E8400 is amazing!

Conroy · Post by **Conroy** » Mon Jan 28, 2008 8:23 am

djkest wrote:Anyway, interesting to see your results so close to mine! I just crashed after 15 min of prime95... testing my 3.8GHz numbers. I may have to mount a fan on the ninja, it's starting to get a little toasty.

Yeah, I thought it was pretty interesting that the numbers were so similar. They were even closer when I was initially using the other prime95 torture tests to load the system.

I tried out CrystalCPUID 64, but I was unable to use it to set anything. On the screen to set a single voltage and multiplier ("Intel Speedstep"), it ignored whatever I set , and on the RMClock-like screen, it didn't display any CPU info to set.

Beyonder · Post by **Beyonder** » Mon Jan 28, 2008 11:54 am

smilingcrow wrote:
Beyonder wrote:Looking at this thread, I can't help but wonder: how feasible would it be to run an E8400 system with a decent video card on a picopsu? I'd love to run a minimalist system (say, 3850?) with an E8400, one of those Western Digital green drives, and a good video card for my primary system...
No chance of that working as the PicoPSU 120 is only rated for 84W continuous output on the +12V rail.

Could you elaborate on this a bit more? Is the issue that the CPU and video card share the +12V? Is it out of the question to run an E8400 on a picopsu, or is it out of the question to run the E8400 _and_ an 3850/7600GS/7600GT/etc. ?

One thing I was figuring is the picopsu is going to be more efficient at the wattages (70-100W) being displayed in this thread. That said, i realize the wattage figures here haven't been loading the video card either.

smilingcrow · Post by **smilingcrow** » Mon Jan 28, 2008 12:10 pm

Most of the power in modern PCs is drawn from the 12V rail which is why the Picoâ€™s headline rating of 120W can be misleading. 84W DC is enough to run an E8400 as the data posted in this thread shows. 98W AC at 3GHz (undervolted) with a GPU that consumes at least 15W AC at idle suggests that if you used an IGP you would use ~83W AC or ~67W DC at load. That leaves roughly 17W DC for a VGA card which isnâ€™t very much. In practice the CPU might not be so heavily loaded whilst gaming compared to running Prime95 so that would give a bit more headroom for the VGA card. Thereâ€™s no way a HD 3850 would safely fit into the power envelope though.

WR304 · Post by **WR304** » Mon Jan 28, 2008 5:41 pm

smilingcrow wrote:Iâ€™ve seen a lot of reports suggesting that certain Wolfdale CPUs have problems reporting accurate DTS temps for lower values and sometimes the temperature reported are static at say 43C at idle and only increases when under load and the temp rises above this value. Something to keep in mind if anyone has unusually high idle temps with these chips.

That's what my one seems to be doing.

With a Gigabyte GA-X38-DS4 motherboard (F2 Bios), and E8500 CPU, idle temperatures are exactly the same at stock speeds and vcore (9.5x333, 3.16ghz and 1.225v) as when it's running overclocked at 9.5x422, 4.009ghz and 1.344vcore. They jump higher when the CPU is loaded though.

In the BIOS, CPU temperature appears as 30c approx, whilst Coretemp 0.96.1 displays the core temperatures as 45/42c.

In a Gigabyte GA-X38-DS4 motherboard, getting the E8500 CPU to 4ghz is literally just a case of typing in a higher FSB number and pushing the CPU voltage up a few notches.

The CPU at 4ghz can boot into Windows at anything over 1.3 vcore. Less than 1.344 vcore isn't stable when stress testing with Prime95 or Orthos though.

Core 2 Duo E8500 @4ghz

The extra speed over a 3.2ghz Core 2 Duo is quite impressive and noticeable when running software.

The CPU socket layout on the GA-X38-DS4 motherboard isn't perfect. It's high up and further away from the exhaust case fan. It seems to be working ok but it looks like another fan to cool the Scythe Ninja may be necessary.

E8500 and Gigabyte GA-X38-DS4 motherboard in Antec Sonata case

mcoleg · Post by **mcoleg** » Mon Jan 28, 2008 7:01 pm

seems mine is doing something similar...

i am keeping it though - it does 4ghz at stock vcore - so i'm good.

djkest · Post by **djkest** » Tue Jan 29, 2008 6:15 am

mcoleg wrote:seems mine is doing something similar...

i am keeping it though - it does 4ghz at stock vcore - so i'm good.

