PDA

View Full Version : To Hyperthread...or to Not Hyperthread...that is my question



HyperMatrix
02-22-2013, 04:05 PM
So in trying to get my 5GHz OC stable, I've been trying "everything." I didn't think it would help but I remember reading that disabling Hyper Threading can often help with higher overclocks. So I disabled it. And so far (fingers crossed) everything is fine. Been running prime for 10 minutes. Temperatures are lower by 10c! They're between 69c-77c per core. Intelburntest numbers are faster by around 17%.

So...what the hell? Do I want to keep working on stabilizing with Hyper Threading on? Or do I take the MUCH lower temperatures and so far completely stable OC at 5GHz? I remember reading benchmarks where they showed HyperThreading was good for some games, and bad for others. What do you guys think I should do? Let's say I keep trying to stabilize 5GHz and I can't...and I have 2 options:

1) 4.9GHz with HyperThreading
or
2) 5GHZ with no HyperThreading

TheZone
02-22-2013, 04:33 PM
I dont what to do Hyper haha maybe Sneaky will like to.

HyperMatrix
02-22-2013, 04:49 PM
What's also weird is that while Intelburntest and Prime95 are both reporting back as being stable, doing cpu intensive video encoding causes virtual dub to crash. Even if it's just 60-70% usage and no higher than about 70c on the cores. Really gotta figure that one out...

EniGmA1987
02-22-2013, 05:26 PM
I think it is more a matter of if what you are doing benefits more from the additional threads at slightly slower speed or not. For instance, if you did a lot of video encoding then even a 200MHz drop per core would be outweighed by having twice as many processing threads. For gaming, it usually doesnt help right now although there are a few exceptions.

whitespider
02-22-2013, 07:13 PM
Hyperthreading is great for encoding. About a 25% boost at the best of times. Gaming is another matter entirely. I have found windows 7 (and probably 8 if I actually used it) performance to be better as well. Snappier.

Switch to hyperthreading when you want to encode, and standard when you want to game. Your bios should have a preset fuction that you can save a stable overclock for both.

4.9ghz vs 5ghz is about 5-10fps in the best case scenarios. Which is often more than hyperthreading will give games. (It detracts as much as it adds)

winterhell
02-23-2013, 03:14 AM
4.9ghz vs 5ghz is about 5-10fps in the best case scenarios

2% clock speed increase results in less than 2% FPS increase.

waperboy
02-23-2013, 04:02 AM
To help putting things in perspective, remember that hyperthreading doesn't really give you an extra full thread, it just means that while the core is. say in an IO wait-state, it can execute some instructions for another thread. It may potentially even hurt performance in some cases, since the "second thread" can tear up the cache and fight for the core execution logic, of which there is only one instance.

HyperMatrix
02-23-2013, 04:34 AM
Well I was looking around and saw benchmarks for x264 encoding showing that a 5GHz 3570k is just barely faster than a 4.8GHz 3770k. So that should mean there is a 4% performance increase to be had for encoding with that codec, if comparing 5GHz HT to 5GHz no-HT. However, of that expected 4% difference, would any of it be attributed to/alleviated by the 8MB cache on the 3770k as opposed to the 6MB L3 cache on 3570k?

I'd love to get 5GHz with HT running. But keep running into instability. Buying new ram tomorrow, but I ran memtest86+ (for only an hour) and received no errors on my existing sticks so I think it's a CPU instability issue and not a memory issue as i had hoped. Been running prime for 8 hours now with no issues (HT disabled) and max temps have hit 74, 81, 81, 76 in that time. Current temps with test running is 68, 75, 74, 67.

Maybe I should research into why hyperthreading has such a hugely negative effect on overclocking and how to deal with it. On the plus side I know I'm 100% guaranteed 5GHz with no HT so that's a bit of a relief.

Shadman
02-23-2013, 04:41 AM
Like I was saying in another thread, if things are crashing then you aren't stable. Prime, Intel, memtest, all are a little synthetic, and while they are really good, sometimes with the algorithm they use, it takes a long time to get to find it. I don't know how long you are letting it run, but run it long enough and you will eventually find something.

