It doesn't, instead it has additional +3x turbo multiplier, because TDP is low. Some 60W CPUs also had +4x turbo.

its also got the best IMC of the series so it should be able to push memory better than other models.

All X56xx are essentially identical, same core stepping and everything. 95W TDP in specs doesn't mean some magical higher binned silicon. You get lucky or you don't.

The Serpent Rider wrote on 2023-05-02, 04:52:

It doesn't, instead it has additional +3x turbo multiplier, because TDP is low. Some 60W CPUs also had +4x turbo.

its also got the best IMC of the series so it should be able to push memory better than other models.

All X56xx are essentially identical, same core stepping and everything. 95W TDP in specs doesn't mean some magical higher binned silicon. You get lucky or you don't.

Intel did actually bin these. The exact binning process probably varied over the year or two these were fabricated.

likely top bin: X5670, X5675, X5680, X5690, W3680 and W3690.
likely second sorting: 980X, 990X, E5649, X5660, W3670
likely third bin: X5650, 970, 980, E5645
bottom bin: The really low frequency chips.

(The 32nm 4-cores are probably from all the bins depending on spec. I don't think intel had that many chips with failed cores.)

I have tested about 40 Gulftown/Westmare CPUs over the last 13 years and while that isn't enough to get a good picture it is enough to be able to tell that Intel did some binning. 😀

In my experience the X5650 often needs alot of voltage for 4.4 GHz while for example 100% of the twenty or so X5670 and X5675 CPUs I have tried can do 4.4GHz at a low, safe and easy to cool voltage.

The Serpent Rider wrote on 2023-05-02, 04:52:

It doesn't, instead it has additional +3x turbo multiplier, because TDP is low. Some 60W CPUs also had +4x turbo.

its also got the best IMC of the series so it should be able to push memory better than other models.

All X56xx are essentially identical, same core stepping and everything. 95W TDP in specs doesn't mean some magical higher binned silicon. You get lucky or you don't.

The X5675 was one of the last steppings produced and has refinements to the IMC the earlier models didn't, this is a known fact not something I just pulled out of my arse, its also got nothing to do with the TDP which is simply lower from using better silicon and a refined fabbing process.

Its well known that Intel makes improvements to the IMC and CPU itself with each stepping and that they do tend to do binning particularly for their server CPUs which have stricter requirements than their desktop cousins. But you'll likely argue the point when there are a vast number of OC benches showing that the final stepping Westmere CPUs have the improved IMC over the earlier models.

TrashPanda wrote on 2023-05-01, 22:14:

XP 32bit supports more than 2 cores, IIRC its 32 logical cores with a max of 2 physical CPUs for XP Pro, XP64 its 64 logical cores and 2 CPUs.

So im not sure why you wouldn't run XP on a 6 core CPU, its more than up to the task, that said Win7 is better overall if you want DX10.

ah yes, i'm conflating cores and processors! i think after sp2 everything is recognised by xp

though i'd still prefer w7 if only for the 64 bits, better driver support and so on (in case we want a more modern graphics card for instance)

a better v8 engine analogy would be using only 32bits of a 64 it cpu is using half the engine

still, i understand why some folk love using latest possible tech with XP and making XP fly along

This is my workstation at work (where I sadly spend most of my waking hours during the weeks).

.

Last year I used this system to bin most of my 6-core Gulftown/Westmare CPUs. I find it important to have productive things to do at work...

Out of the X5670s and X5675s this one was the worst so I have used it for memory testing and for testing degradation. My best X5670 is one I bought new (but not at full retail price) back in 2011 for my SR-2 system.

Even a crappy X5675 is likely still a nice chip 😀

TrashPanda wrote:

The X5675 was one of the last steppings produced and has refinements to the IMC the earlier models didn't

They all are one stepping - B1.

Skyscraper wrote:

Intel did actually bin these.

There are counter arguments for that too. Like existence of i7 920 D0, which basically was go to CPU for really hardcore OC, because it was cheap, numerous and results were comparable to i7 975X with identical stepping. Results on low-end models obviously would vary more wildly though.