No way. It wouldn't run a stress test like Prime95 and pass for 24 hours on stock Vcore. Just booting into windoze is not enough. Is this an E8500 or E8400?

djkest · Post by **djkest** » Tue Jan 29, 2008 6:18 am

WR304 wrote: The CPU at 4ghz can boot into Windows at anything over 1.3 vcore. Less than 1.344 vcore isn't stable when stress testing with Prime95 or Orthos though.

1.344 is your Vdroop. Your Vcore is actually higher in the bios. 1.365 maybe? I'm running 3.8Ghz at 1.30Vdroop.

Your CPU actually runs at a lower voltage than the bios is set for. It droops. So your measurements in windows will be lower than that of the bios setting.

mcoleg · Post by **mcoleg** » Tue Jan 29, 2008 3:02 pm

djkest wrote:
mcoleg wrote:seems mine is doing something similar...

i am keeping it though - it does 4ghz at stock vcore - so i'm good.
No way. It wouldn't run a stress test like Prime95 and pass for 24 hours on stock Vcore. Just booting into windoze is not enough. Is this an E8500 or E8400?

ah, right... there's no pic upload on this forum. i'll upload some at night my time, have some orthos screenies.

WR304 · Post by **WR304** » Tue Jan 29, 2008 3:44 pm

Despite the unreliable CPU temperature readings the PC has worked fine all day today.

I'm still considering putting a fan on the Scthe Ninja but during normal use it seems ok for now. The Scythe Ninja is cool to the touch, and all the other recorded temperatures (motherboard, harddrives, graphics card etc) are in the 30-36c range. That's with a Scythe S-Flex SFF21D 800rpm exhaust fan, a Scythe Slipstream 500rpm fan on the graphics card cooler and a room temperature of 18c approx.

@djkest: The CPU vcore setting selected in the BIOS appears as 1.34375v.

"Loadline Calibration" is enabled in the BIOS also.

In CPU-Z 1.43 vcore appears as 1.344v when the PC is idling.

When running a stress test such as Prime95 it drops to a reported 1.328v.

I'm not sure how accurate CPU-Z actually is for measuring vcore but that's what is being reported by it.

There are some other people's experiences of E8x00 temperature readings in this thread:

http://www.xtremesystems.org/forums/sho ... 999&page=8

mcoleg · Post by **mcoleg** » Tue Jan 29, 2008 9:00 pm

here's the pix:

not the around-the-clock, of course. the whole thing was getting a bit toasty and the bloody fan was at above 2000rpm. i can go around wearing earplugs only for so long...

i am getting a dif. h/s for it, i might retest then.

smilingcrow · Post by **smilingcrow** » Wed Jan 30, 2008 12:18 pm

mcoleg wrote:not the around-the-clock, of course. the whole thing was getting a bit toasty and the bloody fan was at above 2000rpm. i can go around wearing earplugs only for so long...

i am getting a dif. h/s for it, i might retest then.

Sounds like you have a particularly good chip there. What were the DTS readings whilst running Orthos? You arenâ€™t showing them in Speedfan and you might well need to use CoreTemp to calibrate Speedfan as it seems less accurate at determining the Tjunction temp.

mcoleg · Post by **mcoleg** » Wed Jan 30, 2008 12:55 pm

at about 67-70 at the time. i just got tired of f8-ing every time i restart so i gave up on the coretemp for now. it doesn't read the current flock of wolfdales correctly anyways so i figure - why even bother.

coretemp reads tjmax at 105c and vid for the chip is 1.0500.

i had to level the vdroop on 5pe-vm for this overclock to happen though.

on the other hand, it was doing 3.8 at 1.14v vcore in windows (again, it was set on AUTO in BIOS) with vdroop so i might remove the mod. i figure 3.6 is about as fast as i want right now and that would allow to cut on some noise.

smilingcrow · Post by **smilingcrow** » Thu Jan 31, 2008 2:58 am

mcoleg wrote:at about 67-70 at the time. i just got tired of f8-ing every time i restart so i gave up on the coretemp for now. it doesn't read the current flock of wolfdales correctly anyways so i figure - why even bother.