When you run applications, they interface a little differently than those previous programs. I think you should wait for the ram to come in before determining anything though, just in case

HyperMatrix
02-23-2013, 05:07 AM
Like I was saying in another thread, if things are crashing then you aren't stable. Prime, Intel, memtest, all are a little synthetic, and while they are really good, sometimes with the algorithm they use, it takes a long time to get to find it. I don't know how long you are letting it run, but run it long enough and you will eventually find something.

When you run applications, they interface a little differently than those previous programs. I think you should wait for the ram to come in before determining anything though, just in case

That's what I'm saying. I'm 100% fully stable with HT off. In games. Benchmarks. And the most finicky one of all, Video Encoding! (which surprised me, actually). But I had to disable HT to get that stability. With HT enabled, I'd get a lot of:


"A corrected hardware error has occurred.Reported by component: Processor Core
Error Source: Corrected Machine Check
Error Type: Internal parity error
Processor APIC ID: 4
The details view of this entry contains further information."

Along with crashes during video encoding (but not during gaming or other work), and the "occasional" crash while heavily stress testing. By crash I mean the application would crash. Not the entire PC. No matter what I've tried (with my limited knowledge + googling), I haven't gotten complete 100% stability with HT on. I guess I'll have another go at it with the new memory. Obviously I'd prefer to have HT active. But even if not, guess HT-Less I shall be. Don't want to drop below the nice 5GHz badge of honour. :P

p.s. I'm in trouble. My local computer shop just added the GTX Titan to its inventory at $999.99. The price I said I'd buy it at if it became available locally. So now I have the option of pre-ordering 3 online at $1019.99 each. Or...risking it and not pre-ordering, and then having my local shop not have 3 units available for purchase. When the GTX 680 initially launched, for example, they only received 2 units.

HyperMatrix
02-23-2013, 05:07 AM
COMPUTERS ARE HARD!

Super sad face :(

Cpt.Teacup
02-23-2013, 05:11 AM
Did the previous owner use HT?

HyperMatrix
02-23-2013, 05:15 AM
Did the previous owner use HT?

Yeah. On an Asus Maximus IV Extreme which is the Z68 version of my Maximus V Extreme. I'm just really hoping it somehow has something to do with the memory. I honestly had never realized how much of a performance increase you can get by going from 1600MHz to 2400MHz+. Makes me sad I don't have an X79 quad channel setup right now. But the current 50% increase in memory bandwidth should help when I get the new ram.

waperboy
02-23-2013, 06:16 AM
COMPUTERS ARE HARD!


Gamers are soft.

Shadman
02-23-2013, 10:08 AM
Id talk to the manager of the computer shop and ask for three directly from him :p

Crackatoah
02-23-2013, 01:00 PM
Well it depends on what games or programs you are using doesn't it.

If your playing modern games like Crysis 3 and BF3, or most future games. Then disabling Hyperthreading for 5Ghz would be pointless and detrimental to performance because you would get allot more performance with 8 threads and 4.8Ghz for example

If your playing games that use 4 cores or less then I suppose you can disable it but I doubt you will gain much at all by running at 5Ghz as apposed to 4.8 which I guess you can get with Hyperthreading on easily.

The only logical reason for disabling HT to get 5Ghz would be... so you know its running at 5ghz? and taking a performance hit in some stuff and not really maming a difference in others. And the added bonus you know the chip is probably degrading quicker too lol.

Shadman
02-23-2013, 01:16 PM
The only logical reason for disabling HT to get 5Ghz would be... so you know its running at 5ghz? and taking a performance hit in some stuff and not really maming a difference in others. And the added bonus you know the chip is probably degrading quicker too lol.

But of course! 4.9 is not 5. You can't round up ;)

And chip degration? You mean that thing that happens when I already sold my CPU after I get the next 1 or 2 gen processor?

