![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
1366 x58 Xeon Enthusiast overclocks club
- Thread starter primetime
- Start date
Virtually every one in this thread is using the turbo multiplier to get clocks that high and higher. There are exceptions, but for the most part the wheels start coming off on most X58 boards at or before ~220BCLK. So it's pretty much required for the X5650, and even on the 5670/75 chips you're going to want to use it so the chip can clock down and reduce voltage when not needed. Where it gets tricky is not every board supports all the same tricks such as the ability to lock in your turbo multiplier under a full 6 core load. Further, it's often difficult to help others as all the vendors choose to use their own specific vernacular to describe certain settings. Your best bet short of someone in here using your specific motherboard is to check for old overclocking guides for the i7 920 specific to your board.
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,362
the only bad thing about those guides is some have information intended for 45nm cpus and these 32nm chips can not handle the same voltage as the older generation without damage
I was speaking more along the lines of settings cause/effect than for voltage. EIST, Current feedback override, etc.
Bad news mates. Looks like my fun in this thread is over. Southbidge reflow went poorly despite my best precautions. I've been lucky with such ventures in the past. Not so today.
Barring x58 motherboards plummeting in price due to x99, I doubt I'll be back. Ill put my xeons and mobo water block up in the fsft section soon. The mobo is probably fixable if you have access to a machine that can do a proper removal of the southbridge and reball.
Thanks for the sharing of knowledge and helping me push my overclock. Looks like the x5675 I have on the way will have to be tested elsewhere.
Barring x58 motherboards plummeting in price due to x99, I doubt I'll be back. Ill put my xeons and mobo water block up in the fsft section soon. The mobo is probably fixable if you have access to a machine that can do a proper removal of the southbridge and reball.
Thanks for the sharing of knowledge and helping me push my overclock. Looks like the x5675 I have on the way will have to be tested elsewhere.
I am running it without turbo now and in firestrike if that is a good judgment the chip is faster at the same clock speed without turbo as I guess all cores are at 4.2Ghz.. I don't plan to go past 1.35v but it did help me get my ram speed back up to 1600Mhz clocking it like that and i'm at 199BCLK and I may try and drop the vcore a click or two as the max temp so far has been 60c at 77F room temp.
It really help me in World of Tanks as I run max setting now and it powers thru all the maps with good 50 to 100 fps which has always been tough because of single tread and needing a fast core to push it.
It really help me in World of Tanks as I run max setting now and it powers thru all the maps with good 50 to 100 fps which has always been tough because of single tread and needing a fast core to push it.
King Icewind
Supreme [H]ardness
- Joined
- Jan 17, 2008
- Messages
- 5,442
I have had both a L5639 and a x5650 now. A difference of 2.13ghz and 2.66ghz stock speed, and 16 and 20 multiplier. The highest I could achieve on the L5639 was 3.2ghz, and I'm at 4ghz with the x5650. The L5639 took less voltage(s). They both are pretty much the same price on ebay now, so definitely just get the x5650 or if you are set on a L5639 I actually still have mine (hint hint
). One advantange to the L5639 is it's only a 60watt TDP processor which meant it ran nice and cool.
Zoson
Supreme [H]ardness
- Joined
- Aug 4, 2001
- Messages
- 6,113
Remember, by altering the clockspeed and voltage you're running the CPU at, you're actually changing the TDP.
TDP ~ Dynamic Power + Short Circuit Power + Leakage Power
This breaks down to:
ACVf + tAVIf + VI(leak)
Where:
A = activity
C = total capacitance driven by all gate outputs
V = operating voltage(core voltage)
I = current, also in the third term specifies the current leakage
f = clock frequency.
Modern CPU TDP is about 99% dynamic power, so we only need to really be concerned with ACVf.
Source
So the thing that has the biggest overall change in TDP is the core voltage. Since core voltage is directly related to frequency(f) we get the following:
f(max) = [(V-Vthresh)^2]/V
This shows that lowering the voltage threshold will increase the maximum frequency.
Now, this actually also changes gate leakage(the third term in our TDP equation): VI(leak)
I(leak) = e^(-q*Vthresh)/(kT)
Where T is temperature, and q and k are constants caused by physical properties.
