E8500 on DDR2 Performance

[bhelms]

My head hurts, I'm sick and tired of thinking of this lol.

I'm debating between getting the AMD Phenom 9950 and the E8500.

AMD seems like a good choice since it doesn't even use FSB. So a lower RAM wouldn't matter.

However, like you said, the performance drop is not even noticeable.

Also looked at the tri-core 8600B.

Anyways, someone make the decision for me. I'm going to be playing games, and on either DDR 800 or 1066. Whats the best processor for me to get under $200?

[martini161]

e8500. amd's dont really compare performance wise.

[Qchan]

I think this is wrong.

FSB does run at 4x, but this is all during one clock cycle.

DDR runs at double density, it sends date twice per clock.

And what matters is the full clock cyle.

So that would mean, in 1 cycle the FSB is bottlenecked cause the RAM can only send 800mhz in 1 cycle.

You could look at it as being bottlenecked whatever speed the RAM is, as if in your example; once you synced the FSB to RAM (and not the real 1:1 divider, but 1333RAM:1333FSB), the performance of the system would stay the same after a that speed RAM was hit. But as we know that you can keep adding faster RAM and you will still get performance increases from the system.

However, once the FSB is saturated (which is as I posted before), the gains from going to faster RAM are minimal in most situations (a few things like un-raring that are memory intensive benefit a fair bit), so you won't really notice the difference between 800Mhz RAM and 1333Mhz RAM (and that's not even starting to touch on the CAS Latency has on the performance). Which is why I suggested to go with a decent kit of 4GB 800Mhz Cas4 RAM in the first place, you may as well put your money somewhere else it *will* be noticable.

The FSB is accessed 4 times per cycle, not 4 times per MHz. The ram is accessed twice per cycle. So, if you want to set the ratio (CPU:RAM not FSB:RAM) to 1:1, you're better off getting RAM that can function at the speed of the FSB.

You've got that backwards, it is accessed 4 times per Mhz.
[Edit: Oops, I was wrong on that one.]

Let me clarify a couple of points about FSB and RAM rated speeds.

The Mhz is it's actual clockspeed (so 333Mhz for the 1333 rated FSB). And the actual unit of rating for FSB (and RAM, or anything that does more than one thing per Hz of the cycle) is MT/s. So the FSB is actually 1333MT/s and the RAM is actually 800MT/s (for DDR2-800).

Dan

I made an account so I could respond to this particular post.
The FSB isn't measured by MHz. Its measured by clockspeed.

Clock speed is defined as Hz not MHz.
http://en.wikipedia.org/wiki/Clock_rate

And what that other guy said IS true. The FSB is accessed 4 times per clock, not per MHz.
http://en.wikipedia.org/wiki/Front_side_bus

The same goes for DDR. DDR is double density. It is accessed twice per clock. So, at the end of the clock cycle, you have 800 mhz for the DDR and 1066 for the FSB.

[Qchan]

e8500. amd's dont really compare performance wise.

You have no idea what you're talking about.

Price vs. performance, the AMD is either as good or beats Intel. I'm not talking about the fastest Intel vs. the fastest AMD. I'm talking about the price difference between the two processors with similar performance.

[martini161]

well considering mhz and hz are the same thing just that mhz is 1000000 hz, it doesnt really matter. and it seems it is you who has no idea what your talking about clock for clock across the board, intels beat amd's.

[Qchan]

well considering mhz and hz are the same thing just that mhz is 1000000 hz, it doesnt really matter. and it seems it is you who has no idea what your talking about clock for clock across the board, intels beat amd's.

Umm... 1000000 hz doesn't equal 1 mhz. 1000hz = 1mhz. Just to correct you there.

I never said that AMD beats intel clock for clock. You're misunderstanding me. I said AMD beats Intel on price vs. performance. Do you understand?

[martini161]

well considering mhz and hz are the same thing just that mhz is 1000000 hz, it doesnt really matter. and it seems it is you who has no idea what your talking about clock for clock across the board, intels beat amd's.

Umm... 1000000 hz doesn't equal 1 mhz. 1000hz = 1mhz. Just to correct you there.