Crackatoah
02-23-2013, 02:40 PM
But of course! 4.9 is not 5. You can't round up ;)

And chip degration? You mean that thing that happens when I already sold my CPU after I get the next 1 or 2 gen processor?

What do you mean with the 4.9?

And on the second point, learn to not take everything seriously lol :p

winterhell
02-23-2013, 02:46 PM
What do you mean with the 4.9?

It means that 5.0GHz is a cute round number that gives people some peace of mind. 4.9 seems much lower than 5.0 at first glance, not unlike how all the prices in the store end on the numbers 99 or 95 to trick you into thinking the price is not that high.

Crackatoah
02-23-2013, 02:52 PM
Oh right, well I never said 4.9 I said 4.8 thats why I was confused lol.

Shadman
02-23-2013, 04:02 PM
4.8 is even further from 5.0! Blasphemy!

And I wasn't taking it seriously, I'm saying real overclockers don't give a hoot about degration, that only slows us down >:D

Cpt.Teacup
02-23-2013, 04:57 PM
Yeah. On an Asus Maximus IV Extreme which is the Z68 version of my Maximus V Extreme. I'm just really hoping it somehow has something to do with the memory. I honestly had never realized how much of a performance increase you can get by going from 1600MHz to 2400MHz+. Makes me sad I don't have an X79 quad channel setup right now. But the current 50% increase in memory bandwidth should help when I get the new ram.

It might be a good idea to try the overclock on the original motherboard with the previous owner's settings just to see.

Shadman
02-23-2013, 05:41 PM
But the RAM will still be different. At least, I didn't hear about the RAM being part of the deal too

Crackatoah
02-23-2013, 06:06 PM
4.8 is even further from 5.0! Blasphemy!

And I wasn't taking it seriously, I'm saying real overclockers don't give a hoot about degration, that only slows us down >:D

I was joking about the degradation part hence the smiley face lol.... like he cares.

What im saying is, overclockers you overclock for performance, not a number.

Yes sure, overclock for a number for a benchmark to get a high score, but overclocking 24/7 to 5ghz and getting worse performance than 4.8ghz (with HT on) for example is stupid lol

HyperMatrix
02-23-2013, 06:56 PM
No need to fight guys. I may have resolved it. New Ram + New Offset voltage and no encoding crash yet. Passed intelburntest (high heat). Will run prime95 later to monitor heat and extended performance. But seems to be doing well so far. :) Also, just pre-ordered 3 Titans.

And yes, to me encoding crash is the ultimate test of stability. Because cpu's that pass prime95 and intelburntest can still fail in that department.

Sneaky
02-24-2013, 08:31 PM
That's a nice OC there Hyper, If I wasn't away all weekend, I could have told you it was most likely your ram :p.
More often than not CPU OC limits are a cause of ram, some MB's do not like all their DIM slots
populated, also some MB's can be fussy with what ram you use.

HyperMatrix
02-24-2013, 08:39 PM
Well I'm going to do a video on proper OC stability testing because I realized the CPU has automatic error correction that sometimes works and slows you down and sometimes fails. So when people say some games aren't overclock friendly, it just means they have an unstable OC. I ended up finally setting it to 4935mhz with memory at 2520mhz. It's close enough for me to be happy with it. Though I will consider getting the swiftech h220 to see if extra cooling will somehow let me get that extra 65mhz. Doubtful. But good excuse to get a top notch cooler. :)

Sneaky
02-24-2013, 08:40 PM
Yeah when ya cranking it up that high a proper WC set up will help ;)

Cpt.Teacup
02-24-2013, 11:20 PM
You could also try replacing the TIM on your VRM's. Make sure you're using all of the power headers, they help stabilize the power.

Crackatoah
02-25-2013, 01:05 PM
Well I'm going to do a video on proper OC stability testing because I realized the CPU has automatic error correction that sometimes works and slows you down and sometimes fails. So when people say some games aren't overclock friendly, it just means they have an unstable OC. I ended up finally setting it to 4935mhz with memory at 2520mhz. It's close enough for me to be happy with it. Though I will consider getting the swiftech h220 to see if extra cooling will somehow let me get that extra 65mhz. Doubtful. But good excuse to get a top notch cooler. :)

Word of warning don't bother getting the swifttech h220, it looks like it should be good but very surprisingly its actually no better than a H100i which in turn isn't much better than a regular air cooler such as the D-14/K2/SA.