Basically, when overclocking, the starting TDP of a chip doesn't mean anything unless its starting voltage and frequency are the same as another chip with a higher TDP. It's a sliding scale where a 130W 1.2v cpu at say 3.5GHz can be turned into a 95W TDP cpu by simply lowering the voltage and clockspeed. This is another reason why buying the top bin can net you significantly better results.
TDP ~ Dynamic Power + Short Circuit Power + Leakage Power
This breaks down to:
ACVf + tAVIf + VI(leak)
Where:
A = activity
C = total capacitance driven by all gate outputs
V = operating voltage(core voltage)
I = current, also in the third term specifies the current leakage
f = clock frequency.
Modern CPU TDP is about 99% dynamic power, so we only need to really be concerned with ACVf.
Source
So the thing that has the biggest overall change in TDP is the core voltage. Since core voltage is directly related to frequency(f) we get the following:
f(max) = [(V-Vthresh)^2]/V
This shows that lowering the voltage threshold will increase the maximum frequency.
Now, this actually also changes gate leakage(the third term in our TDP equation): VI(leak)
I(leak) = e^(-q*Vthresh)/(kT)
Where T is temperature, and q and k are constants caused by physical properties.
Basically, when overclocking, the starting TDP of a chip doesn't mean anything unless its starting voltage and frequency are the same as another chip with a higher TDP. It's a sliding scale where a 130W 1.2v cpu at say 3.5GHz can be turned into a 95W TDP cpu by simply lowering the voltage and clockspeed. This is another reason why buying the top bin can net you significantly better results.
King Icewind
Supreme [H]ardness
- Joined
- Jan 17, 2008
- Messages
- 5,442
Yeah obviously increasing power increases heat output but a processor with a lower designed TDP running at the same overclocked speed and similar voltage as a higher TDP processor would have a lower TDP, would it not?
Zoson
Supreme [H]ardness
- Joined
- Aug 4, 2001
- Messages
- 6,113
The way the binning process works, the highest TDP chips would most likely run at a higher clockspeed if you were to figure out where the 95W threshold is and underclock it. While not *always* true; basically the highest bin chips will run the fastest speeds at the lowest voltages, thus also running at the lowest TDP for those speeds.
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,362
taking zosons advice i lowed my block and went with the highest multi.....sure enough i was able to lower all my unsafe voltages and achieve the same or higher clocks.....fully stable and im not having to worry about blowing the cpu up....i have had some success getting tubo 24 stable as well but still needs a bit of tweaking to be 100% stable at 24x...but its not far from it at all at 4.6 but until then 4.2 will do
Yeah the one Achilles Heel of the 5650, the 22x Max Multi hence my 210x22
verylostindeed
Gawd
- Joined
- Dec 27, 2004
- Messages
- 660
Hi all - have been running i7 920 @ 4ghz for longer than I can remember. Thinking of trying out x5650. I've got a custom water loop, so should have adequate cooling. Do you guys think the performance difference above the 920 is worth it if primary use case is gaming?
Well you can run firestrike and see if the Xeon is an upgrade as at 4.2Ghz my x5660 did Physics Score14434 which was
Physics Test 45.82 fps
I hear all the SB /IVY Haswell guys saying those would be upgrades to my Xeon in gaming but I have seen any big number from them in the Physics Test yet.
Physics Test 45.82 fps
I hear all the SB /IVY Haswell guys saying those would be upgrades to my Xeon in gaming but I have seen any big number from them in the Physics Test yet.
Yeah I ran a 920@4Ghz myself for years and the Xeon is definitely an upgrade. The degree of upgrade will vary game to game but its a boost the more cpu dependent you are. If you have a custom loop I'd cautiously say 4.5Ghz should be quite achievable for you. I know the 5650 is only $70 but there is that one ebayer selling the 5675 for like $125-ish. Might consider that if you want the bigger multipliers. You might get closer to 5Ghz with that chip and your custom loop
Its almost linear. I was getting a little over 10K with my quad core 920, my hexacore Xeon is over 15K in the aforementioned Firestrike Physics Test.
My 3930K isn't all that far beyond my X5670, most of the difference is due to the 200MHz clock difference. It's going to depend on the benchmark though, I'd need a lot more than +200MHz to match the 3930K in Cinebench for instance.
[email protected] - 10811 Physics
[email protected] - 17027 Physics
[email protected] - 17915 Physics
I've just got my x5650 today to have a play around with in my old Asus P6X58D-E and it seems for once I've hit the jackpot in the silicon lottery.