I never said that AMD beats intel clock for clock. You're misunderstanding me. I said AMD beats Intel on price vs. performance. Do you understand?

umm, no. 1Mhz=1000000hz. learn the orders of magnitude my friend: http://en.wikipedia.org/wiki/Mega- and yes, i do understand, but your still wrong he aksed about performance only, not price vs. performance

[DMB2000uk]

I made an account so I could respond to this particular post.
The FSB isn't measured by MHz. Its measured by clockspeed.

Clock speed is defined as Hz not MHz.
http://en.wikipedia.org/wiki/Clock_rate

And what that other guy said IS true. The FSB is accessed 4 times per clock, not per MHz.
http://en.wikipedia.org/wiki/Front_side_bus

The same goes for DDR. DDR is double density. It is accessed twice per clock. So, at the end of the clock cycle, you have 800 mhz for the DDR and 1066 for the FSB.

Hi, Welcome to LR forums.

Was the bold and red really necessary? I'm a pretty reasonable guy and would have taken into account what you had to say even if you hadn't of highlighted it.

Ah yes, I did make a slight mistake and put Mhz when I meant to put Hz. And when you read it with that correction, you should realise that 4 times per hertz is 4 times per clock.

You are getting your terms mixed up a lot on the next bit.

DDR = Dual Data Rate. Double density = RAM chips on both side of the RAM stick. But I knew what you meant.

At the end of the clock cycle you have whatever you did that clock cycle, At the end of a second you have the effective speeds of the device. But as I said earlier they can't be counted as (M)Hz as the clock is still set at it's rate before it is dual/quad pumped. So it's known as 'MT/s' or 'effective data rate'. I would normally let this go (as it's widely misused), but seen as though we were discussing technicalities to start with I feel it's only fair to be clear in what specifically it is.

Dan

[Qchan]

well considering mhz and hz are the same thing just that mhz is 1000000 hz, it doesnt really matter. and it seems it is you who has no idea what your talking about clock for clock across the board, intels beat amd's.

Umm... 1000000 hz doesn't equal 1 mhz. 1000hz = 1mhz. Just to correct you there.

I never said that AMD beats intel clock for clock. You're misunderstanding me. I said AMD beats Intel on price vs. performance. Do you understand?

umm, no. 1Mhz=1000000hz. learn the orders of magnitude my friend: http://en.wikipedia.org/wiki/Mega- and yes, i do understand, but your still wrong he aksed about performance only, not price vs. performance

Oops! You're right. I was thinking khz.

You're also right about what he was talking about. However, if he's only talking about performance, it means he doesn't care about the price. However, if he doesn't care about the price, then why stop at the E8500? Why not get something faster?

[Qchan]

I made an account so I could respond to this particular post.
The FSB isn't measured by MHz. Its measured by clockspeed.

Clock speed is defined as Hz not MHz.
http://en.wikipedia.org/wiki/Clock_rate

And what that other guy said IS true. The FSB is accessed 4 times per clock, not per MHz.
http://en.wikipedia.org/wiki/Front_side_bus

The same goes for DDR. DDR is double density. It is accessed twice per clock. So, at the end of the clock cycle, you have 800 mhz for the DDR and 1066 for the FSB.

Hi, Welcome to LR forums.

Was the bold and red really necessary? I'm a pretty reasonable guy and would have taken into account what you had to say even if you hadn't of highlighted it.

Ah yes, I did make a slight mistake and put Mhz when I meant to put Hz. (1)And when you read it with that correction, you should realise that 4 times per hertz is 4 times per clock.

You are getting your terms mixed up a lot on the next bit.

(2)DDR = Dual Data Rate. Double density = RAM chips on both side of the RAM stick. But I knew what you meant.

(3)At the end of the clock cycle you have whatever you did that clock cycle, At the end of a second you have the effective speeds of the device. But as I said earlier they can't be counted as (M)Hz as the clock is still set
at it's rate before it is dual/quad pumped. So it's known as 'MT/s' or 'effective data rate'. I would normally let this go (as it's widely misused), but seen as though we were discussing technicalities to start with I feel it's only fair to be clear in what specifically it is.

Dan

I apologize for the bold... But that is how I talk on message forums. Think of it as my global signature. My font is always red and it's always in bold.

