Opteron for gaming?

[GamingGod]

I dont know much about these processors, how do they do for gaming? If I were able to get ahold of a 146 how would it compare to a a64 3200+? Is there any compatibility issues or downsides to using one? I am mostly just a gamer.

[McBanjo]

According to the info I got then it's no difference between an opteron and an athlon when it comes to gaming.
Opteron might be a little bether since the quality is bether to get server-status and the memorycontroll (probibly a little bether than athlon) is a little different too but that's about it

[shimq1]

All the Opterons have 1MB cache so you don't have to get a 3700+ just to get 1MB cache, and it also overclocks better than Athlon 64's.

[GamingGod]

any reviews to compare them, how much of an overclock could a 146 take? Ive heard people say that a cheap opteron overclocked can be one of the fastest single cores out there.

[acaurora]

If you check out [H}ardOCP on their forums, they have a post with coupons from MonarchComputers. Right now, you can get an Opteron 165 (1 MB x 2, 1.8 Ghz) for only about $275 after taxes/SH. Many people on there are reporting overclocks of UP TO 2.8GHz, and that’s on stock cooling. However, I didn’t read all 45+ pages of posts, only the first few and the last, lol. However I’m too much strapped for cash to get one, I am fine with my single core 3700+ San Diego

[McBanjo]

any reviews to compare them, how much of an overclock could a 146 take? Ive heard people say that a cheap opteron overclocked can be one of the fastest single cores out there.

Not really true. A faster cpu by standard is more likely to get more Mhz if you overklock it compered to a slower one. But you can get a cheap opteron to beat a standard FX-60.
The % overclock is generally higher on a slower/cheaper cpu as well

Is that what you meant or have I missed something?

[Mats]

any reviews to compare them, how much of an overclock could a 146 take?

My 146:

2.0 GHz, 1.0 V
2.85 GHz, 1.39 V
3.0 GHz, 1.57 V

On a DFI nF3. I sold it though, I want dual core instead. Those extra MHz doesn't make much difference in real world, it's already fast as it is. Especially in gaming, the graphics card becomes more and more important.

I see you have a P4, if you're not upgrading very often I'd suggest getting a dual core, either a X2 3800+ or a Conroe when they show up after the summer. Those extra 300 MHz you can get out of a single core just looks good on paper, and in XS forums, but they can never replace an extra core.

When I reached 3 GHz I thought "Yeah, great!.... and now?". Even then the CPU load could go up to 100 % load and get sluggish in Windows just because of some bad app or whatever. It's in those situations you realize that this brand new CPU acts just like any old Pentium 2 (only faster). The chances it would happen with a dual core are minimal. A dual core CPU gives you a different functionality that is hard to see in benchmarks. Besides, they will be more important in future games. But, if you know that you really need single core then don't listen to me.

Purely speculating: Expect prices to go down when Conroe shows up, Intels pricing are pretty much in line with todays PD, but the CPU's are cooler and much better performing. I guess AMD will cut the prices big time. And get socket AM2, the 939 won't stay for long.

[GamingGod]

so how long until am2 dual cores get down to a reasonable price like sub 200?

[stupid]

so how long until am2 dual cores get down to a reasonable price like sub 200?

It will be a few months, AM2 CPUs are debuting at the same price or slightly higher than their current S939 counterparts. I'm guessing that when Intel releases their Conroe CPUs AMD will do a minor price drop. This is pure speculation though, but I'm basing it on the fact that AND will want to maintain it's market share especially if Conroe outperforms by a large margin Athlon 64 AM2.

[Igor]

[
When I reached 3 GHz I thought "Yeah, great!.... and now?". Even then the CPU load could go up to 100 % load and get sluggish in Windows just because of some bad app or whatever. It's in those situations you realize that this brand new CPU acts just like any old Pentium 2 (only faster). The chances it would happen with a dual core are minimal. A dual core CPU gives you a different functionality that is hard to see in benchmarks. Besides, they will be more important in future games.

My thoughts exactly. In plain ol' Windows usage there seems to be no difference between 1.8GHz and 2.8GHz (my Opteron 144 ). When zipping lots of files or converting video files the computer feels just as slow... it may get it done a bit faster at 2.8GHz, but it still annoys the crap out of me. My old hyperthreading pentium 4 felt alot snappier in a similar situation.

