e8600 Slower than e8500???

[Evilsizer]

it isn't a new architecture like nehalem. I'm pretty sure E0 is just a test to see how far the cpus can overclock when they loosen the cycles for L1 cache. The 2~3% decrease in performance clock per clock sounds about right. What is MOST important, is to find out if the cpu is fasterin the end, E8600 highest OC vs. older stepping highest OC.

right, which now makes sense... as this would/may allow intel to release a faster clocked Neha out the door then the first planned. though im still dougting intel will release one faster then the 3 they have alread talked about releasing.

 

[Stilletto]

it isn't a new architecture like nehalem. I'm pretty sure E0 is just a test to see how far the cpus can overclock when they loosen the cycles for L1 cache. The 2~3% decrease in performance clock per clock sounds about right. What is MOST important, is to find out if the cpu is fasterin the end, E8600 highest OC vs. older stepping highest OC.

Well, my last e8500 topped out at 4.02 stable with acceptable voltage. I am running this 8600 at 4.42, and of course, it smokes on the number crunching at that speed, as it damn well better..
HOWEVER, it is giving me less graphical performance(slightly) at 4.42ghz than the e8500 was at 4.02
That is my main beef, and the purpose of this long and arduous quest so far..

 

[wingman99]

it isn't a new architecture like nehalem. I'm pretty sure E0 is just a test to see how far the cpus can overclock when they loosen the cycles for L1 cache. The 2~3% decrease in performance clock per clock sounds about right. What is MOST important, is to find out if the cpu is fasterin the end, E8600 highest OC vs. older stepping highest OC.

No i agreed it's not a new architecture, but i cant find out what all the power saving features are that they have in cooperated in the Eo step, they seem to be hidden from the public, they could easily change the transistors without changing the architecture.

They also could shorten Cache prefetch to save energy.

 

[muddocktor]

OK, enough is enough already!!!

wingman99, you keep dragging this out and threadjacking and I won't stand for this any more. If you don't have anything to add about the purpose of this thread (which is discussing whether the E8600 is slower clock for clock than the E8500, BTW), then you had best quit posting in it. I won't tolerate any more threadjacking!

Evilsizer, I wish you had hit the report post button earlier instead of answering back to wingman99. This thread has some nice info in it but it's hard to read due to the bickering back and forth.

I'll try to clean the crap out now, so this thread will be closed for a little while.

Stay on topic, people!

 

[muddocktor]

OK, I cleaned the bickering out. This thread is back open.


Keep it civil and keep it on topic!!!

 

[saxile]

So far, the cache speed/latency according to my testing is looking to be slower indeed.....

Well, my last e8500 topped out at 4.02 stable with acceptable voltage. I am running this 8600 at 4.42, and of course, it smokes on the number crunching at that speed, as it damn well better..
HOWEVER, it is giving me less graphical performance(slightly) at 4.42ghz than the e8500 was at 4.02
That is my main beef, and the purpose of this long and arduous quest so far..

I think the cache speed is what is lowering the graphic performance, that would be my best guess. I'm not hardcore like most of the other's here, but I do have a good bit a knowledge, of course I could be wrong and that isn't it, but I truthfully don't think mhz is all to this battle, look at AMD's performence on their AMD 64 series compaired to the Intel chips of that time, a 2.0ghz AMD ate a 3+ghz Intel for dinner. Now if we could only figure out a way to overclock the cache on these chips and do it safely look at the speed increases we would get.

I'm not knocking on Intel on this one as you can tell I'm Q6600 in my rig currently, but when I was running my phenom 9850BE at 2.5ghz with lower ram settings and same video settings I was getting about 1k more marks in 3dmark06 than with this Q6600 at 3.85

 

[jason4207]

I'm not knocking on Intel on this one as you can tell I'm Q6600 in my rig currently, but when I was running my phenom 9850BE at 2.5ghz with lower ram settings and same video settings I was getting about 1k more marks in 3dmark06 than with this Q6600 at 3.85

Can you show proof of this? I find it hard to believe.

 

[Evilsizer]

ok now that i remember what the name of the program is... i did recall a while back there was a L1/L2 benching section in some program. turns out it was sissoft's sandra, why would this matter. simply at the same fsb/cpu speed, run the "cache and memory" bench. it will provide bandwidth data like ram bandwidth. that way if cpu clocks/fsb are equal we would then be able to figure out for sure if the cycle times have been increased/loosened. i do recall another "tool" that would report cycle times for L1/L2. i think it was by H. oda, i check out his site and does have one CPUID type program. its from 2004, i tried to run it but it forced my computer to restart. thank god for FF caching we pages and this post! i did how ever lose my numbers from sandra.

 

[Stilletto]

ok now that i remember what the name of the program is... i did recall a while back there was a L1/L2 benching section in some program. turns out it was sissoft's sandra, why would this matter. simply at the same fsb/cpu speed, run the "cache and memory" bench. it will provide bandwidth data like ram bandwidth. that way if cpu clocks/fsb are equal we would then be able to figure out for sure if the cycle times have been increased/loosened. i do recall another "tool" that would report cycle times for L1/L2. i think it was by H. oda, i check out his site and does have one CPUID type program. its from 2004, i tried to run it but it forced my computer to restart. thank god for FF caching we pages and this post! i did how ever lose my numbers from sandra.