People aren't really sure why it doesn't perform that well compared to the regular all in ones, maybe its the CPU block or pump its using or just a combination of the unit.

Get a proper loop :D

HyperMatrix
02-25-2013, 01:19 PM
Word of warning don't bother getting the swifttech h220, it looks like it should be good but very surprisingly its actually no better than a H100i which in turn isn't much better than a regular air cooler such as the D-14/K2/SA.

People aren't really sure why it doesn't perform that well compared to the regular all in ones, maybe its the CPU block or pump its using or just a combination of the unit.

Get a proper loop :D

Well I wanted to get the best performance possible with a simple all in one kit. All reviews and forum posts I've seen on it have been positive. Where are you seeing the negative reports? I placed a preorder for a unit based on benchmarks on a couple sites, and positive forum posts from people on OCN and others. Links to reported shortcomings/issues would be appreciated. Because all the benchmarks I've seen show anywhere from 3c-10c lower temps compared to the h100i. And as it has a better pump, I think there should be some nice gains if you switch out the fans that come with it to bigger, louder, and meaner fans. Because clearly the ones they use are low-rpm fans meant for the noise sensitive individual. I'm also hoping I'll be able to not only switch out the fans, but go with a push/pull config. It's a great high quality 240mm radiator. And it has an amazing pump. Since I'm not using it for a full-system cooler, and it's just to service the CPU, it *should* be as good as it gets with push/pull config, no?

Update: Just saw the stock fans that come with the Swiftech are 54cfm. I'm using fans with twice that output right now on my H100. That should mean there's some real good room for upgrade. :)

Crackatoah
02-25-2013, 04:56 PM
http://www.youtube.com/watch?v=aPfLKkyIJOI&list=UU_SN80_V2GymyCWM2oTYTeg&index=1

Is also within a few degrees of other cheaper units.

If I was you I would just get an actual loop, it really doesn't take much time to install it and if your running such high end stuff overclocked to the max I would think its worth it and a good investment.

whitespider
02-25-2013, 05:36 PM
http://www.youtube.com/watch?v=aPfLKkyIJOI&list=UU_SN80_V2GymyCWM2oTYTeg&index=1

Is also within a few degrees of other cheaper units.

If I was you I would just get an actual loop, it really doesn't take much time to install it and if your running such high end stuff overclocked to the max I would think its worth it and a good investment.

I like this guy. But he bitches about EVERYTHING that's not his specific part-to-rule-them-all. That said, I agree with most of what he's saying in that review.

Crackatoah
02-25-2013, 05:59 PM
I like this guy. But he bitches about EVERYTHING that's not his specific part-to-rule-them-all. That said, I agree with most of what he's saying in that review.

I also agree, he is way too picky and and always very strongly states his opinion, but his testing and actual reviews are always very good to go by.

Sneaky
02-25-2013, 07:23 PM
Those H220's aren't bad but I am more of a fan of the XSPC rasa kits, they are more customizable.
You can pretty much use any Radiator/CPU block you like.

Cpt.Teacup
02-25-2013, 08:18 PM
At this point I would go with a real water loop, too. I think that above $100 AIO coolers are a waste of money.

HyperMatrix
02-26-2013, 03:38 AM
At this point I would go with a real water loop, too. I think that above $100 AIO coolers are a waste of money.

I feel bad doing it as I should have gotten a maximus v formula if I were going full and proper water cooling so I could have the motherboard cooled for stability as well. Having a hard time deciding. Partially because I'm completely unaware of what's needed to set up my own loop. So there's that over-complexity fear factor. Have any specific units I should look at and consider? I have a cooler master cosmos 2 ultra case.