I only had 20 minutes or so before I had to go to work but this one booted into Windows and I was able to run Prime95 at 4000MHz with default voltage (1.18v) and default QPI volts (1.2v) aswell. That's with DDR3 at 1600MHz.
Its seems that when on 22x multiplier it holds on all cores under load on the P6X58D-E according to TurboGT. However I cant seem to get a clear answer on this, is the 22x actually working on all cores or is it an error? Fortunately my board did 220bclk with my old i7 930 and the lack of QPI volts needed at 200MHz with the Xeon suggests it should be OK.
I can't wait to test it further as it got into windows at 4.4GHz (22x200MHz if it is indeed correct) with 1.24v too.
I only had 20 minutes or so before I had to go to work but this one booted into Windows and I was able to run Prime95 at 4000MHz with default voltage (1.18v) and default QPI volts (1.2v) aswell. That's with DDR3 at 1600MHz.
Its seems that when on 22x multiplier it holds on all cores under load on the P6X58D-E according to TurboGT. However I cant seem to get a clear answer on this, is the 22x actually working on all cores or is it an error? Fortunately my board did 220bclk with my old i7 930 and the lack of QPI volts needed at 200MHz with the Xeon suggests it should be OK.
I can't wait to test it further as it got into windows at 4.4GHz (22x200MHz if it is indeed correct) with 1.24v too.
Makaveli@BETA
2[H]4U
- Joined
- Mar 24, 2004
- Messages
- 2,433
Hey guys - I just picked up a x5650 on eBay to replace my 930. I have very little experiencing OC'ing and would be very appreciative of some hand-holding as I put this Xeon on my Rampage III and get it running in the 4ghz territory. How do I get this ball rolling?
Hyperthreading is on. It completed an hour of Prime95 at 1.37v with 1.25v QPI and all other voltages at default:
It seems to be a great chip. I think Im pretty much at the limit of my cooling though. Its only a megahalems rev B with a very quiet fan. Ironically I probably wouldnt have upgraded to my 4770K if these had been available so cheap a year ago as it beats it in quite a few things.
Go right to your max multiplier [22] and then set BCLK to 182. That might be all you need for a 20 second solution. You might need more than default voltage, kinda a lottery on how much voltage you need for 4Ghz.
Beyond 4Ghz it gets much more involved as every 100Mhz necessitates tweaks to voltages, possibly memory timings, etc.
But yeah for a quick and easy 4Ghz I haven't heard of a 5650 that would not do 4Ghz
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,362
hard to be unhappy with the sata speeds
[/URL][/IMG]
Thanks for the quick and dirty! I figure the 4Ghz should help me put off the upgrade itch until next years tic.
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,362
Wouldn't it be nice if apps that are single thread programmed could utilize the power of all 12 threads? I wonder if that would be even possible at some point......we could always dream. it be like never having to upgrade......i guess it does to some degree load balance somewhat. I would by an os that could do that in a hart beat. At least we do well in muddleheaded apps like video encoding and bf4 at least
Zoson, I know we have had this discussion both in private message and in threads at several different forums sites. However, I still wonder to this day if this vcore maximum is not somewhat reduced when going to a 95, 80, 60 or even a 40 watt Xeon part? I've seen first hand that once 1.35v is hit temps drastically rise (especially on the x5650's, x5660's and even the x5670's), even on high end cooling. To my eyes, this tells me a detailed story that these lower watt equivalents, which were probably binned because of their lower inability to handle higher voltages, are not as versatile as the 130 watt binned chips. Once my Xeon is set to 1.35v, or anything above that temps rise like mad (1.25v to 1.35v is a 10C to 20C difference on many of them), and I have one of the worlds best TIM's, an excellent rad and arguably 5 of the best push/pull fans on the market. I think telling people with a x5650, or even the x5660 or x5670 that more than 1.35v is not a very good idea, when nearly 100% of all the information we have to go on is based on enthusiast class 130 watt i7's. Sure, they are damn near identical in every way, but not exactly the same. Intel did have a binning process for these Server chips, and thus they probably had a reason to limit the x5650 to only a 20 multi and 1.35v voltages.
Do we have absolute irrefutable proof that it is safe for any 95 watt or lower Xeon, to venture past the 1.35v vcore spec? Or is this all hearsay based on the experiences of users with 130 watt chips? Again, I'm just going by what I see in the forums. It appears, to me anyway, that these lower wattage chips do not have the same voltage stamina as these 130 watt parts. Is that also your conclusion?