The Serpent Rider wrote on 2023-05-02, 09:04:

Skyscraper wrote:

Intel did actually bin these.

There are counter arguments for that too. Like existence of i7 920 D0, which basically was go to CPU for really hardcore OC, because it was cheap, numerous and results were comparable to i7 975X.

Well that's a Bloomfield so a different kind of fruit!

The early C0/C1 Bloomfields were binned for sure as most i7 965 will do 4.2GHz without too much trouble but many C0 i7 920 run into a brick wall at around 4GHz.

I bought my first i7 920 at launch day, a pretty poor C0 that managed 4.0GHz but totally brick walled at ~4.1GHz regardless of voltage and cooling. As soon as the D0-stepping came out in 2009 I bought a new i7 920, it was not much better than my C0 and the IMC was even worse. During 2009-2010 I did build a few s1366 systems for friends/family/work and I was always on the lookout for a better clocking CPU compared to what I had but I didn't have much luck. I even tried a W3520, the Xeon flavor of the i7 920 but never any of the expensive higher clocked models. When I upgraded and built my EVGA SR-2 system late 2010 I still had not found a Bloomfield chip that would do more than 4.2 GHz at air cooled "24/7 settings".

These days when the CPUs are a couple of euro each I have bought some higher clocked ones and gotten even more with motherboards and I have now probably tested about 30 Bloomfield CPUs in total. I can not say that I have seen any noticeable differences in how D0 i7 920, 930, 950 and 960 clock on average but the sample size is still a bit small. I bought 8x W3565 and they (at least the 4 I have tested so far) seem to clock better doing 4.4 - 4.5GHz with decent air cooling. The W3565 is an OEM version of the W3570 but with lower QPI-bus speed. Because all of my W3565 CPUs are from the same batch it's hard to draw any broader conclusions other than that this very batch of W3565 CPUs is decent. 😁

I'm pretty confident that Intel binned Bloomfield D0 silicon but at some point almost every chip came out great.

.

If I were looking for a great clocking s1366 chip today I would buy a few higher tier Xeon models from different sellers to get CPUs from different batches to better the odds of getting at least one good one. The other choice would be to sift through old xtremesystems threads to see if I can find CPUs on Ebay with batch numbers close to a known good CPU... but this is tedious work. If I would be satisfied with a decent CPU I would just buy 3 of the model I wanted from the same seller, test them and use the best clocking one. Luckily even top tier s1366 Xeons are cheap as chips.

It would be awesome to build a system on one of these: (Supermicro X8DTN+-F Intel (MBD-X8DTN+-F) Motherboard)

Skyscraper wrote on 2023-05-02, 09:50:

I bought my first i7 920 at launch day, a pretty poor C0 that managed 4.0GHz but totally brick walled at ~4.1GHz regardless of voltage and cooling. As soon as the D0-stepping came out in 2009 I bought a new i7 920, it was not much better than my C0 and the IMC was even worse. During 2009-2010 I did build a few s1366 systems for friends/family/work and I was always on the lookout for a better clocking CPU compared to what I had but I didn't have much luck. I even tried a W3520, the Xeon flavor of the i7 920 but never any of the expensive higher clocked models. When I upgraded and built my EVGA SR-2 system late 2010 I still had not found a Bloomfield chip that would do more than 4.2 GHz at air cooled "24/7 settings".

I also was an early adopter of the original i7 and had a C0 920, and like you hit a brick wall when trying to overclock. If memory serves mine wouldn't even get past 3.2 or 3.4, but that could have been a limitation of the motherboard as I likely got a mid range board. I was pretty disappointed in that chip, especially having come from an E6600 which overclocked quite well. I ended up jumping on the Sandy Bridge platform and my 2500k overclocked like a champ, got it to 4.8 on my first try without tweaking much.

Server boards don't support overclocking. You're stuck with the stock frequency. You gain SMP and ECC but lose that.