Sneaky
02-26-2013, 04:51 AM
IIRC the Cosmos II can fit a 360 Radiator up the top of the case.
I would recommend the XSPC Raystorm D5 EX360 Kit
http://www.pccasegear.com/index.php?main_page=product_info&cPath=207_160_45&products_id=20463
875


Not too expensive and easy to put together, you could also add GPU's to the loop if you wanted and also an extra Radiator
in the future if you wanted.

HyperMatrix
02-26-2013, 04:56 AM
IIRC the Cosmos II can fit a 360 Radiator up the top of the case.
I would recommend the XSPC Raystorm D5 EX360 Kit
http://www.pccasegear.com/index.php?main_page=product_info&cPath=207_160_45&products_id=20463
875


Not too expensive and easy to put together, you could also add GPU's to the loop if you wanted and also an extra Radiator
in the future if you wanted.

Perhaps this could be my first foray into the world of water cooling. Anyone else vouch for this model or recommend a different one?

Crackatoah
02-26-2013, 12:11 PM
+1 on the XSPC 360 kit, great value for money and good for a beginner, everything you need.

HyperMatrix
02-27-2013, 07:43 PM
IIRC the Cosmos II can fit a 360 Radiator up the top of the case.
I would recommend the XSPC Raystorm D5 EX360 Kit
http://www.pccasegear.com/index.php?main_page=product_info&cPath=207_160_45&products_id=20463
875


Not too expensive and easy to put together, you could also add GPU's to the loop if you wanted and also an extra Radiator
in the future if you wanted.

Just placed orders for this, coolant additive, silver killcoil, and 4 of those ridiculous $38 Delta fans. Was going to get 6 but Newegg Canada only had 4 in stock at the time. Figure I need to take my first baby step into water cooling.

Crackatoah
02-27-2013, 07:53 PM
Please cancel your order on the delta fans lol. One of the great things about water cooling with decent sized rads is that you can use quiet fans and get great temperatures, the best fans for performance was well as being quiet are the syth 1850 rpm's. Theres not a hole lot of benefit in terms of degrees getting the lawnmower sounding deltas compared to them, you could count the temperiture difference on 1 hand minus the thumb, but you could count the noise difference with the loss of 1 ear LOL

Sneaky
02-27-2013, 08:19 PM
Nice Hyper .... prety sue you will be hapy with the perfomance.
Those fans will be good your 4 GPU fans will mke more nose than the delta's

HyperMatrix
02-27-2013, 08:21 PM
Nice Hyper .... prety sue you will be hapy with the perfomance.
Those fans will be good your 4 GPU fans will mke more nose than the delta's

Well I figure if I have all 6 of them, don't need to run them at max. But for some benchmark/crazy testing i'll have the option to crank it up. =D Also, fan controller? What should I get. I made sure to pick up PWM versions of the delta fans.

Sneaky
02-27-2013, 08:28 PM
Do you want to control ALL your case fans or just the 6 on the rad?

HyperMatrix
02-27-2013, 08:48 PM
Do you want to control ALL your case fans or just the 6 on the rad?

Just the 6 on rad.

Sneaky
02-27-2013, 09:04 PM
Well you have two options:
1) Have a separate controller dial for each of the six fans
2) Run the fans speed control in tandem (two fans on each controler dial).

I think number two is the way to go as each set of fans will throttle the same, in this instance you really only need
a controller that will accomodate three fans as opposed to one that has 6 separate channels.

Now, if you do this, you need a PWM adapter cable that will join two fans into one connection:
http://www.pccasegear.com/index.php?main_page=product_info&cPath=19_1344&products_id=17917

You will need a controller that can accomodate two fans on the one channel, I have a fan controller that came with my
PSU and it doesn't like two fans on the one channel, it greatly decreases the adjustment of speed.

Here is a list of lots of different controllers, you should be able to find each one of these on newegg.
http://www.pccasegear.com/index.php?main_page=index&cPath=9_511

I prefer a slider or dial type as opposed to the touchscreen ones but that's my preference.