In gaming the situation is different though. More clockspeed means more fps. The difference of course varies between games, but you will notice whether you're running a 1.8GHz cpu or a 2.8GHz cpu. Especially in games that use more cpu, such as racing simulations (my favourite genre ). Currently dual core cpus don't give that much more gaming performance, but that will change, no doubt about it.

[Mats]

Igor: What's your impression with running memory at 280 MHz, is it really worth it (in gaming)? I mean, paying €100 extra for fast RAM and getting 4 FPS more in a game is not a good deal to me. Or is it better than that?

Dunno why, but working with RAM is the most boring thing I know. tCAS this and tRAS that. I just want to buy a decent stick or two and just use them as they are. I still haven't seen any good reason for overclocking them or trying to keep them at the same speed as the CPU bus in real world usage, not XS benchmarking world. That's for K8 at least.
What do you think?

[high5]

on K8 memory has negligible influence on real world performance. there is a decent article on madshrimps.be regarding this issue:

http://www.madshrimps.be/?action=getarticle&articID=325

[QuietOC]

I still haven't seen any good reason for overclocking them or trying to keep them at the same speed as the CPU bus in real world usage, not XS benchmarking world. That's for K8 at least.
What do you think?

There isn't a direct connection between the K8 memory bus(es) and the HT connection with the motherboard chipset, and since they always share the same clock they are always synchronous.

Considering the performance of S754, S939 processors have excess memory bandwidth. Even the IGP's have trouble utilitizing very much of the extra bandwidth.

If you want better slightly better performance, just get lower latency memory (the 2-3-2 stuff is pretty cheap), but that is not worth much of a price premium.

S754 with an nVidia or ATI IGP will see a big benefit from high bandwith DDR, but it is cheaper just to get a PCIe video card.

[Mats]

There isn't a direct connection between the K8 memory bus(es) and the HT connection with the motherboard chipset, and since they always share the same clock they are always synchronous.

Yeah, but you know what I mean, when the CPU runs at x*multiplier, and the RAM runs at x. What people call running at 1:1.

[Igor]

Igor: What's your impression with running memory at 280 MHz, is it really worth it (in gaming)? I mean, paying €100 extra for fast RAM and getting 4 FPS more in a game is not a good deal to me. Or is it better than that?

Haven't done - or seen - any tests in real life gaming performance between 200MHz and 280MHz, but on my system in 3DMark06 the difference is less than 10% (4300pts vs. 4700pts). RAM timings were the same on both runs (3-3-3-7-1T). Faster RAM translates to more raw CPU power, so CPU hungry apps and games benefit more. Worth the extra 100€? Probably not, but it doesn't hurt either

Dunno why, but working with RAM is the most boring thing I know. tCAS this and tRAS that. I just want to buy a decent stick or two and just use them as they are. I still haven't seen any good reason for overclocking them or trying to keep them at the same speed as the CPU bus in real world usage, not XS benchmarking world. That's for K8 at least.
What do you think?

Yeah, it's boring stuff, that's why I mostly browse through forums like Overclockers UK, DFI street and Xtremesystems and copy some good settings. My RAM probably could run at 290-300MHz with some time consuming tweaking, but I still can't reach 1:1 with my Opty 144 at 312MHz bus, so I don't bother.

My motive behind RAM overclocking is the search for more CPU power. Racing sims are CPU hogs with all the physics calculations going on...

[high5]

i just remembered i did some Doom3 benching about 2 weeks ago. i dug out my notes, so in case anyone's interested...

system is:

Abit NF8
Venice 3000+ (s754)
2×512MB SuperTalent PC3200 CL2 (Infineon CE-5)
Sapphire X800GT 256MB

results:

CPU @ 240×10, RAM @ 200 2-3-2 1T: 69.6fps
CPU @ 240×10, RAM @ 200 2-3-2 2T: 69.0fps
CPU @ 240×10, RAM @ 200 2.5-3-2 1T: 69.5fps
CPU @ 240×10, RAM @ 200 3-3-2 1T: 69.4fps

CPU @ 240×10, RAM @ 240 3-3-2 1T: 69.9fps
CPU @ 240×10, RAM @ 240 3-3-2 2T: 69.6fps

CPU @ 200×10, RAM @ 200 2-3-2 1T: 68.3fps

i used demo1 that comes with D3, at 1024×768, 'High quality', all advances options (except AA & Vertical sync) were turned on.

