Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mac Pro GeekBench-Whats your score?
- Thread starter MattDSLR
- Start date
- Sort by reaction score
I have an SR-2 Setup with 2 x X5650's, 48GB of Kingston HyperX 1600MHz RAM, Silverstone ST-1500 Watt PSU and GTX 480 GPU. I wanted to find out if you can provide any kind of input with the site you referenced to as I'm having a hard time understanding d00d's information. I can understand some but not all of it. Whatever help you can provide would be great. BTW, I'm only asking you because from what I can see others have tried to ask for help with limited replies as he keeps defaulting to having the individual figure it out themselves. Although it looks like d00d has done a great job, I'm just having a few issues understand some of it...
Also, I'm thinking on possibly getting X5680's or X5690's (and selling my X5650's) and I also see that you under clock your CPUs. Can you provide help in under clocking if I get these CPUs as I don't think it's possible to under clock the X5650's but only the X5680's and X5690's? Thanks for whatever you can do...
Thanks Tutor,
I already have things up and running using myHack and it works fine (but to a point), but I will try DSDT_SR2_3.zip file (I think it's the 3rd version that I should try as it's the latest one). I did want to go over the BIOS setup he has explained to follow:
BIOS settings;
Advanced Settings:SATA Configuration:Configure SATA#1 as [AHCI]
Power Management Features:ACPI Configuration:General ACPI Configuration: Suspend Mode: [S3(STR)]
Power Management Features:ACPI Configuration:Chipset ACPI Configuration: High Precision Event Timer: [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® SpeedStep() tech [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® TurboMode tech [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® C-State tech [Enabled]
I'm gathering this is the setup you're talking about. I've had others give conflicting input about disabling TurboMode and C-State while others said it doesn't matter. I'll use it though.
I asked for all this help because for the last few weeks things have been working fine with myHack and the rest of the install process and I was reaching 3.8GHz on the 5650's but I don't think the 48GB of Kingston RAM was working well with the CPUs, because I started seeing things on Mac windows starting to get "jaggedy" (like 2 frames per second; kind of like an old film - not smooth) in it's operational flow. Example when I use command + arrow up, it separates all the windows that I have open and spreads them apart into one screen. The flow of the animation was very smooth in the beginning, but now is not smooth and very jagged. When I took out all the RAM except for one stick the animation flow was nice and smooth and when I put in 12GB (6 sticks) of RAM the animation started to lose it's nice flow and when I put all the sticks of RAM in it was back to being bad and then after 2 to 3 hours would lock up and I would have to restart the machine.
So I did a MemTest86 RAM test for 12 hours one day and then 16 hours the next and everything passed just fine. But somethings telling me that the pair of 5650's don't play well with 48GB of RAM, but do okay with 24GB. I was thinking of getting of getting this RAM:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820161459&Tpk=wintec 48gb
as it's still 48GB of RAM but only 6 sticks (and not 12). I read somewhere that too many DIMMS may be an issue. But I'm not the expert in that area. But again, this is just my guess and my gut. I had a similar setup with my UD7 Gigabyte Mobo and using a i7 970 (at 4.2 GHz) and 24GB of RAM and it did the same exact thing. As soon as I put in the i7 930 and the Xeon W3680 (same as 980X) and only 12GB of RAM they both cranked at 4.2GHz just fine. That i7 970 for some reason just had issues. I'm thinking maybe it's how it communicates with the RAM. I see d00d had only 24GB of RAM running on his system. So maybe it's got something to do with that. Again, who knows as their aren't that many SR-2 Hackintoshes out there to really compare anything too.
Lastly, you said that you didn't have the 5650's and 5690's but only the 5680's. Great, but I did want to find out how much RAM you have on that system and if underclocking is the best method as I read that you reached 39,000+ in your Geekbench scores. Like I said, if I can get a hold of a pair of 5680's and sell my 5650's then I'd like to try your method of underclocking if you're willing to show me how you made that happen in your BIOS setup. Thanks again for getting back to me and taking the time to show me as much detail as you did.
