General Intel HEDT Xeon/i7 Discussion

[Zando_]

4 hours ago, wild0077 said:

And the last one I remember right now is the "CPU Switching Frequency". Again, if someone could please explain to me what exactly does it do, I'd be immensely grateful. Also, could an overclocked cpu benefit from tweaking this as well?

Likely the VRM switching frequency, my EVGA X58 boards allow me to adjust this as well. I wouldn't touch it unless you're down for some exhaustive testing: https://www.overclock.net/threads/fixed-cpu-vrm-switching-frequency.1709130/. It could be helpful if you're willing to sit down and manually find the optimal frequency for your specific VRM transistors on your specific board. I assume it's only useful to folks pushing for world records or on LN2, same as a lot of the advanced OC settings on this kind of board. 

5 hours ago, wild0077 said:

The second one is "CPU Phase Control". In many X99 guides they mention something about setting it to "Performance", however this MSI board have only "Normal", "Optimized" and "Disabled" options. Which one would be recommended for overclocking scenarios?

https://www.overclock.net/threads/cpu-phase-control.793584/. At least for older boards, it seems to control how many phases it uses for voltage delivery, it can drop down to run more efficiently but it won't be as stable under OC load, so the OC.net folks recommend disabling it. 

3 hours ago, Crunchy Dragon said:

I'd imagine this assists with power delivery and has something to do with the VRM phases. "Optimized" should mean the same as "Performance".

Given the info above, I'd assume optimized is using less phases, not more, so it would be worse for OC. It would be great if all OEMs just said what the thing actually does vs slapping some stupid name on it. 

5 hours ago, wild0077 said:

The first option is the "VR 12VIN OCP Expander", could someone please explain what exactly does it do? Could any X99 overclocked cpu benefit from tweaking this at all? 

Not likely unless you're doing extreme overclocking. OCP = OverCurrent Protection, given that name I would assume it raises the OCP limits for the 12v input, which would be useful if you were drawing an ungodly amount of current and it kept tripping. 

5 hours ago, wild0077 said:

While I do believe that is normal, I mean, we are talking about a used soon-to-be 10 years old cpu, lmao, I didn't expected the wear to just "kick in" like this, after only 1,5 years of mild use, AKA gaming.

Running the exact same benchmark (and same version of said benchmark) to compare? It likely just wasn't completely stable to begin with and whatever test you're using now exposed an instability you never came across during the original tweaking. Getting chips 100% rock solid stable isn't easy in my experience, I tend to get them to ~98-99% stable but I'll still have a funky issue every now and then and have to re-tweak. That + thermal headroom is why I don't push my chips as hard anymore. 

[wild0077]

7 hours ago, Zando_ said:

Likely the VRM switching frequency, my EVGA X58 boards allow me to adjust this as well. I wouldn't touch it unless you're down for some exhaustive testing: https://www.overclock.net/threads/fixed-cpu-vrm-switching-frequency.1709130/. It could be helpful if you're willing to sit down and manually find the optimal frequency for your specific VRM transistors on your specific board. I assume it's only useful to folks pushing for world records or on LN2, same as a lot of the advanced OC settings on this kind of board. 

Well that was a good read, it seems like power hungry cpus could actually see a type of benefit in tweaking that setting, I actually remember messing with it on a Z97 board, back in 2015 or something. The issue for me is just that this board have A LOT more of options for VRM Switching Frequency, and the numbers go pretty high compared to that old Z97 board. That, combined with the fact that I simply could not find anyone that already tested it with this specific X99 board, made me steer away from it entirely. Is it possible to actually cause damage to the cpu or the board itself by tweaking the VRM Switching Frequency alone? 

 

 

7 hours ago, Zando_ said:

At least for older boards, it seems to control how many phases it uses for voltage delivery, it can drop down to run more efficiently but it won't be as stable under OC load, so the OC.net folks recommend disabling it.

Thank you, I had a clue about this one, but had never found anyone really talking about it.

 

 

7 hours ago, Zando_ said:

It would be great if all OEMs just said what the thing actually does vs slapping some stupid name on it. 

Tell me about it.......     Not to mention the super dodgy "explanations" on the bios, that if the option even has one lol.

 

 

7 hours ago, Zando_ said:

Not likely unless you're doing extreme overclocking. OCP = OverCurrent Protection, given that name I would assume it raises the OCP limits for the 12v input, which would be useful if you were drawing an ungodly amount of current and it kept tripping. 