By the way, I finally hit 4.6ghz on my R3E and x5650... and I did so at only 1.343v (at the R3E setting of 1.35v) Using LLC at full per your recommendation has really helped this chip. Thank You!
Zoson
Supreme [H]ardness
- Joined
- Aug 4, 2001
- Messages
- 6,113
Keep in mind that the *transistors* that make up your chip are the same 'revision' transistor as what's in my 130W TDP chip. Aside from wafer quality, there isn't going to be a difference in the tolerance to voltage and heat of the actual gates that make up your CPU.
So, while yes your chip does have a different layout, and thus required a separate set of bins, the actual process to binning these chips(and pretty much all chips) remains the same. The manufacturer picks a handful of TDP values that they set as their targets, and a handful of clockspeed targets that the silicon should be able to reach at those TDP values.
Then, all they do is see which cpus can reach those specified speeds at voltages that keep them under or at the desired TDP value.
This is why, for example, that you can assume an i7 970 will not clock as well as an i7 990x. They share the same TDP target, but the 970 reaches that TDP at a lower clockspeed. What that's basically telling us is that the 970 needs more voltage to reach the higher clockspeed of the 990x, and as a result would have put it over the 130W TDP target.
Now, the thing you have to remember about the 95W vs 130W Xeons is that, in fact, a 95W chip may be able to do the same clockspeed at the same voltages as a 130W part. They're two totally separate bin targets and one chip being in the top bin of 95W is basically the equivalent as being in the top bin of 130W(well, not exactly, but for simplicity sake).
A really GREAT example of this is the original AthlonXP-M CPUs that amd released. They were generally much lower clocked parts, but they also ran at really low voltages. It turned out that those chips typically were the _best_ overclockers you could find at the time.
Really though, I've said it before and I'll say it again. The top end extreme cpus have the absolute best silicon. Which is why you see the craziest overclocks on the X series parts.
TL;DR: I wouldn't expect much difference in overclock-ability between the top bin 95W part and the top bin 130W part. The transistors in our chip are the same, just the layout is different, so the actual electrical tolerance should be the same at the end of the day.
Also, I just wanted to let you know that my time with this platform is about to end. My 5930k system should be up and running by September 21st. I'll still be able to help you, but I will be selling off my parts so I won't be able to provide hard data any more.
Ok Zoson, congrats on the new platform bud. Its always fun moving off into new pastures. Ten years ago I built a new system 2-3 times a year, but since about 2006 I dropped down to once a year and since this R3E, well not at all, its just done everything I have wanted it to. Anyway, I want to be the first to say thanks for all the information you have provided here and over at xtreme systems. And I also want to again thank you for informing me about the crazy price drops on these Xeon parts. I'm sure it was difficult for you guys, well some of you anyway, with these expensive 990X's watching hexa core xeons get so cheap. When I retired in 2011 I pretty much reserved myself to the fact I would never again be able to splurge like I used to be able to. I want x99, or what ever comes afterwards, but my time in the Enthusiast sector (which has been over a 20 year ride for me) may be over. x99 will possibly (or Skylake-E if it exists) be my next system, but only with used parts and not until at least 2017, or maybe even 2016 if I'm lucky. I just hope my R3E has the willpower to last that long, and I pray these Xeons stay available and affordable, just in case... LOL
Take care Zoson. Oh and don't forget to tease us with benchies please
P.S. today I was looking at i7-5960X, 8-core, 16 thread, 64Gb of DDR4 support running at 2133 mhz. INSANE drooling!
wabbitseason
[H]ard|Gawd
- Joined
- Jun 16, 2010
- Messages
- 1,511
I'm thinking of jumping on one of these. Which cheap one (<$200) out there right now will run coolest and stablest at around 4GHz? x5670? Thanks.
x5650, x5660, x5670, x5672, x5675. Those are all 95 watt parts. There are even lower wattage ones, but may not provide the 4ghz you want. If your willing to spend $200, try and find a retired x5675 or x5670. Otherwise, the x5650 provides the best bang for your buck...
hard to be unhappy with the sata speeds
[/URL][/IMG]
lol magic drive.