More on being a paradoxical contrarian and finding a oneness in all things in life
This is the ram that I use - CORSAIR XMS3 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 2000 (PC3 16000) Desktop Memory Model CMX12GX3M3A2000C9 [ http://www.newegg.com/Product/Product.aspx?Item=N82E16820145326 ]. I have been running 48 gigs of it in WolfPack1 that has a pair of 5680s and 24 gigs of it in WolfPack2 that has a pair of 5675s. Note that it's, at least, two steps higher than DDR 1333 (the speed rated for our boards). That's because raising the BCLK also raises the memory speed, requiring ram with greater tolerance ability to run safely at higher speeds. I don't have any experience or other information about the quality of the brand you are looking at. Also, I have no problem with your choosing ECC ram - it does, however, tend to run a little slower than non-ECC ram. However, I can tell you that I would never get far as I have tweaking a system to yield better performance without using ram with the headroom to withstand running at higher speeds, without relaxing the timings. I haven't had any problems with the Corsair ram in either system. So, if you like Wintec ram and want the benefits of using ECC ram, I do recommend that you avoid purchasing DDR3 1333 or 1600 MHz and instead consider DDR3 1866 or DDR3 2000 MHz ram as I have purchased. If you're using a BCLK of about 200, then your current ram is going to be overly stressed, unless you designate it, in bios, as 1066 or 800 MHz ram. Pushing your ram too hard might be the cause of your tearing problem. In sum, I have never experienced that tearing problem. I can designate my ram, in bios, as DDR 1333 ram without any headroom-related problems.
There is no way that I could get the performance that I gotten if I did not underclock/turbobias my systems. I know that underclocking/turbobiasing can be done to 5675s and 5680s because I've done it to mine. I suspect that it can be done with 5690s and other members of the 5600 family, including your 5650's, but since I don't have any these other chips I don't know for sure. As a related aside, I've also done it to single and other dual processor systems with chips preceding the intro of the 5600s.
Please take a look at page one of top Geekbench scores [ http://browse.geekbench.ca/geekbench2/top ]. You'll see the scores achieved by 2, 4, 6, 8, and 12 processor systems (as well as seemingly anomalous scores of 37,900 for a 1 processor/4 core Xeon E5472 running at 2.999 GHz on a Parallels Virtual Platform and of 32,052 for a 1 processor/64-core SunSpark system running at 2.66 GHz). Note the processor speeds for the 2 processor systems. Leaving the seemingly anomalous 1 processor scores aside, all of the two processor scores, except for the 33,066 score (d00d - 3.7 GHz - OSX) and the 33,344 score (Echrei - 4.112 GHz - Windows), are for my WolfPacks. Even the unattributed scores are mine (I use Geekbench2 and Cinebench 11.5 to tune my systems). If you then look at pages 2 and 3 of the top scores, you'll see a trend - higher 24/7 clocked 2 CPU systems yield lower performance returns than my lower 24/7 clocked systems. Wouldn't you think it paradoxical if someone advised you, "If you want a fast computer, then slow it down, but if you want an even faster computer, then slow it down some more, but if you want the fastest computer, then really slow it down when compared to all of the other systems like it? I believe that most people would consider this advice paradoxical. However, as we get older, we learn the value of pacing ourselves and only using energy when there is a need to do so and using only so much energy as is called for under the circumstances. Isn't this like turbo boosting theory?*
For my WolfPack1 and WolfPack2, I use the settings that you've listed above as well as settings along the lines that I have listed in post #70 in d00d's thread. D00d advises us to enable all of the native power management related features of a Hac and to stay within spec VID, i.e., to pace it. Seasoned overclocker say do just the opposite. Is d00d just being a contrary dude? His advice is certainly contrary to what most overclockers advocate. Realizing that what someone has told you is contrary to what many others have told you can create cognitive dissonance->frustration. However, that one gives advice contrary to the crowd's does not mean that the one is wrong. Apple has espoused the philosophy, "Think Different[ly]." D00d's advice, because he had backed the truth of that advice up with his results, forced me to think different and to consider taking his advice a step further than he had done. So, I decided to extend d00d's advice by underclocking my systems and maximizing their turboboosting capability, i.e., to pace them more than any other comparable systems [See, also, URL in sig]. If you think about how low those Sandy Bridge E5s are clocked and how wide their turbo range is extended, I hope that you'll see parallels, truths, and a oneness that I hope that you can extend to enrich all of us.