This is some of the stuff I have been working on. I have done some testing, this time against an e8400, with the help of wingman99. I am going to post some results shortly.

 

[Stilletto]

Here are the results of the comparison. The first was submitted for comparison by wingman99 with his e8400 clocked at 3ghz. The second is my run with the e8600 clocked at 3ghz, same 333x9.5 settings (the bios did overclock by 5mhz, but I am considering that inconsequencial). These results are encouraging, (except for the sse3 FPU) and what we are to expect out of a good stepping. However, the next post will contain concerning results....

 

[Evilsizer]

i didnt remember pcmark, thought they stopped that program. what you should do then since you have it is run the L1 and L2 benches in there. as to me that bench doesnt show what we are after.

 

[Stilletto]

Now the L1 and L2 cache is what concerns me, and I may see a problem here.
Here is a shot of what I got with the same 3.0ghz settings...keep in mind that with the e8400 and e8500, the results are consistently 16cycles and 15cycles, L1 and L2 respectively. With the e8600, this is one of the lowest results I was able to get:

 

[Stilletto]

what you have posted looks like info from 3dmark that matches up. i didnt remember pcmark thought they stopped that program. what you should do then since you have it is run the L1 and L2 benches in there. as to me that bench doesnt show what we are after.

Sry, I wuz readin' yer mind there

 

[Evilsizer]

Sry, I wuz readin' yer mind there

ok if that was the slowest with the E8600, what was the slowest with the E8400?

 

[wingman99]

Here are the results of the comparison. The first was submitted for comparison by wingman99 with his e8400 clocked at 3ghz. The second is my run with the e8600 clocked at 3ghz, same 333x9.5 settings (the bios did overclock by 5mhz, but I am considering that inconsequencial). These results are encouraging, (except for the sse3 FPU) and what we are to expect out of a good stepping. However, the next post will contain concerning results....

Holy smokes you should of told me you were going to use that screen shot, i forgot I was running a game server 50% cpu usage. Here is one stock E8400 3.0Ghz clean Edit: i changed my memory to stock 5-5-5-15 are you doing the test with speed step ON or OFF, mine is on. Edit: I forgot i had turbo on,I changed that.

 

[Stilletto]

ok if that was the slowest with the E8600, what was the slowest with the E8400?

I could not get much better with my e8500...15/16 L1 and 15L2.

Holy smokes you should of told me you were going to use that screen shot, i forgot I was running a game server 50% cpu usage. Here is one stock 3.0Ghz clean Edit: i changed my memory to stock 5-5-5-15 are you doing the test with speed step ON or OFF, mine is on

Mine is on as well. This shows that there is a common performance curve between the two, all except for the cache latency. However, when clocked at 4.42ghz, the 8600 is starting to catch up....I am getting 16cyclesL1 and 15cyclesL2 at those speeds

What really bakes my noodle is that in Vista, the latency is ridiculous. 25cycles for L1 and 18cycles for L2. This may explain some of the oddness occuring under Vista. THis does NOT explain, however, the lower graphics performance this chip is getting vs the previous steppings. This is starting to hurt my head

EDIT: I think I will concentrate on running some benchies in some games, like many posters have requested. The two I think that are most likely candidates would be COD4 and Crysis. Those results will hopefully provide some hard evidence that I am desperately seeking

 

[wingman99]

I could not get much better with my e8500...15/16 L1 and 15L2.


Mine is on as well. This shows that there is a common performance curve between the two, all except for the cache latency. However, when clocked at 4.42ghz, the 8600 is starting to catch up....I am getting 16cyclesL1 and 15cyclesL2 at those speeds

What really bakes my noodle is that in Vista, the latency is ridiculous. 25cycles for L1 and 18cycles for L2. This may explain some of the oddness occuring under Vista. THis does NOT explain, however, the lower graphics performance this chip is getting vs the previous steppings. This is starting to hurt my head

If you want we can do some SiSoftware Sandra Lite XII.SP2c bench test's tomoro night, so you don't have switch out your Cpu.

 

[Stilletto]

If you want we can do some SiSoftware Sandra Lite XII.SP2c bench test's tomoro night, so you don't have switch out your Cpu.

That's a plan. I will also try to get some FPS performance in games for some real world data. Before this is over, I will undoubtedly need a case of Artic Silver

 

[wingman99]

That's a plan. I will also try to get some FPS performance in games for some real world data. Before this is over, I will undoubtedly need a case of Artic Silver

Yes sounds good, i still think the problem is the cache, i was worried about the EO stepping before you posted, that's why i sent you the link with pc wiz 2008 easy to use, i was hoping you were doing some testing and you said you were I'm sure you were trying everything, I also wanted to see some things for my self that's why i redid that screen shot. this is the kind of stuff i love to do, see what really going on with intel and there changes.

 

[froggyluv]

EDIT: I think I will concentrate on running some benchies in some games, like many posters have requested. The two I think that are most likely candidates would be COD4 and Crysis. Those results will hopefully provide some hard evidence that I am desperately seeking

Yes,please do. I've been scouring the internet seekng out the elusive E0 stepping 8400/8500, hearing of it's overclocking prowess but as I am primarily a gamer, your thread has me worried that perhaps E0 ain't the way to go. It may be a benchmark king, but like my grandpappy always said " Bench for show-Frame-rates for dough".