You guys are making me want to upgrade! I've been running this 920 at 21x200 with HT on a P6T Deluxe v2 (with the 0006 special bios) for a few years now, and haven't felt the need to upgrade yet (except maybe in modded Skyrim, or to reduce power consumption). I haven't tried pushing my board past 200MHz Bclk, I'm already asking a lot from my basic 1333MHz Corsair XMS at 1600MHz (still 9-9-9-24-1T).
So outside of canned benchmarks or certain CPU-intensive programs (video compression, etc), are you guys noticing any improvements in your day-to-day operations? While I would certainly like to improve my Handbrake processing times, I don't use it often. I can see how it would help with some larger multi-player games, but I only play single-player or co-op. I'm just not sure if spending $300-400 on a Xeon and some more DDR3 would be worth it for my usage, as much as I'd like to have some fun again. If there's a clear advantage in my gaming experience or in my day-to-day stuff, then it would be a no-brainer!
Do these Westmere chips have the same issue with even multipliers as the Bloomfields? I know I'd need to flash my bios to accept a Xeon, and thus lose my P6T's ability to lock in the turbo multi. I'd prefer to at least do 21x200, so I'd be looking at an X5660 at minimum.
So outside of canned benchmarks or certain CPU-intensive programs (video compression, etc), are you guys noticing any improvements in your day-to-day operations? While I would certainly like to improve my Handbrake processing times, I don't use it often. I can see how it would help with some larger multi-player games, but I only play single-player or co-op. I'm just not sure if spending $300-400 on a Xeon and some more DDR3 would be worth it for my usage, as much as I'd like to have some fun again. If there's a clear advantage in my gaming experience or in my day-to-day stuff, then it would be a no-brainer!
Do these Westmere chips have the same issue with even multipliers as the Bloomfields? I know I'd need to flash my bios to accept a Xeon, and thus lose my P6T's ability to lock in the turbo multi. I'd prefer to at least do 21x200, so I'd be looking at an X5660 at minimum.
Zoson
Supreme [H]ardness
- Joined
- Aug 4, 2001
- Messages
- 6,113
Was reading some Haswell-EP stuff, and realized intel makes a point clear here that I wasn't really able to support with documentation previously.
I had said, previously, that all cores in Gulftown/Westmere turbo together. Even if only one core is in use, all cores turbo up to the turbo clockspeed.
The proof is in the pudding, individual clock states on a per-core basis is only now supported in Haswell EP:
http://images.anandtech.com/doci/8423/PowerHEp.png
This is possible because of FIVR (fully integrated voltage regulator) that intel's been touting as a large benefit in Haswell.
I had said, previously, that all cores in Gulftown/Westmere turbo together. Even if only one core is in use, all cores turbo up to the turbo clockspeed.
The proof is in the pudding, individual clock states on a per-core basis is only now supported in Haswell EP:
http://images.anandtech.com/doci/8423/PowerHEp.png
This is possible because of FIVR (fully integrated voltage regulator) that intel's been touting as a large benefit in Haswell.
Orion Silverhelm
n00b
- Joined
- Dec 25, 2011
- Messages
- 12
lets say i have the opportunity to buy a 5675 or a 5680 at the same price point, whats to make me choose one over the other? 5675's came out a year after the 5680, and 5680 has slightly higher clock and higher tdp at 130w vs 95w.
If you were me, and you have a solid water cooling set up on an asus rampage 2 extreme running an i7920 at 4.0-4.2 stable for 5 years, which would you go with? Bear in mind, I AM going for a xeon upgrade, I just don't know which one. My goal is to shoot for 4.8-5ghz 24/7 stable on water with hyper threading on supporting 24gb 2000mhz ram.
If you were me, and you have a solid water cooling set up on an asus rampage 2 extreme running an i7920 at 4.0-4.2 stable for 5 years, which would you go with? Bear in mind, I AM going for a xeon upgrade, I just don't know which one. My goal is to shoot for 4.8-5ghz 24/7 stable on water with hyper threading on supporting 24gb 2000mhz ram.
primetime
Supreme [H]ardness
- Joined
- Aug 17, 2005
- Messages
- 7,362
whichever has the highest multiplier almost always gives best results....with that being said there's no guaranty you will hit your goal of 4.8 but its got to help. In my results i can use much lower pll and qpi voltage for lower blck speed. Course somebody posted an incredible deal on 5675s a while back but i haven't searched current ebay prices recently, so if the price was close then might as well go for the higher multi. (5680)