* Isn't it also like martial arts theory, the saying, "Less is more," and many other things in like?
This is the ram that I use - CORSAIR XMS3 12GB (3 x 4GB) 240-Pin DDR3 SDRAM DDR3 2000 (PC3 16000) Desktop Memory Model CMX12GX3M3A2000C9 [ http://www.newegg.com/Product/Product.aspx?Item=N82E16820145326 ]. I have been running 48 gigs of it in WolfPack1 that has a pair of 5680s and 24 gigs of it in WolfPack2 that has a pair of 5675s. Note that it's, at least, two steps higher than DDR 1333 (the speed rated for our boards). That's because raising the BCLK also raises the memory speed, requiring ram with greater tolerance ability to run safely at higher speeds. I don't have any experience or other information about the quality of the brand you are looking at. Also, I have no problem with your choosing ECC ram - it does, however, tend to run a little slower than non-ECC ram. However, I can tell you that I would never get far as I have tweaking a system to yield better performance without using ram with the headroom to withstand running at higher speeds, without relaxing the timings. I haven't had any problems with the Corsair ram in either system. So, if you like Wintec ram and want the benefits of using ECC ram, I do recommend that you avoid purchasing DDR3 1333 or 1600 MHz and instead consider DDR3 1866 or DDR3 2000 MHz ram as I have purchased. If you're using a BCLK of about 200, then your current ram is going to be overly stressed, unless you designate it, in bios, as 1066 or 800 MHz ram. Pushing your ram too hard might be the cause of your tearing problem. In sum, I have never experienced that tearing problem. I can designate my ram, in bios, as DDR 1333 ram without any headroom-related problems.
SR2Mac2 said:
There is no way that I could get the performance that I gotten if I did not underclock/turbobias my systems. I know that underclocking/turbobiasing can be done to 5675s and 5680s because I've done it to mine. I suspect that it can be done with 5690s and other members of the 5600 family, including your 5650's, but since I don't have any these other chips I don't know for sure. As a related aside, I've also done it to single and other dual processor systems with chips preceding the intro of the 5600s.
Please take a look at page one of top Geekbench scores [ http://browse.geekbench.ca/geekbench2/top ]. You'll see the scores achieved by 2, 4, 6, 8, and 12 processor systems (as well as seemingly anomalous scores of 37,900 for a 1 processor/4 core Xeon E5472 running at 2.999 GHz on a Parallels Virtual Platform and of 32,052 for a 1 processor/64-core SunSpark system running at 2.66 GHz). Note the processor speeds for the 2 processor systems. Leaving the seemingly anomalous 1 processor scores aside, all of the two processor scores, except for the 33,066 score (d00d - 3.7 GHz - OSX) and the 33,344 score (Echrei - 4.112 GHz - Windows), are for my WolfPacks. Even the unattributed scores are mine (I use Geekbench2 and Cinebench 11.5 to tune my systems). If you then look at pages 2 and 3 of the top scores, you'll see a trend - higher 24/7 clocked 2 CPU systems yield lower performance returns than my lower 24/7 clocked systems. Wouldn't you think it paradoxical if someone advised you, "If you want a fast computer, then slow it down, but if you want an even faster computer, then slow it down some more, but if you want the fastest computer, then really slow it down when compared to all of the other systems like it? I believe that most people would consider this advice paradoxical. However, as we get older, we learn the value of pacing ourselves and only using energy when there is a need to do so and using only so much energy as is called for under the circumstances. Isn't this like turbo boosting theory?*
For my WolfPack1 and WolfPack2, I use the settings that you've listed above as well as settings along the lines that I have listed in post #70 in d00d's thread. D00d advises us to enable all of the native power management related features of a Hac and to stay within spec VID, i.e., to pace it. Seasoned overclocker say do just the opposite. Is d00d just being a contrary dude? His advice is certainly contrary to what most overclockers advocate. Realizing that what someone has told you is contrary to what many others have told you can create cognitive dissonance->frustration. However, that one gives advice contrary to the crowd's does not mean that the one is wrong. Apple has espoused the philosophy, "Think Different[ly]." D00d's advice, because he had backed the truth of that advice up with his results, forced me to think different and to consider taking his advice a step further than he had done. So, I decided to extend d00d's advice by underclocking my systems and maximizing their turboboosting capability, i.e., to pace them more than any other comparable systems [See, also, URL in sig]. If you think about how low those Sandy Bridge E5s are clocked and how wide their turbo range is extended, I hope that you'll see parallels, truths, and a oneness that I hope that you can extend to enrich all of us.
