Xeon Owners Club

[Edwired]

It will get used again when the box gains a thick layer of dust..

Maybe

The joys of dusting that off

 

[Mr.Scott]

Really? Wow..

Yes.

Mr.Scott`s Cinebench - R15 score: 3431 cb with a Ryzen 9 3900X

The Ryzen 9 3900X @ 4400MHzscores getScoreFormatted in the Cinebench - R15 benchmark. Mr.Scottranks #343 worldwide and #116 in the hardware class. Find out more at HWBOT.

hwbot.org

 

[freeagent]

It is a sweet setup though, it rips pretty good.. he had a pretty good tune on it last time I saw it in action. That SR2 is effing sweet too. If I had one I would probably use it as my daily just to look at it

Edit:

Beat me to it

 

[Toothless]

It is a sweet setup though, it rips pretty good.. he had a pretty good tune on it last time I saw it in action. That SR2 is effing sweet too. If I had one I would probably use it as my daily just to look at it

Edit:

Beat me to it

If my case wasn't so big and I had a see-through panel, I'd do it. Stick the 1080ti in there and watch it go vroom but that build is around 60-80 pounds and too big for any places.

 

[freeagent]

Yes I hear you there man! I love the way my x58 looks, still sexeh af after all these years

I still have my Stacker STC-01 and when that is full of hardware it makes me happy it has wheels

 

[SeanBeanDiesInTheEnd]

I combed over a handful older threads, but 1 question stands out for me. Xeons are built for stability and I can just enter DRAM values manually from XMP settings shown in AID64. Can xeons operate using XMP profiles?

I've read some people saying that when XMP is set they can't even get the system to post with their xeons unless DRAM in the BIOS is set to manual.

Been fighting boredom fiddling with potential settings for when I finally do get the x5675 in my hands to make it fly.

 

[lexluthermiester]

Can xeons operate using XMP profiles?

As long as the motherboard supports XMP profiles, then yes. If it does not but you can still enter the RAM timings into the bios manually, you should still be fine. Worst case scenario is that you have no such support and the RAM will run at default SPD timing profiles, which are generally JEDEC compliant specs.

Remind us all, what board do you have?

 

[SeanBeanDiesInTheEnd]

As long as the motherboard supports XMP profiles, then yes. If it does not but you can still enter the RAM timings into the bios manually, you should still be fine. Worst case scenario is that you have no such support and the RAM will run at default SPD timing profiles, which are generally JEDEC compliant specs.

Remind us all, what board do you have?

Cheers, it’s a p6T deluxe v2 on the latest bios. Runs XMP no problem for the 920, I think what caught my eye was the stories needing hard CMOS resets (bizarre) so gotta ask the pro’s. Personally I think it was just bad DRAM voltages or not enough QPI since the stories had no resolution if they solved things.

This one is something I've even been curious personally RE manual entries, @Regeneration posted for another fellow this example to input XMP values manually:

Is it crucial to fill the values past the red line of the BIOS screenshot as he had in his edited screengrab or can they be left to auto? Basically is there any benefit to filling out the rest.

 

[lexluthermiester]

Cheers, it’s a p6T deluxe v2 on the latest bios. Runs XMP no problem for the 920

You'll be fine then. When it comes to RAM timing profiles, Xeons are equally compatible with the mainstream counter-parts.

 

[SeanBeanDiesInTheEnd]

You'll be fine then. When it comes to RAM timing profiles, Xeons are equally compatible with the mainstream counter-parts.

Excellent

Welp I think I have a good base to work in increments for a stable 4.20GHz OC, and a fun benchmark to 4.5GHz+. Hopefully the chip likes low (v) values that keep things below spec!