Why did you need to F8 to get CoreTemp to work!
Does Speedfan not read the DTS in Wolfdale?

smilingcrow · Post by **smilingcrow** » Thu Jan 31, 2008 3:02 am

Xbitlabs has an article showing the power consumption of Wolfdale.

The difference at load between 1.6GHz 512MB cache (65nm) versus 3.16GHz 6MB cache (45nm) cache is only 16.6W. Over twice the performance for under 17W. Wow.

mcoleg · Post by **mcoleg** » Thu Jan 31, 2008 6:54 am

"Why did you need to F8 to get CoreTemp to work!" - vista 64-bit.
"Does Speedfan not read the DTS in Wolfdale?" - no, it does not appear so.

wim · Post by **wim** » Thu Jan 31, 2008 4:03 pm

E8400 and E8500 is same??

smilingcrow · Post by **smilingcrow** » Thu Jan 31, 2008 4:09 pm

wim wrote:E8400 and E8500 is same??

Thereâ€™s only a ~5% difference in speed which can easily be offset by a lower VID. Donâ€™t forget the chips are speed binned so the top of the range can be that little bit better.

Matija · Post by **Matija** » Fri Feb 01, 2008 9:29 am

So, anyone want to guess what the issue with these CPUs is?

Faulty CPUs not reporting temps properly, or just software issues?

I wanted an E8400, but now I'm not so sure...

Conroy · Post by **Conroy** » Fri Feb 01, 2008 10:20 am

My guess is a combination of bios support and faulty sensors

I'd guess that the stuck temps are a bios problem, since most of the reports I've seen seem to be from similar motherboards. I'd also guess that the incorrect readings at low temps could be a sensor issue..

1) I'm not sure, but I think that kind of problem is not uncommon for the first batches of new chips.
2) Weren't the sensors originally intended for use at high temps mostly, to throttle the cpu down in case of overheating?
3) It looks like mobile penryns have been delayed due to defective thermal sensors, e.g. http://www.engadget.com/2008/01/15/inte ... yn-delays/

Of course, this is just what i've gathered reading lots of posts and googling, so I could be way off.

WR304 · Post by **WR304** » Fri Feb 01, 2008 4:51 pm

You can see how much variation there is between retail chips from this screenshot of my second E8500. This one is fitted in a GA-P35-DS4 motherboard.

It's not as good as mcoleg's E8500 but appears better than the first E8500 I bought (pictured above in the GA-X38-DS4 motherboard).

This chip has a lower VID of 1.0875v, compared to 1.1125v for the first E8500.

It works at 4ghz with a lower vcore of 1.27500v (BIOS) compared to 1.34375v (BIOS) for the first E8500.

In the motherboard BIOS this chip has a default vcore of 1.22000v, whilst the first one is higher at 1.22500v.

In theory that should make the second one cooler running, as it requires less voltage. Neither is reading temperatures right so it's difficult to say.

Both processors work ok though.

Second E8500 CPU at 4ghz in GA-P35-DS4 motherboard

cmthomson · Post by **cmthomson** » Sun Feb 03, 2008 10:33 am

This kind of improvement in operating voltage is pretty common at the start of the product cycle of Intel CPUs.

The first batch is shipped when the fab achieves sufficient yield to be profitable (and the design has been debugged). The engineers then continue to tweak the fab processes to maximize yield and repeatability across fabs. This almost always means that CPUs built a month or two after the initial batch overclock better at lower voltages.

This was really pronounced when the Conroe (E6600) was introduced. By the fourth week of production, the difference was quite large.

To take advantage of this, you need to know about date codes. These are printed on the serial number sticker on the box, as well as on the IHS of the CPU, but are not readable by software. The date code is a two-digit year, followed by a two-digit week number. For example, a chip made this week would have a date code of 0806.

If you buy online, it's hit-and-miss what you'll get. If possible, you should get your CPU at a brick-and-mortar, where you can inspect the date code. If that's not an option, you may be able to keep RMAing CPUs until you get what you want (a hassle, for sure).

dragmor · Post by **dragmor** » Sun Feb 03, 2008 5:49 pm

cmthomson wrote:This kind of improvement in operating voltage is pretty common at the start of the product cycle of Intel CPUs.