Umm... what exactly would be an ungodly amount? XD   This cpu is actually pretty power hungry even at 4.5ghz, I know that technically it's a 50% overclock, but still, I actually have it power limited on bios, because all I want is the IPC increase for the games, but now these days the dreaded unpleasantly long shader compiling thing hitting the cpu 100% is getting more and more common, I didn't felt like rushing to task manager and disabling cores every time that happens lol. At 100% loads this will actually push over 200W if I leave it "unshackled", and with that, temps above 80-90s easily, which I find very uncomfortable for some reason.
Tripping that protection would simply shut down the system, right? No BSOD or reboot, just shut down?  If so, that has actually never happened, so I guess I really don't HAVE to tweak that.

 

 

7 hours ago, Zando_ said:

Running the exact same benchmark (and same version of said benchmark) to compare? It likely just wasn't completely stable to begin with and whatever test you're using now exposed an instability you never came across during the original tweaking. Getting chips 100% rock solid stable isn't easy in my experience, I tend to get them to ~98-99% stable but I'll still have a funky issue every now and then and have to re-tweak. That + thermal headroom is why I don't push my chips as hard anymore. 

Well, actually, no, not the same version, but, I did noticed something that escaped my attention before: back then I was running 2666MHz CL12, it was the same kit though, but I was experimenting with the stuff. Now I am daily running 3200MHz CL16 for quite some time, and checking HWiNFO (it was open as well in the screenshot), the "Uncore/SA Voltage Offset" is set to 0.500V now, while the screenshot from 2021, running 2666 CL12, shows only 0.125V.
I believe I left this option on "AUTO" in both occasions. The thing is HWiNFO does not give me a voltage reading, nor the bios, only the offset, but I do think that could be a bit high, right?
Oh, and one last thing, but this one is related to the cpu itself (I think, lol). I noticed that the core temps have a pretty wild delta, like right now for instance, after a quick CPU-Z benchmark, cores #1, #3, #5 and #7 reach 62ºC, 62ºC, 63ºC and 63ºC respectively, meanwhile cores #0, #2, #4 and #6 went 66ºC, 69ºC, 75ºC and 71ºC respectively. Is this normal behavior? I mean, I do know that some core delta is normal, but this high? I swear I've seen 15 to 18ºC difference between hottest and coolest cores back when I was running benchmarks and stuff. Repasted, reseated cooler, even tried different TIM application methods, but nothing changed. If the SA Voltage Offset is indeed too high, could it be further increasing this delta between cores?


Thank you so much for your answer, and for your time, I really appreciate it.

[Zando_]

1 minute ago, wild0077 said:

Umm... what exactly would be an ungodly amount?

5960X extreme overclocking amount. The Xeons don't have the OC headroom to get that saucy in the first place. And yep, OCP shuts the system off instantly, so you would very much know if it tripped.

 

That core delta looks completely normal. The RAM itself could have something to do with it though. I've never directly tested, but according to @Damascus (he has ran a lot of X99 SFF builds) running above 2666Mhz increases CPU temperatures, so it definitely could influence the CPU itself. I've always run my RAM at 2800Mhz (my EVGA kit) and 3200Mhz (my HyperX kit) without issue on a 5820K, 5960X, and 6950X.

6 minutes ago, wild0077 said:

because all I want is the IPC increase for the games, but now these days the dreaded unpleasantly long shader compiling thing hitting the cpu 100% is getting more and more common, I didn't felt like rushing to task manager and disabling cores every time that happens lol. At 100% loads this will actually push over 200W if I leave it "unshackled", and with that, temps above 80-90s easily, which I find very uncomfortable for some reason.

Keep a lookout for deals on Broadwell-E chips. They have a much lower OC headroom (most cap out at 4.4Ghz or so, and the 6950X often caps out at 4.2), but markedly higher IPC and put out a lot less heat. My 6950X at 4.2Ghz core/3.5Ghz uncore puts out the same single core score in Cinebench as my 5960X at 4.7Ghz core/3.7GHz uncore did. So same core performance at ~200-300Mhz lower clocks and lower temperatures to boot. Better in games than my 7980XE is due to still being ringbus arch, the X299 chips moved to mesh.

 

Looping back around to the VRM stuff... it shouldn't break anything. I read into it more back when I was tweaking X58 stuff, but that was 3+ years ago now so I don't have much more to offer. Don't recall anything about it kersploding the CPU though, and I think I would have remembered that.

[the pudding]

What's better than one 7980XE? 

 

three of them

[the pudding]

just from some quick testing with a U12A, one is a bit so so, 4GHz 1.2v, one is alright, 4.2 1.17v and one seems to be pretty good with 4.4GHz with 1.2v, looking forward to seeing what this last one does under water 

[Crunchy Dragon]

1 hour ago, the pudding said:

just from some quick testing with a U12A, one is a bit so so, 4GHz 1.2v, one is alright, 4.2 1.17v and one seems to be pretty good with 4.4GHz with 1.2v, looking forward to seeing what this last one does under water 

You're making me want to invest money into my aforementioned 5820K overclock testing idea

[the pudding]

11 hours ago, Crunchy Dragon said:

You're making me want to invest money into my aforementioned 5820K overclock testing idea

You're welcome

[Slayer3032]

More hardware has found its way back around to me, X5670 with way too small of a cooler and a EVGA E758 with a 750Ti. The Rosewill PSU is probably faulty as well as the system was always weirdly unstable and was also originally unstable on a Z170 board. However it's also built into my first case purchase, an Antec Nine Hundred which has had my Athlon X2, EVGA 680i/Q6600 and X58A-UD5 in it.