Gear:
P6T Deluxe v2 (BIOS 1202)
12gb DDR3-1600 CL9 G.Skill Ripjaws
650w EVGA supernova P2
GTX 1660S 6GB


(profile 1)
4.20GHz

Load Line Calibration: Enabled
CPU Differential Amp.: 800mV / Auto
CPU Clock Skew: Delay 300ps / Auto
CPU Spread Spectrum: Disabled
IOH Clock Skew : 300ps / Auto
PCIE Spread Spectrum: Disabled
————————
AI Overclock Tuner: Manual (9-9-9-24 @ 2n) / X.M.P (profile 1)
CPU Ratio Setting: 21 | 23
Intel (R) Turbo Mode Tech: Disabled
BLCK Frequency: 200 | 183
PCIE Frequency: 100
DRAM Frequency: DDR3-1467MHz
UCLK Frequency: 2935MHz - 3200MHz
QPI Link Data Rate: 6605MT/s
CPU Voltage: 1.30v / 1.31250v / 1.32v / 1.33125v / 1.343v / 1.35v
CPU PLL Voltage: 1.825 / 1.84 / 1.88
QPI/DRAM Core Voltage: 1.25v / 1.26250v /1.30v / 1.325v / 1.35v
IOH Voltage: 1.12 / 1.16
IOH PCIE Voltage: Auto
ICH Voltage: Auto
ICH PCIE Voltage: Auto
DRAM Bus Voltage: 1.50



(profile 2)
4.53GHz / 4.6GHz / 4.8 GHz?

(Load Line Calibration: Enabled
CPU Differential Amp.: 800mV / Auto
CPU Clock Skew: Delay 300ps / Auto
CPU Spread Spectrum: Disabled
IOH Clock Skew : 300ps / Auto
PCIE Spread Spectrum: Disabled
————————
AI Overclock Tuner: Manual (9-9-9-24 @ 2n) / X.M.P (profile 1)
CPU Ratio Setting: 23
Intel (R) SpeedStep (TM) Tech: Disabled
BLCK Frequency: 196 | 197 | 200 | 209
PCIE Frequency: 100
DRAM Frequency: DDR3-1579MHz
UCLK Frequency: 3100MHz - 3206MHz
QPI Link Data Rate: 7074MTs - 7110MT/s
CPU Voltage: 1.36v / 1.37500v / 1.38125v / 1.39375v / for 4.6GHz - 1.40625v, 1.4375v (avoid if possible) / for 4.8GHz - 1.4825v (avoid if possible)
CPU PLL Voltage: 1.825 / 1.84 / 1.85 / 1.88 / 1.9
QPI/DRAM Core Voltage: 1.25v / 1.26250v / 1.26875v / 1.300v / 1.325v / 1.35v
IOH Voltage: 1.12 / 1.18 / 1.20 / 1.30
IOH PCIE Voltage: Auto
ICH Voltage: Auto
ICH PCIE Voltage: Auto
DRAM Bus Voltage: 1.50

 

[Toothless]

You honestly could push to 1.4v safely for a quick bench. When you get your x5675 we'll have to do some head to head benchmarks.

 

[SeanBeanDiesInTheEnd]

You honestly could push to 1.4v safely for a quick bench. When you get your x5675 we'll have to do some head to head benchmarks.

You’re on!

I genuinely believe I’m going to hit the TDP throttle limit of my p6T @ 4.4, I have a gut feeling. Not looking forward to cross flashing the bios since this is our family htpc for COVID entertainment, I may have to cut loses early LOL.

 

[Edwired]

Don't think it will hit tdp limit that quick on your board.

Downside to the crossflashing as you will gain some settings in the bios and some functionality might get broken as it might be risky to brick the board if the process is done incorrectly as you would need an engineering sample of the bios flash and a correct command

 

[oldwalltree]

I'm going to give it every chance to. I think there's around 7 in total, including MX-4. My goal is to focus on the affordable(ish) products so the specialty ones like liquid metal are out. No doubt they would win anyway but it's not worth buying 1 gram for $30 or whatever it is. But yeah it will be an interesting shootout, it will probably be a couple of weeks before the results are up.

Have you tested KPX? I grabbed a 3g tube on amazon yesterday for 12 bucks.

 

[SeanBeanDiesInTheEnd]

Don't think it will hit tdp limit that quick on your board.

Downside to the crossflashing as you will gain some settings in the bios and some functionality might get broken as it might be risky to brick the board if the process is done incorrectly as you would need an engineering sample of the bios flash and a correct command

Fingers crossed!