* Isn't it also like martial arts theory, the saying, "Less is more," and many other things in like?
Last edited:
My other hobby is driving a car reasonably fast on a race track. As it turns out, the neophyte will always try to drive faster by pushing the accelerator pedal down more. But those of us with experience know that you have to be smooth to be fast. And to be smooth, you have to go slow.
In other words, in order to be fast, you have to be slow.
Now, with that said, dammit, get something working on Mac Pro 5,1s for underclocking!
jas
Well, just to let you know I took your advice and just purchased 4 sets of the Corsair XMS3 12GB RAM kits (totaling 48GBs) today as I was looking at that RAM a few days ago anyway. Thank you for making that decision easier for me. Also, I'm guessing that "WolfPack1" (WP1) and "WolfPack2" (WP2) are what you're calling your PCs.
That was my gut too; that the RAM is just not working well with those CPUs as a possible result of what you just mentioned. Thanks for the clarity. So when I input my info in the BIOS (for the existing X5650 CPUs and the new Corsair XMS3 RAM that I will have soon) I should just leave the Memory Frequency on DDR-1066, right? Also, with the rest of my OC I followed this person's BIOs set up in the first 6 pics that he provided with his "180 BaseClock" Setup:
http://www.evga.com/forums/tm.aspx?m=268280&mpage=1
One thing I noticed is that on the 5th pic the C-State tech and ACPI T State was Disabled. So I don't know (since I never used d00d's setup yet) if that affected anything. If I don't use d00d's setup do I still need to follow his BIOS settings as a basic standard to follow (below)?:
BIOS settings;
Advanced Settings:SATA Configuration:Configure SATA#1 as [AHCI]
Power Management Features:ACPI Configuration:General ACPI Configuration: Suspend Mode: [S3(STR)]
Power Management Features:ACPI Configuration:Chipset ACPI Configuration: High Precision Event Timer: [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® SpeedStep(™) tech [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® TurboMode tech [Enabled]
Frequency/Voltage Control:CPU Configuration:Intel® C-State tech [Enabled]
Or is that just for his setup that works with his custom Kext files?
Quite honestly, I would love to learn as much as I can from you in this area and under clock the HADES out of these CPUs if you can show me. I would take your input and apply it and be your "student" (in a sense), so I can learn to do it the right way once I get those X5680's when they come in a couple of weeks from now. If you have some kind under clocking BIOS set up that you can provide I will try it out and tweak it once those CPUs come in; as I will have the same parts you have to get things working more efficiently and faster. Now I realize that even though I now have the same Mobo, (and very soon I'll have the same) RAM and CPUs I do know that things will be a bit different (as every CPU and other parts) need different tweaking, but I would be very grateful for that info even if I could just try out your current BIOS under clocking setup to see if it works and go from there. I do very much appreciate your time and detail in what you provided so far and want to continue to learn… Thank you…
PS - Just to let you know, I do want learn d00d's setup, but it's a bit hard to understand how to do it properly as I don't want to mess things up and I don't quite understand everything he's explaining and I'm afraid that I'll receive the same response that he has given a few others on there. Plus, I know that he (like all of us) are very busy with other things and this is just something that he wanted to contribute to help those that are on the same level of understanding that he has. Again, if I have a question about his install process and I ask for some clarity (if I get confused), I hope that you could possibly explain it in a different way for me so I can understand what it is he is saying. Again, thanks for your time…
Last edited:
For pickup - See post #18 in thread "All We Know About Maximizing CPU Related Performance."
Latest score on the same 2.26 x 2 Mac Pro 4.1
http://browser.primatelabs.com/geekbench2/1239880
Score = 14063 with more ram this time of 20 gb.
http://browser.primatelabs.com/geekbench2/1239880
Score = 14063 with more ram this time of 20 gb.