 

No idea what I'm going to do with it. My best spare PSU is a 550w. I also have a EVGA GTX 970 now as well which would really put my 300w~ at the wall X5675 build at the exact limit of my Superflower Leadex III Gold. I don't ever plan on drawing that absolutely ever on that machine though.

 

I really wish some X79/X99 stuff would just fall into my lap but that's likely not happening, the P6T/X5660 home server could probably use a bit more of a power efficient platform than X58 lmao.

[Crunchy Dragon]

Starting to hear rumors of EVGA pulling out of the motherboard market. That's a bit spooky, hopefully it isn't so.

 

Otherwise, I may accelerate my plans and snag EVGA boards on X79, X99, and probably Z690 for good measure while I still can.

[the pudding]

On 7/7/2023 at 10:29 PM, Crunchy Dragon said:

Starting to hear rumors of EVGA pulling out of the motherboard market. That's a bit spooky, hopefully it isn't so.

 

Otherwise, I may accelerate my plans and snag EVGA boards on X79, X99, and probably Z690 for good measure while I still can.

I'd hope not, I'd love a W790 dark haha 

 

[the pudding]

Speaking of xeon dark, snagged this chonky lad for a good deal local 

[da na]

Ever wondered what four Core 2 Duos glued together would perform like? Here's a server with two Xeon X5365, an LGA771 chip that is for all intents and purposes a Q9650 with less L2 and more L1. It does not quite double the Q9650's performance but I would put this more up to the slow FBDDR2 memory in the server vs. the DDR2-1200 in the system I tested the Q9650 on. 

[Crunchy Dragon]

20 minutes ago, da na said:

Ever wondered what four Core 2 Duos glued together would perform like? Here's a server with two Xeon X5365, an LGA771 chip that is for all intents and purposes a Q9650 with less L2 and more L1. It does not quite double the Q9650's performance but I would put this more up to the slow FBDDR2 memory in the server vs. the DDR2-1200 in the system I tested the Q9650 on.

That would make sense.

 

It would be nice to see the labels for the spreadsheet columns. I'm still not fully sure what CB score is here, but I'm also not sure if I'm seeing levels of cache or lithography....lol

[Pasi123]

3 hours ago, da na said:

Ever wondered what four Core 2 Duos glued together would perform like? Here's a server with two Xeon X5365, an LGA771 chip that is for all intents and purposes a Q9650 with less L2 and more L1. It does not quite double the Q9650's performance but I would put this more up to the slow FBDDR2 memory in the server vs. the DDR2-1200 in the system I tested the Q9650 on. 

The X5365 is an older 65nm CPU so it's more like a QX6850. The 45nm Q(X)9650 equivalent Xeons would be the E5450 and X5450

 

I actually have a server with 2x E5450 or X5450 (idk which because it reports as having both lol) but I haven't ran R10 on it so here's R11.5

[Crunchy Dragon]

My turn to buy more workstation hardware.

 

Have a Xeon E5-2690v4 on the way, plus a kit of 4x16GB DDR4 -- I may upgrade my motherboard in the future, and might want 128GB down the road -- as it turns out, the plan to revive the X99 machine is to rebuild it into a server and run virtual machines. It'll more than likely end up pulling NAS duty as well. I also don't actually have all the parts for the spare CPU cooler I thought I had, so I snagged a replacement and upgrade in the form of a Noctua NH-U12S. Hopefully it'll keep the 135W 14-core monster quiet and happy.

 

You know the drill. Pictures will arrive as parts do.

I'm very quickly reminded just how old the Define R5 is as a case. Wow.

[Crunchy Dragon]

CPU cooler and SSDs have arrived! Still waiting on the actual CPU and the RAM, but I was able to get my BIOS updated no problem. Somebody(probably me) had updated it to the crucial version(F22) in the past, so I updated to the most recent version(F24c) just for peace of mind.

 

Cable management is looking pretty alright, considering it's the Define R5 that I pretty notoriously did not enjoy working in. Things are a little better because there's just about nothing in the case right now. Feel free to rate it:

Spoiler

 

Spoiler

Eagle-eyed viewers may recall this is almost the exact hardware configuration I had in my ATX-modded PowerMac G5, which is the reason for the velcro.