I quite like this one, it has an LCD so shows temps and fan RPM but uses dials for control (accepts 30w per channel):
http://www.pccasegear.com/index.php?main_page=product_info&cPath=9_511&products_id=17216

HyperMatrix
02-27-2013, 10:53 PM
Well I'd need it to be a 6-channel kit as the fans I'm getting are rated at 24w each. So I can't split 2 of them over a 30w channel. So I'm thinking about this simple 6-knob one: http://www.newegg.ca/Product/Product.aspx?Item=N82E16811995075

Only one I can see on Newegg.ca that has 6 channels with enough wattage to cover my fans, and that is PWM based, not RPM based. Ideally I'd have loved to have 3 fans controlled per knob (so the 3 push fans, and 3 pull fans). But doesn't seem like it's an option. So I'm going to have to go individual controls. And all the LCD based ones are either 5-fan models, or 6-fan but at lower wattage than what mine need.

Sneaky
02-27-2013, 11:13 PM
Haven't heard of that brand before but it should do what you want it too,
pitty your fans are 24w or there are not any controllers with a higher wattage per channel

HyperMatrix
02-28-2013, 02:39 AM
Haven't heard of that brand before but it should do what you want it too,
pitty your fans are 24w or there are not any controllers with a higher wattage per channel

Yeah. Apparently they are the highest static pressure fans in 120mm x 25mm format. Something like 5-6x stronger than the fans that come with the h220 I was going to buy. So I'm hoping this kit will be enough even after a few CPU upgrades. Maybe in the future add in a second radiator and go 3 of these fans per radiator. For example if I decided to water cool the titans.

Crackatoah
02-28-2013, 05:16 AM
Those delta fans aren't very good a variable speeds, they are a bit jumpy an not accurate so some may be faster than other ect if you turn the seed down.

Also even when at a low rpm they are pretty loud, at high rpm they will be WAY louder than all 4 of your gpus fans running at 100% (literally like a Hoover at full pelt, if not louder)

Cpt.Teacup
02-28-2013, 08:41 AM
What model fans are we talking about?

HyperMatrix
02-28-2013, 10:15 AM
What model fans are we talking about?

The super ridiculous Delta FFB1212EH-pwm. Will be great for benchmarks. My 3770k boots into windows at 5250-5300MHz. Ran a bench at like 5214 or something around that. This is with the side open and a large fan blowing to cool down the vid cards, mobo, and CPU. Since I had voltage up to around 1.6v. Lol. But it stayed cool. :P

On a side note....I hate how good sb-e chips are for benchmarks. Even though they're not as good for 90% of current gen games. :( so even though my system is designed for the fastest gaming possible, that's not represented in benchmarks.

Fimconte
02-28-2013, 06:11 PM
I'll trade you my 2600k for your 3770k ? ;D

HyperMatrix
02-28-2013, 06:15 PM
I'll trade you my 2600k for your 3770k ? ;D

Sb-e! Not sb! :P

frunction
03-02-2013, 01:51 AM
I just installed my single Titan, actually getting slightly better framerates than my two 680s @1080p. The minimum fps stays higher and it is much smoother. I wasn't expecting such good results...

HyperMatrix
03-02-2013, 02:06 AM
I just installed my single Titan, actually getting slightly better framerates than my two 680s @1080p. The minimum fps stays higher and it is much smoother. I wasn't expecting such good results...

Yeah I was really surprised. It just "feels better." What can I say. :P I currently hold the #1 Single-Card Unigine Valley bench on OCN. I go to some extreme lengths to get my numbers though...I've also got 2nd place in both Unigine Valley and Unigine Heaven. Both being beat out by a Quadfire 1350MHz setup on a loop which the owner claims cost more than $3000. And also on a 6-core cpu instead of little 4 core.

883

My only issues are with 1 card I have that's a little bit retarded....I'm either going to return or exchange it. The other issue is that these cards really heat up a lot and I find the fan on them to be less effective than the one on my GTX 680. Could also be the nearly double-sized die. :P