Going north of 1.4v is when the headache MAY happen from what I'm reading, but anything below that is gravy. Shame because the majority of 4.6GHz+ OC's need that juice, only a few blessed silicon's run south of 1.4v - not looking forward to messing with AFUDOS...

 

[Edwired]

As far as I know your board throttle the multiplayer judging on the info when running 4.4ghz

 

[Toothless]

Fingers crossed!

Going north of 1.4v is when the headache MAY happen from what I'm reading, but anything below that is gravy. Shame because the majority of 4.6GHz+ OC's need that juice, only a few blessed silicon's run south of 1.4v - not looking forward to messing with AFUDOS...

Have some fans on the VRM when pushing that hard. My Rampage III Formula needed some cooling there and on the socket when I was pushing for clocks. Never knew a pump on an AIO could get so hot.

 

[8BitZ80]

Have you tested KPX? I grabbed a 3g tube on amazon yesterday for 12 bucks.

I have some KPX but haven't tried it yet. I couldn't resist ordering some because I've heard good things about it and it's not too expensive.

I've been spending a LOT of hours on testing different testbenches and ended up throwing pretty much all of my original methodologies out the window. They just didn't provide the level of accuracy and repeatability that I would expect.

One of the things I actually didn't expect with testing thermal paste is just how bloody unpredictable the changes in temperature are. Basically, you do one benchmark, come back later and then get a different result on the second run. These are the results when I tried a one-and-done approach:

The rate of heating (higher value = faster heating) doesn't scale with ambient temperature at all, it's all over the place, so that wasn't going to work. But now that I've taken that into account and have been doing multiple runs per test, the results are waaaaay more accurate with the same testbench:

See, what most reviewers do is run a benchmark until the testbench reaches maximum temperature. I wasn't happy with that approach because there's too many issues with it. It takes a long time, the final temperature isn't repeatable, it kicks a lot of heat out into the room which changes the ambient temperature too much.

My benchmark runs are only 2 minutes each and I can evaluate a thermal paste completely in 10 minutes. Better yet I'm able to measure load temperatures to within a tenth of a degree and will be able to adjust results to account for sub-decimal changes in ambient temperature. It should be accurate enough to detect tiny differences in performance between different thermal pastes.

 

[bobbybluz]

My main working test for thermal pastes is how well they work when I'm transcoding large video files. This test is somewhat less effective when using my 16 and 22 core X99 Xeon rigs because they're so much more efficient (and not overclockable) then my other A/V rigs are. I have several different brands of paste on hand and still prefer Gelid GC Extreme over the others. I recently had a tube of GC Extreme that had began hardening far too soon. I contacted Gelid and they sent me a new one for free.

 

[80-watt Hamster]

I have some KPX but haven't tried it yet. I couldn't resist ordering some because I've heard good things about it and it's not too expensive.

I've been spending a LOT of hours on testing different testbenches and ended up throwing pretty much all of my original methodologies out the window. They just didn't provide the level of accuracy and repeatability that I would expect.

One of the things I actually didn't expect with testing thermal paste is just how bloody unpredictable the changes in temperature are. Basically, you do one benchmark, come back later and then get a different result on the second run. These are the results when I tried a one-and-done approach:

View attachment 198589

The rate of heating (higher value = faster heating) doesn't scale with ambient temperature at all, it's all over the place, so that wasn't going to work. But now that I've taken that into account and have been doing multiple runs per test, the results are waaaaay more accurate with the same testbench:

View attachment 198606

See, what most reviewers do is run a benchmark until the testbench reaches maximum temperature. I wasn't happy with that approach because there's too many issues with it. It takes a long time, the final temperature isn't repeatable, it kicks a lot of heat out into the room which changes the ambient temperature too much.

My benchmark runs are only 2 minutes each and I can evaluate a thermal paste completely in 10 minutes. Better yet I'm able to measure load temperatures to within a tenth of a degree and will be able to adjust results to account for sub-decimal changes in ambient temperature. It should be accurate enough to detect tiny differences in performance between different thermal pastes.

Isn't sub-decimal precision an order of magnitude beyond margin of error? As in, it sounds like you're going to try to calibrate tenths against hundredths. Unless I've missed something. That happens a lot.