Errius wrote on 2023-05-02, 13:40:

Server boards don't support overclocking. You're stuck with the stock frequency. You gain SMP and ECC but lose that.

I've never really been big on overclocking!

Irinikus wrote on 2023-05-02, 13:53:

Errius wrote on 2023-05-02, 13:40:

Server boards don't support overclocking. You're stuck with the stock frequency. You gain SMP and ECC but lose that.

I've never really been big on overclocking!

Have you ever tried a 50% overclock over what the fastest available for sale is?

I've been overclocking since the 80486 days.

Socket 7, Slot A, Socket A, LGA775 and so on get really great gains from running over stock.

LGA 775 really starts to fly around 450Mhz fsb.

Basically, stock really sucks once you have experienced overclocking when you are limited at stock speed.

For retro and vintage stuff it doesn't matter so much because you can just use faster stock hardware.

But when you can get a 50% overclock on some things with almost not tweaks whatsoever... Q6600 from 2.4 to 3.2 Ghz at usually stock voltage or maybe .1v voltage over stock and there is no reason at all to not overclock it

I still have my Q6600 that I ran at 3.84Ghz for quite a few years.

Brawndo wrote on 2023-05-02, 13:11:

I also was an early adopter of the original i7 and had a C0 920, and like you hit a brick wall when trying to overclock. If memory serves mine wouldn't even get past 3.2 or 3.4, but that could have been a limitation of the motherboard as I likely got a mid range board. I was pretty disappointed in that chip, especially having come from an E6600 which overclocked quite well. I ended up jumping on the Sandy Bridge platform and my 2500k overclocked like a champ, got it to 4.8 on my first try without tweaking much.

If you didn't make it past 3.4GHz then something else and not the CPU was in all likelihood the limiting factor. The C0 stepping wasn't great but the absolute worst early i7 920 chips would at least make it to 3.8 GHz with aftermarket cooling.

Sandy Bridge was a bit more easy going though! 😁

I stayed with socket 1366 in form of the EVGA SR-2 as my main system right up until Alder Lake launched. I have had other systems for gaming as the X5690s started to show their age. I have in fact owned (and probably still own) every Intel consumer and HEDT platform except for x299 and Z490/590 (LGA1200 ) which I haven't played with yet. My curiosity usually makes me buy interesting computer stuff I haven't tested as soon as they get cheap enough on the second hand market... which seems to take more and more time...

cyclone3d wrote on 2023-05-02, 14:43:

Have you ever tried a 50% overclock over what the fastest available for sale is? […]
Show full quote

Have you ever tried a 50% overclock over what the fastest available for sale is?

I've been overclocking since the 80486 days.

Socket 7, Slot A, Socket A, LGA775 and so on get really great gains from running over stock.

LGA 775 really starts to fly around 450Mhz fsb.

Basically, stock really sucks once you have experienced overclocking when you are limited at stock speed.

For retro and vintage stuff it doesn't matter so much because you can just use faster stock hardware.

But when you can get a 50% overclock on some things with almost not tweaks whatsoever... Q6600 from 2.4 to 3.2 Ghz at usually stock voltage or maybe .1v voltage over stock and there is no reason at all to not overclock it

I still have my Q6600 that I ran at 3.84Ghz for quite a few years.

I get your point, but have a soft spot for multiprocessor systems!

Sorry, late the party so discussion gone a bit beyond this, but I don't see a great case for socket 1366 beyond personal nostalgia, though I guess minor historical significance since it was start of the Core i line, even though not start of Core. However personal nostalgia is an itch that tends to keep itching until you scratch it. Myself, I have no logical reason to need an AMD 40Mhz 386, but I had one in the day, and I will likely be unable to resist if one crosses my path.

