New W3520 Xeon Golden??????

[pentium777]

You decide I am not a noob to overclocking by a long shot I go back to 386 overclocking days.

I just received from the EGG a W3520 for my new rig. I've only changed the PLL to 1.25v, along with adjustments to vcore obviously and ram adjustments. Running 12GB (3x4GB) of Mushkin 998798

W3520 Batch #3951A552 Costa Rica Pack Date 1/26/10

Using the STOCK HSF because my water cooling setup isn't ready yet I wanted to test this chip to see if going 32nm with a 980X was worthwhile.

Testing shows OCCT Linpack stable (never erroring) however temps reaching 80C I have it to autoshutdown the test.

4Ghz @ 1.20v actual 1.25v temps reach 80C in 16 minutes, end test.
4Ghz @ 1.175v actual 1.8v/1.22v(load) temps down 5C and I think it'll reach the 1 hr mark without reaching 80C

I had to turn down AC in my office to megacool and house AC to 70F when people talk about the stock intel HSF there is no exaggeration on what a POS it is, I have a TRUE on my E8400 that runs 4-4.2Ghz but no 1366 bracket for it and I'm using that PC right now.

Figured since I only found one reference to a batch 3951A500 that I'd post this new batch that I just bought from Egg about 4-5 days ago. Maybe someone else will want a one.

So any opinions on this chip? Looks good with water to hit high? Or do you think that 4.2-4.5Ghz will take an extreme amount of voltage past 1.4v? I'll update as I further test this chip.

[hlonipha]

man I don't want to discourage you but 16 min of occt is not quite near to be a golden chip. run p95 for a few hours or linx for at least 10 passes full memory usage. My last chip was a W3520, is in the overclocking database @ 4.4ghz with 1.42v i think maybe a tad more. they do run hot so the water cooling will be needed to try to keep it stable and 1.39v that at a minimum you will need for 4.4ghz and that is running p95 for more than 5-6 hours and 20 passes of linx.

As far as the 980x you do not buy that chip for its future prove or if it is going to be worthwhile investment!!! You buy it cause you can !!!! Unless of course you do video editing of some sort that will fully utilize all 6 cores. By far it has been the easiest overclock I ever had and runs cool too. If you can take that xeon to 4.4ghz or a bit higher keep it and enjoy your system.

If you have money to spend then buy a couple of Intel x-25 M's and set them up on raid 0, you will notice the much increase overall speed on your system.

I just finished building a system for a friend and we used raptors for his os hdd and I thought I was on a 486 machine in comparison with mine. Those ssd made the difference.

Good luck

[jcool]

First of all, get rid of that boxed cooler
Second, see if you chip can do this - that's my 3848A352 on water, it needs around 1,13V for 4Ghz prime. And yes, there are better ones out there. Of course, whether a chip is "golden" or not is mostly a subjective matter. Some want a chip that primes high, others one that benches high, and some want high VID chips that are great for aircooling.

[pentium777]

I am using 3x256GB corsair P256 SSD which are the samsung 220 / 200 read/write in Raid0 config.

Still trying to determine if I am going to keep Raid0 or use Windows 7's TRIM, I am unsure how good the GC is and if GC works in Raid0 I think it does but moving forward I don't want to have my drives degrade in performance when I reload and move from Vista to Win 7

[hlonipha]

I figure that by the time mines start deteriorating, it will be time to get the 3rd or 4th gen ssd drives anyhow, so raid them all the way!!!!!!

[PatRaceTin]

my i7 920 (sold) rock stable at

4.5Ghz @ 1.40000 VCore on water cool system

[pentium777]

Further testing showed 4.2Ghz 1.25v stable. I've found I seem to be stuck at a 212 Bclk wall though so it would seem that the max I'll get out of this chip is 212x21=4.452Ghz. Sadly I was hoping for a 4.5-4.8Ghz chip bummer. Still think that it is better than average from what I've seen lately.

[PatRaceTin]

you may try to bump up QPI/VTT Voltage for that

[trn]

Thats a good chip, My w3520 3945a (not a golden batch) runs LinX stable @ 4.2GHz 1.275v load which is pretty good for a newer chip. Nothing beats the old 3845's and 3849's though. Mine hits a Bclk wall at 219 were my PCIe absolutely die at that bclk despite pcie frequency etc.

Further testing showed 4.2Ghz 1.25v stable. I've found I seem to be stuck at a 212 Bclk wall though so it would seem that the max I'll get out of this chip is 212x21=4.452Ghz. Sadly I was hoping for a 4.5-4.8Ghz chip bummer. Still think that it is better than average from what I've seen lately.

[CryptiK]

It's good but I wouldn't call it golden in term s of low voltage requirement by a long shot.

My 920 does 4GHz at 1.112v, 4.2GHz at 1.168v, and 4.5GHz at 1.32v with HT on. I don't know if I'd even call mine golden, but it's pretty good.

[CryptiK]

It's good but I wouldn't call it golden in term s of low voltage requirement by a long shot.

My 920 does 4GHz at 1.112v, 4.2GHz at 1.168v, and 4.5GHz at 1.32v with HT on. I don't know if I'd even call mine golden, but it's pretty good.

[Rubycon]

It's good but I wouldn't call it golden in term s of low voltage requirement by a long shot.

My 920 does 4GHz at 1.112v, 4.2GHz at 1.168v, and 4.5GHz at 1.32v with HT on. I don't know if I'd even call mine golden, but it's pretty good.

I would call that spectacular IF you can run it fully loaded (Linpack) for 5 hours or more.