 

[8BitZ80]

Isn't sub-decimal precision an order of magnitude beyond margin of error? As in, it sounds like you're going to try to calibrate tenths against hundredths. Unless I've missed something. That happens a lot.

I don't think so. Ambient temperature is being measured to the tenth and I'm using HWInfo readings to measure to the tenth. That's what I meant to say.

 

[SeanBeanDiesInTheEnd]

Hey friends, slightly off topic but still related to xeon hopefully as it involved some overclocking and refreshing the memory on settings and such before messing with the x5675. TL;DR not sure what happened to my RAM...

Earlier this morning I was messing around with the 920 trying to get it stable again @ 4.2 before I put it to pasture. Ran:
-1.4250v (high I know) on the vcore
-1.9v on the CPU_PLL
-1.4250v on the QPI.
-DRAM set manually to (9-9-9-24 @ 2n) at its 1.5v rating

What I thought was degraded actually ran prime95 for about 2 hours till it blue screened. I dialed it back to 3.8GHz and my previous stable settings, but my previously confirmed RAM displayed in the BIOS, windows, and CPU-Z of 12288mb of RAM in triple channel now shows 8192mb in dual channel only.

AIDA 64 AND CPU-Z acknowledge there's 6x2gb sticks available, but only running 8192mb in dual channel. Things I've done is:
-reseat each single stick to confirm they work, all work
-set the DRAM to it's lowest speed 1033, nada
-bump the voltage on the DRAM bus to 1.56v, nada

(this is where it gets weird)
-ran “wmic MEMORYCHIP get BankLabel, DeviceLocator, MemoryType, TypeDetail, Capacity, Speed” in CMD and DIMM 0/1/4/5 is only showing @ 1333, DIMM 2/3 not showing at all?

I've read threads of the P6T deluxe v2 board having bizarre acting DIMM's but wondering if the gurus can help make sense of this.

 

[Edwired]

That sounds like the cpu isn't sitting correctly in the socket resulting that as I ran into the same problem on asus p6x58d-e with xeon 5675. What heatsink you using?

I usually tighten/loosen the four corners and blew out the dust bunnies out of the socket and inspect the socket for misaligned pins as some might be missing the fine tips on it. Use a good magnifying glass and tiny sewing needle to do the job BUT WARNING GIVEN is take it slow and steady straighten them out one by one

 

[lexluthermiester]

Edwired might have a point, check your socket to be sure it's seated properly and no pins are damaged.

However, I think that CPU is dying. It's giving off all the symptoms of a die that is cascade degrading.

 

[SeanBeanDiesInTheEnd]

That sounds like the cpu isn't sitting correctly in the socket resulting that as I ran into the same problem on asus p6x58d-e with xeon 5675. What heatsink you using?

I usually tighten/loosen the four corners and blew out the dust bunnies out of the socket and inspect the socket for misaligned pins as some might be missing the fine tips on it. Use a good magnifying glass and tiny sewing needle to do the job BUT WARNING GIVEN is take it slow and steady straighten them out one by one

Running a corsair hydro H80, still full, but plan on flushing and refilling the coolant in the next few weeks. So reseating the CPU is closer to the bottom of the list on what I think is the issue, but I won't rule it out.

Reason I think this way is that I haven't physically removed the 920 from it's socket in over 2 years. Yes I've cleaned around it and reapplied thermal paste annually, but It has been running the 3.8GHz OC rock solid UNTIL I pushed it a bit today in the BIOS. I have had the CPU seat issue once upon a time in my early days as that was a common problem affecting the integrated memory controller this CPU has, p6T boards amplify this problem.

This one has me miffed boys. Combing through the Prime95 logs and EventViewer points to a memory failure that caused the BSOD (Bugcheck code 209).

Edwired might have a point, check your socket to be sure it's seated properly and no pins are damaged.

However, I think that CPU is dying. It's giving off all the symptoms of a die that is cascade degrading.

Man, I guess I accelerated the last PIP it had left this morning LOL. Bummer I don't have another CPU to toss into the unit till I get my hands on the x5675...