 4th generation PC parts and stuck Win7 64 bit on the HD and maxed out all settings with 75 fps and hires 1024 textures and shadows on 2048...

to where on the same pc using winxp 32 pro it fell down to 58 to 61 fps in light action..... and down to mid 40s in hvy action. on a 9500GT 512 meg DDR3 vidcard and 3 gigs of ram with an AMD opteron185 Dualcore 2.8 ghz cpu and using onboard sound......

my own PC gets a solid 5 to 6 FPS increase from 69/70 in winXP pro.... to solid 75 FPS in win 7 Ult 64 bit with my ASUS AMD athlon X2 6400+ 3.2 ghz dual core and 4 gigs corsair XMS2 pc6400 ram and creative xtremegamer prof soundcard and Asus 3870 512 meg DDR4 video PC P&C 750watt silencer PSU and Antec 900 case.....with same settings.... and I have 4 HD's in it 2 DVD DL RWs etc...and I built it like 3 or so yrs ago....... it still works great ( knocking on wood heheh )


Jimson,
if you can get ya hands on Win7 Pro 64 bit , go for it....... you'll most likely see an improvement... and may not even experience any flickering or drops for the most part . ( unless you use like 1900 x 1200  screen res )...... say 1600 or lower like me 1280 x 1024 and you be fine I'd think.....
if you can get a better video card, then do that too..... but comparing that old pc I throwed together to your current PC parts.. who should be able to smoke the results I got from that thrown together PC> with ease....

best of luck to ya...

hope this helps...