Edited with more results.

[QuietOC]

2×512MB SuperTalent PC3200 CL2 (Infineon CE-5)
Sapphire X800GT 256MB

results:

CPU @ 240×10, RAM @ 200 2-3-2 : 69.6fps
CPU @ 240×10, RAM @ 240 3-3-2 : 69.9fps
CPU @ 200×10, RAM @ 200 2-3-2 : 68.3fps

It looks like you're GPU limited in Doom 3 at these settings.

Also, the two 512MB modules are forcing you to run at 2T Command Rate (unless you have single-sided modules), which impacts memory performance more than the other timings.

If you find a test that isn't GPU limted by your X800GT, try one test with just one 512MB module installed and 1T enabled. It will probably be faster than the 2 x 512MB even with only half the memory.

[high5]

It looks like you're GPU limited in Doom 3 at these settings.

we are talking real-world performance here. my point is to show that memory has little influence on overall system performance in that case on K8 architecture.

Also, the two 512MB modules are forcing you to run at 2T Command Rate (unless you have single-sided modules), which impacts memory performance more than the other timings.

FYI i'm running 1T all the time.

and modules are double-ranked. i don't know why do you think i should be running at 2T. this is Venice i'm using, not some old Clawhammer/Newcastle that would have those problems.

[QuietOC]

It looks like you're GPU limited in Doom 3 at these settings.

we are talking real-world performance here. my point is to show that memory has little influence on overall system performance in that case on K8 architecture.

No, all you showed is that the performance of Doom 3 at those settings is limited by your video card. It doesn't say anything about "overall system performance", or the effect of memory on such.

I play a wide variety of games, and they all have different system demands. Not all of them are GPU-limited by my overclocked X800GTO.

As far as the command rate... I haven't seen any reports of 1T working with multiple double-sided DIMMs on a single channel (except one S939 motherboard). If it works, that is good.

[high5]

No, all you showed is that the performance of Doom 3 at those settings is limited by your video card. It doesn't say anything about "overall system performance", or the effect of memory on such.

well, i can run timedemo at 640×480 with low detailes but that wouldn't be real-world performance anymore, would it?

it point is to show whether RAM at higher frequency will help at settings people actually use when they play game(s). that is a direct answer to a poster who asked whether extra bucks for premium RAM is worth it.

yeah, of course you could say my X800GT is bottleneck, so what? in modern games, performance will always be somewhat limited by video card b/c the better VGA you have, the higher resolution w/ more detailes you'll play your games with.

BTW the link i provided to article at madshrimps also shows that when it comes to video encoding, RAM influence is also very small there...

As far as the command rate... I haven't seen any reports of 1T working with multiple double-sided DIMMs on a single channel (except one S939 motherboard). If it works, that is good.

you haven't seen it? that isn't something rare. i ran this same RAM with my old Newcastle at 1T. though OCing sucked using 1T...

i know a guy from other forums who pushed his TCCDs up to 271MHz on NF8 with CAA2C Claw... using 1T, of course.

and since you mentioned Command rate, i did a few more demo1 runs to add those results to the mix. i'll edit my previous post.

[QuietOC]

and since you mentioned Command rate, i did a few more demo1 runs to add those results to the mix. i'll edit my previous post.

Stop, your "test" is GPU limited... it won't matter what you CPU or memory runs at--unless you severly cripple one of them. It's can't show anything other than your GPU is slow.

And, yes, streaming applications (video) are not affected by latencies, Netburst is better than K8 for that type of work.

How about some common sense. What is latency? Its the time from when something is wanted until the time you have it. You do NOT test latency by average frame rates!

What low latencies help is removing choppiness--in games that will be minimum frame rates. It is that lag you notice when textures are being fetched across the system bus. It improves the responsiveness of search algorithms, etc. And latency in general has a very noticeable affect on real world computer performance even if it doesn't show up in the typical benchmark numbers.

You may want all the lag in your "fast" system that you can get. Go aheady and get a Pentium 4--it's designed for you.

[high5]

again, yes, it's GPU limited, but it (along with madshrimps article) answers the question Mats asked.

and believe it or not, many people still think that AMD systems lose performance if you switch away from '1:1'.

BTW thanks for the kind P4 offer, but it doesn't fit my price/performance ratio.