I know this is true with AMD and their APM method, but I thought Intel used a Copy Exact method for its fabs i.e. tweaks are made in the R&D fab and then pushed out to the production fabs i.e. the production fabs don't tweak and the pushes are infrequent.

C2D's have a lot of variation even in the same batch.

smilingcrow · Post by **smilingcrow** » Mon Feb 04, 2008 3:03 am

mcoleg wrote:on the other hand, it was doing 3.8 at 1.14v vcore in windows (again, it was set on AUTO in BIOS) with vdroop so i might remove the mod. i figure 3.6 is about as fast as i want right now and that would allow to cut on some noise.

With some motherboards when you set the VCore to Auto in the BIOS it actually over-volts. This seems to be true for certain Gigabyte and Asus boards; what board are you using?

mcoleg · Post by **mcoleg** » Mon Feb 04, 2008 6:35 pm

p5e-vm, asus

the board tries to buffer the voltages as fsb is raised but vcore droops pretty bad without mods on it.

i am not particularly sure what is the default vcore on this chip. vid is 1.0500v but that's a very loose indicator from my experience.

so, if it overvolts a bit, no biggie.

there's some pretty complicated voltage regulation build into p5e-vm. i saw vcore dropping to 1.000v a couple of times when idling; it doesn't always do it though. it also has "damper" option in bios, to prevent vdroop, i understand; however, that option doesn't seem to work so a mod is needed if one plans to overclock with this board. hopefully, future bios releases will take full advantage of the hardware.

cmthomson · Post by **cmthomson** » Tue Feb 05, 2008 5:38 pm

dragmor wrote:
cmthomson wrote:This kind of improvement in operating voltage is pretty common at the start of the product cycle of Intel CPUs.
I know this is true with AMD and their APM method, but I thought Intel used a Copy Exact method for its fabs i.e. tweaks are made in the R&D fab and then pushed out to the production fabs i.e. the production fabs don't tweak and the pushes are infrequent.

What you say is not inaccurate: the Intel production fab engineers don't do tweaks. However, at least early in the cycle, the R&D engineers update the specs to improve yields. Pushes are indeed very infrequent after a couple of months, but very early in the cycle they are more frequent.

C2D's have a lot of variation even in the same batch.

Of course. But that's true of nearly all CMOS-based chips. Variations of up to 40% are not uncommon. This is why Intel (and nearly everybody else) has four or so speed bins for each design: it's not because the designs are different, it's because the specific chips have so much variation.

smilingcrow · Post by **smilingcrow** » Wed Feb 06, 2008 1:51 am

mcoleg wrote:so, if it overvolts a bit, no biggie.

Well it probably explains why you appeared to be able to run at 4GHz at stock voltage which surprised people. I guess it wasnâ€™t running at stock voltage if you had used Auto in the BIOS.
I changed the setting in my Gigabyte board from Manual to Auto and the power consumption at load increased by 25W @ 2.43GHz so it can be a not inconsiderable difference. Did you measure power consumption at 4GHz? Or try running it at 3.6GHz with both Auto and Manual settings and compare power draw.

mcoleg · Post by **mcoleg** » Wed Feb 06, 2008 7:45 pm

what is exactly the number i should use for vcore then? i'd give it a try.

smilingcrow · Post by **smilingcrow** » Thu Feb 07, 2008 2:41 am

mcoleg wrote:what is exactly the number i should use for vcore then? i'd give it a try.

I donâ€™t know what Asus use but in recent Gigabyte boards you would set the voltage to manual and not a numeric value. Manual uses the default VCore, Auto will over-volt and the other option is to set a specific fixed value.

mcoleg · Post by **mcoleg** » Thu Feb 07, 2008 7:39 pm

nope, not here. it's either auto or numeric value.

i have thought, ever since these cpus were pre-viewed, that they'd need about 1.2-1.25 on load.

srsly though, what is considered a default vcore for wolfdales? anyone?

Conroy · Post by **Conroy** » Thu Feb 07, 2008 8:09 pm

my chip defaulted to 1.225v, with cpu-z reporting that it's actually getting 1.216v

I think each chip may have a different default though