I numbered and underlined the points I will respond to.

1) Yes. That is what I said in the previous post.

2) Sorry. That was a slip of the lip. I meant double rate and not density. However, I'm glad you knew what I meant

3) Sorry for cutting slices in your paragraph, but I want to iterate that you and I are both talking about Hz and MHz. Not about MB/s. So, if we're talking about Hz and MHz, then we're talking about per clock cycle and not per second. So, talking about MT/s would be completely irrelevant. I did not dispute anything in relation to MT/s.

[DMB2000uk]

Number 1 was me just making sure we were on the same page after my correction

Yes, we are talking about Hz etc. A Hz is (roughly) defined as a clock cycle per second. So that is why I brought seconds into it, and its why after only one cycle it won't have done very much at all.

MT/s isn't irrelevant, as that is what everyone commonly miss-calls the MHz of RAM/FSB. Its actual clock rate/frequency rating is still the base frequency (MHz/Hz), but after it's been dual or quad pumped it strictly should only be called either it's 'effective clock rate', or named in terms of how many transfers (aka cycles) it can do in a second; that is: MT/s (MegaTransfers per second (or in the case of Hz just Transfers per second)).

This is all far to complicated to be discussing this late (I'm in the UK and it's gone 11pm), and I think we are mostly agreeing with each other, but are getting hung up on a couple of little details so am going to call it a night. We've kind of gone off track from the OP's question anyway, so we should probably wrap it up soon


Dan

[bhelms]

Well yup, ya'll lost me. Doesn't bother me though, I don't care to try and understand it that deeply. It was, however, interesting to read and follow to the best of my ability.

I still haven't decided on a processor for under $200

I was like... yes the e8500. Wait.. what about the tri-core from amd for $50 cheaper and it is barely behind the e8500... man that extra core would be nice in the future when more multi-threaded games come around, since I will have the processor for around 2 years. Oh then how about a quad? I'd give anything in the world for me to be able to make a decision. This is more of a rant about myself and my inability to make decisions more than anything.... lol

[martini161]

actually believe it or not intel duals beat AMD tri cores right now in terms of gaming. so the e8500 is your best bet.

[DMB2000uk]

And in convient coincidences, bit-tech has done a review of 1066Mhz RAM vs 800Mhz.
The real world benchmarks start here:
http://www.bit-tech.net/hardware/2008/1 ... -1066mhz/4

And you can see that there is very little difference between the two kits.

So 4GB 800Mhz CAS4, and an E8500 are your recommended choices sir.

Dan

[bhelms]

A Q6600 wouldn't be a better choice for a long term use? Since I don't update but every like... 2 years? I'm just nervous about new stuff on the horizon doing better with more cores. And the performance increase of an E8500 over the Q6600 is marginal right now anyways isn't it? And are you sure getting CL 4 is more important than 1066?

I believe Supreme Commander runs best with 4 cores.

Thank you two very much for your patience with me

[DMB2000uk]

If you are happy spending the extra money for 1066Mhz RAM, then feel free to do so. Just don't expect any major performance jumps over the 800Mhz kits.

If you want a quad, get this: http://www.newegg.com/Product/Product.a ... 6819115041

Unless you are happy with overclocking a Q6600 lots to keep up it's performance over the next couple of years.

Dan

[bhelms]

And the E8500 wouldn't need to be overclocked to keep up performance? Also, what I meant was, 1066 CL5 or 800 CL 4. They are about the same price.

http://www.newegg.com/Product/Product.a ... 6820231209

http://www.newegg.com/Product/Product.a ... 6820231166

[bhelms]

Made my deicsion, thanks Dan and Martini161! E8500 on DDR2 1066 And a brand new 700w modular PSU.

[DMB2000uk]

Made my deicsion, thanks Dan and Martini161! E8500 on DDR2 1066 And a brand new 700w modular PSU.

For anyone else reading: The 1066 RAM was only $10 more, so it was a 'you may as well get it' type of thing.

Dan

[Racerboyf1]

Hey, if you are going to spend $300.00 on a video card why not look at this one? http://www.newegg.com/Product/Product.a ... 6814161247