 

Future plans include moving drives around physically so I can fit a decently sized RAID10 array for the NAS, and more immediately, replacing the Quadro K600 with a GT 1030 or something else with a decent amount of horsepower that's passively cooled. This system is pretty quiet already but I stopped the GPU fan from spinning and it went from "quiet" to "silent".

 

I should have the whole thing mostly up and running over the weekend. I really enjoyed working with the Noctua U12S, this might be my go-to "budget" air cooler. Seems like it has pretty good performance and low noise output for just $50, and should handle the upcoming 135W behemoth with no problems.

 

Not sure when I'll get it fully set up as a NAS, that's a bit of a stretch goal at the moment(similar to moving the system to a new case that I like working in more). I'll probably move some hard drives from my workstation over just so that can function as some network storage so that I can move files around without requiring my workstation to be powered on.

[Crunchy Dragon]

RAM is in.

 

Don't look at the SSDs, they're M.2s that are very good at blending in. Better to not waste your time with that one, it's really not worth it. Will likely be installing Proxmox and spinning up some VMs over the weekend.

[Crunchy Dragon]

So I took a bit of a gamble on Ebay, because I was feeling adventurous.

Bought an Asus X99 Deluxe bundled with an i7-5930K for $115 before taxes, free shipping. Was not listed as "working" or "tested".

 

Showed up today with a nifty kit of GSkill Ripjaws4 16GB DDR4, and I tested it to find out everything works just fine. Working to swap the new board into the Proxmox Box because I really don't like Gigabyte and I was reminded of that fact when I started working on the Proxmox Box.

 

Feels like I really stole this one, not gonna lie. From what I could find, X99 Deluxe is one of the better/best X99 boards, and those don't typically sell for less than $150 unless they're broken, and that's just for the board. 5930K is kind of a boss of a CPU still, being 6c/12t at 3.5Ghz with 40 PCIe lanes.

[Crunchy Dragon]

So the X99 Deluxe worked yesterday, worked with all my parts yesterday, but gave up on booting with the Xeon installed today.

 

Made a thread detailing things further:

 

TL;DR is X99 Deluxe will not boot with the Xeon installed(despite booting with it yesterday), X99 Deluxe will boot with Core i7 CPUs, X99-UD4 still boots fine with everything, so no hardware is actually dead.

 

[Pasi123]

Cinebench 2024 on my X5670 @ 4.4GHz. I had to run it in a macOS VM because the Windows version requires AVX2 while the Mac version only needs SSE4.2

 

[Crunchy Dragon]

8 hours ago, Pasi123 said:

I had to run it in a macOS VM because the Windows version requires AVX2 while the Mac version only needs SSE4.2

That's a little bit funny, I wonder how much better performance would be if you were able to run it natively.

[the pudding]

Finally got my X299 micro 2 back off of my brother haha, decided to use it for now, keep the X299 dark for playing with and for benching with my other more badder 7980XE, got a nice OCUK binned 7980XE in this one, 4.8 1.3v, likely just stick to this until I get my W3175X for the SR3 haha 

[Pasi123]

I might have bought a X299 board + CPU. I still need to order more parts for a complete build but for now I put it in to the case of my Z77 system.

 

Current specs:

Intel Core i9-10900X at stock

Asus TUF X299 Mark 1
8GB DDR4 2400MHz single channel (it's the only DDR4 stick I have lol)

Asus GTX 760 DC2

Antec C400 with an added ID-Cooling fan

Antec Sonata III

80GB Seagate HDD from 2009
Windows 11 Pro (originally installed on an Asus P5B with a C2D)

 

 

[da na]

Xeon X5550 engineering sample sent to me.

It is recognized as an "Intel                   000" in the BIOS and Windows. It appears to have half the L1 per core of the production chip.

HWiNFO recognizes it as an X5550 engineering sample despite the wonky brand name. Oddly it is recognized as Nehalem, rather than Gainestown.

[Pasi123]

I finished my X299 build few days ago.

 

CPU: Intel Core i9-10900X 10c/20t
Cooler: DeepCool Assassin III
Case: Fractal Meshify 2
Memory: 64GB Kingston DDR4 3200MHz (just 2x 32GB in dual channel for now)
Motherboard: Asus TUF X299 Mark 1
Graphics Card: Asus GeForce GTX 1080 Strix Advanced
SSD: 2TB Samsung 970 EVO Plus (OS and programs)
SSD: 2TB WD Blue SN570 (games)
HDD: 8TB Seagate Exos 7E10
Power Supply: Seasonic Focus GX-850
OS: Windows 10 LTSC 2021

 

(with a GTX 760 for testing)

 

GTX 1080 inside the case. The case has a dark tinted glass so pretty much the only visible things are the RGB LEDs on the GPU and motherboard