I've lots of 3520s pass through my hands and so far the best (close to yours actually!) was from last May. The newer ones, while some are close, are still not as good. 4GHz should be guaranteed and 4.2 is getting close to the limit (average over all specimens tested). Always with HT ON and VCORE never tested higher than 1.4V/1.4V VTT (actual)

[jcool]

Thats a good chip, My w3520 3945a (not a golden batch) runs LinX stable @ 4.2GHz 1.275v load which is pretty good for a newer chip. Nothing beats the old 3845's and 3849's though. Mine hits a Bclk wall at 219 were my PCIe absolutely die at that bclk despite pcie frequency etc.

Yeah BCLK seems to be one huge issue on newer chips. I can hit 245+ on my 3848 easy.. on water.

[CryptiK]

I would call that spectacular IF you can run it fully loaded (Linpack) for 5 hours or more.

I've lots of 3520s pass through my hands and so far the best (close to yours actually!) was from last May. The newer ones, while some are close, are still not as good. 4GHz should be guaranteed and 4.2 is getting close to the limit (average over all specimens tested). Always with HT ON and VCORE never tested higher than 1.4V/1.4V VTT (actual)

5 hours? No need for that IMO. I ran 20x LinX at max mem setting at 4.5GHz @ 1.32v for my submission in the '4.5GHz LinX Stable Club' and that is plenty stable enough for me.

[Rubycon]

5 hours? No need for that IMO. I ran 20x LinX at max mem setting at 4.5GHz @ 1.32v for my submission in the '4.5GHz LinX Stable Club' and that is plenty stable enough for me.

If a system has 12+ GB 20 passes may be OK. However on lower memory systems it's just not running long enough IME. Hence a 300 minute test is good as you can change the memory amount and run for the same period of time. With smaller memory amounts this means a lot more tests as well!

p.s. This is very important with i7 980X as they don't seem to fully stress the cores unless 512MB size is selected and with 20 passes, this will only run for a few minutes.

[CryptiK]

Not sure if we're on the same page here, I used max mem setting which was over 5GB of memory selected. It takes over an hour to do 20 passes.

Anyway, to get back on topic, the OP's CPU is pretty good for a new CPU. Not many decent ones have appeared since the good old Week 45 - 49's from 2008.

[[XC] Oj101]

Totally off topic, but clocking 386s? That was awesome! BIOS settings? Nope, too old. DIP switches? Nope, still too old I only got to clock a 386 in around 2001 but man they were impressive. ~50% increase with passive cooling? What went wrong?

[pentium777]

Further testing proved 212 BCLK not viable, was Linx stable for 50 passes and tested for 2 days but something happened and it won't hold anymore, found this out when I raised to 2:8 on ram (1696) and it was failing.

So far tonight testing with 211 BCLK I still have CPU set for low multi and was able to get 1690 ram speed on 12GB (3x4GB) Mushkin 998798 with 7-9-8-24-1T, no errors memtest no errors linx testing 3GB and 12GB.

I've been holding off on raising CPU past 4.2Ghz (1.25v was stable) until I get this mounted in system with quad rad water cooling. Then I'll be raising it up and apparently 4.4Ghz will be my limit due to 211 bclk what seems to be a wall for it.

What is advisable for the uncore? It is currently at 16X should I try to tweak the uncore to 17X and see if it is still stable or does it even help performance much at all?

[pentium777]

No one else with new W3520s? How come people are buying 920s and 930s instead? Just curious as I thought these Xeon W3520s were higher binned and better chips.... Supposed to be identical to I7 920 in every single way except it allows ECC.

Is there a reason more people aren't going for these W3520s?

[drdeath]

5 hours? No need for that IMO. I ran 20x LinX at max mem setting at 4.5GHz @ 1.32v for my submission in the '4.5GHz LinX Stable Club' and that is plenty stable enough for me.

20X is not enough. I ran 100 passes and at 90 I was like wooohooo. 2 min later... Crash. IMO you are not 100% stable at 20 passes. Runnable, yes.

[trn]

Yeah.. odd, they sure don't make em like the used to (3849B's and similar.) If I was going to get some colder weather and break 219 bclk i'd try pulling the vid card and sata drives and just run a IDE hardrive and remote desktop in to see if that avoided the PCIe problems at 219. However, I don't really care because all my i7 quads are going out to crunching pasture to run nice happy lives at 4.0GHz.

Yeah BCLK seems to be one huge issue on newer chips. I can hit 245+ on my 3848 easy.. on water.

[emertX]

Hmmm...that's just subjective I think!

Have a look at mine:
4,2GHZ @ 1,176V

4,6GHZ @ 1,344V


I think, that's a bit better than yours but also a bit hotter!

[FedericoUY]

Hmmm...that's just subjective I think!

Have a look at mine:
4,2GHZ @ 1,176V

4,6GHZ @ 1,344V


I think, that's a bit better than yours but also a bit hotter!

Hey, that's one of the lower voltage chips i've seen, along with cryptik's. So far I'm really happy with mine; does 4000 @ 1.18, 4210 @ 1.247 and 4503 @ 1.37 prime95 stable (batch 3928). Been benching at 4630 @ 1.388 but not sure if it is 100% stable.
Hey emertx, cryptik, have you tried max bclk?
How far does it goes?
Cheers.

[emertX]

[l0ud_sil3nc3]

I just picked up systemvipers old w3540 and it is batch 3846B with a 23x multi. I am hoping I can get 4.5 ghz stable on air 24/7.