Anyway, I would say that the price for these hasn't really bottomed out yet, and might hit a very low bottom, as natural "age out" bottom would I suspect be in a year or two. However, the Windows 10 support deadline looms and there's gonna be 10+ years worth of stuff made functionally obsolete at one stroke, a massive deluge of newer, objectively better hardware, that will be practically worthless, except to the fraction who are glad to retro or linux re-purpose it. There is still going to massively more "sellers/disposers" than "buyers/adopters" though. In general therefore, the shorter term value of 1366 looks to be around free to negative (pay me to haul it away) BUT, huge but, if you are wanting very specific pieces, you probably want a chance to look for a needle in everyone's sewing supplies, before the deluge happens and you're trying to find a slightly different needle in a tidal wave of needles. So moving earlier might be of benefit, even if you pay a bit more than might theoretically be possible later. However, this may also have the characteristics of a mass extinction event, a pinch point, mass scrappage, so after that occurs, what tickles your fancy from first to fourth gen core might be hard to find... and should a general sense of nostalgia for the period later form, become expensive.

Irinikus wrote on 2023-05-02, 15:44:

I get you point, but have a soft spot for multiprocessor systems!

You can have both and at the same time have one of the most solid motherboards ever built.

https://www.bit-tech.net/reviews/tech/motherb … -sr-2-review/1/

A fun fact is that the board lets you put totally different CPUs in the sockets so you could run a cheap low multiplier 4-core in one socket and a top 6 core in the other socket and overclock both though the BCLK. When adjusting multipliers both CPUs need to support the multiplier though as there aren't user accessible manual multiplier settings for each CPU, just one for both.

The board only supports Xeon 5500 and Xeon 5600 series and when running two CPUs there is a good chance they need to be from the same generation.

Skyscraper wrote on 2023-05-02, 18:03:

You can have both and at the same time have one of the most solid motherboards ever built. […]
Show full quote

Irinikus wrote on 2023-05-02, 15:44:

I get you point, but have a soft spot for multiprocessor systems!

You can have both and at the same time have one of the most solid motherboards ever built.

https://www.bit-tech.net/reviews/tech/motherb … -sr-2-review/1/

A fun fact is that the board lets you put totally different CPUs in the sockets so you could run a cheap low multiplier 4-core in one socket and a top 6 core in the other socket and overclock both though the BCLK. When adjusting multipliers both CPUs need to support the multiplier though as there aren't user accessible manual multiplier settings for each CPU, just one for both.

The board only supports Xeon 5500 and Xeon 5600 series and when running two CPUs there is a good chance they need to be from the same generation.

The last one that sold on eBay went for $700... Hah.

If I could find a working one for cheap, I would be all over it.

cyclone3d wrote on 2023-05-02, 18:13:

The last one that sold on eBay went for $700... Hah.

If I could find a working one for cheap, I would be all over it.

The last time I checked Ebay a couple of years ago there were still some dodgy but working boards in the $400 - $500 range. I'm not really surprised the SR-2 keep going up in value.

.

The pension fund...

Skyscraper wrote on 2023-05-02, 18:03:

You can have both and at the same time have one of the most solid motherboards ever built. […]
Show full quote

You can have both and at the same time have one of the most solid motherboards ever built.

https://www.bit-tech.net/reviews/tech/motherb … -sr-2-review/1/

A fun fact is that the board lets you put totally different CPUs in the sockets so you could run a cheap low multiplier 4-core in one socket and a top 6 core in the other socket and overclock both though the BCLK. When adjusting multipliers both CPUs need to support the multiplier though as there aren't user accessible manual multiplier settings for each CPU, just one for both.

The board only supports Xeon 5500 and Xeon 5600 series and when running two CPUs there is a good chance they need to be from the same generation.

That's indeed pretty beastly!!!! 😀

I'll have to start a search!

I realize it's a bit of a necro, but I was googling around what to do with my Intel desktop board and i7-920 .. and this post came up.

If you want the board and CPU, hit me up. Happy to sell it to a good home.

Edit: the board has the skull logo on it, model DX58S0 (Intel "extreme series" X58)
And the CPU is
SLBEJ COSTA RICA
2.66/8M/4.80/08
39458231

Oh, and I've also got a Xeon w3670 too