Last edited:
Nice! It's pretty impressive what H@cks can get. Recently built up a new rig, X79 based, with a 3930K and a GTX 670, ML 10.8.2 with everything working except sleep. OC to 4.2ghz with very little effort.... GB score attached. I still have a bit of tweaking to do as Mac OS seems to think it's running at 3.8 and gives me a lower score than that the 23500 I get in Win 7 64-bit.
Attachments
You're not alone - Compare pics below. The problem to be solved is how to take full advantage of native power management in OSX with Sandy/Ivy Bridge chips, as is obviously being taken full advantage of under Windows. That's why this six core SB-E scores 27,605 under Geekbench in Win 7 and only 23,317 under Geekbench in OSX, even with the memory clocked down a bit in Windows but every other setting is the same. With d00d's help I cracked this nut with dual 5680 Westmere chips getting a OSX Geekbench 2 score of 40,100 (see URL in sig., below) which is to my knowledge - higher than anyone has yet to get in Windows with Westmeres. I'm still working only a solution for Sandy/Ivy bridge chips.
Attachments
Last edited:
I'm pondering building it back up since I just got most of it back we'll see. If I do it again I'm going to actually build and real water cooling system.
Socket 2011 doesn't have speed step yet correct?
AFAIK, no speedstep as of yet. I'm guessing that's holding us back a bit.
IMPRESSIVE score that 27.6K! What clock speed are you at?
I've considered going water with this rig; I have a CM Evo now and for air cooling it does a remarkable job. The one thing I liked about the WC setup (even a closed loop system like the H100) is the lack of 'clutter' in the case... no huge HSF unit, etc. On the other hand I paid $25 for the EVO and my tower sits in my the slot for a tower in my desk, so no eye candy this time around...
Last edited:
AFAIK, no speedstep as of yet. I'm guessing that's holding us back a bit.
IMPRESSIVE score that 27.6K! What clock speed are you at?
I've considered going water with this rig; I have a CM Evo now and for air cooling it does a remarkable job. The one thing I liked about the WC setup (even a closed loop system like the H100) is the lack of 'clutter' in the case... no huge HSF unit, etc. On the other hand I paid $25 for the EVO and my tower sits in my the slot for a tower in my desk, so no eye candy this time around...
27k is all tutor..
He'd be the man when is comes to creating über performance Westmeres!
I was hoping someone had come up with a way to implement speed step without Apple making a 2011 Mac..I'll be patient then
AFAIK, no speedstep as of yet. I'm guessing that's holding us back a bit.
IMPRESSIVE score that 27.6K! What clock speed are you at?
I've considered going water with this rig; I have a CM Evo now and for air cooling it does a remarkable job. The one thing I liked about the WC setup (even a closed loop system like the H100) is the lack of 'clutter' in the case... no huge HSF unit, etc. On the other hand I paid $25 for the EVO and my tower sits in my the slot for a tower in my desk, so no eye candy this time around...
That Sandy Bridge 6-core CPU is clocked at 4.16 GHz (non-turbo)[see GB11.5 pic], but max turbo is over 4.8 GHz. Best under $1,500 system configuration that I've seen. I expect that I can increase the Geekbench score to beyond 30,000 on OSX when I get native power management in full bloom which will allow me to begin underclocking it. I'll be striving for a turbo ratio of 12h or 18 bins from 2.4 GHz. BTW I'm using an H80 for cooling. Just wish that WolfPackPrime0 had the same potential, then we'd be discussing how to get a Geekbench score of over 90,000 with 4 CPUs.
Last edited:
To your point about 2011 Macs... I do wonder what chipset the new iMacs will be using... and if X79 support will come with that platform. My guess is no since they are still 4 core machines, but you never know what other support Apple will bake into that release (presumable 10.8.3 or .4)
Regardless, lack of sleep and speedstep support is not hurting much, the machine absolutely screams.
I may go for an H100 soon, I have a corsair case made for that cooler.
What did you use to overclock?
______________________________________________
Should have kept reading
What did you use to overclock?
______________________________________________
Should have kept reading
The overclock doesn't work after Leopard, this is my current score under Snowy:
http://browser.primatelabs.com/geekbench2/1247635
Last edited:
