| Index | Recent Threads | Unanswered Threads | Who's Active | Guidelines | Search |
 |
World Community Grid Forums
Category: Community
Forum: Chat Room
Thread: The CPU BEAST Intel i7 970 - 97 results per day |
| No member browsing this thread |
|
Thread Status: Active Total posts in this thread: 30
|
|
| Author |
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges: 
|
Again, the main advantage of the SB is the reduced power consumption. They are not really designed to be high end systems, they just happen to be, relative to previous generations of processor. The high end SB processors will come soon enough, and with a high end price tag to match. Anyway, I see the present 2600's as an alternative to a 920 or similar, but for crunching they are not an alternative to the 6core/12 thread processors. With 970 being roughly 2x as expensive as the 2600, even if you somehow does manage to get 20% more credit/day the 970 is not a good choise in my opinion.  "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges: 
|
It's an easy upgrade from an i7-920 to an i7-970, compared to a full system build. Overall, less expensive.
----------------------------------------1155 boards are only dual channel. Try finding a motherboard for an i7-2600 that can run 2 GPU's at X16. There is only 1 that I know of and it costs around £250. There are no 1155 boards that offer triple or quad SLI at X16. So basically what I am saying is that the CPU is not meant to be in a high end system. Even WRT the power, if you compare the TDP's relative to the performance, there is not that much in it (about 8% more efficient). [Edit 1 times, last edit by skgiven at Sep 30, 2011 12:41:19 PM] |
||
|
|
GTCola2010
Cruncher Joined: Jul 21, 2010 Post Count: 40 Status: Offline Project Badges:
|
Thanks for this conversation! I like to read about hardware reviews.
----------------------------------------This caught my attention (written by skgiven): ... but comparisons are meaningless without operating system, comparable clocks, HT on/off, task types being run... Even then task length and credit variation makes it a ball park comparison. You might need to go to another Boinc project and run identical task types to get really accurate results Skgiven , do you have in mind a Boinc project that has very consistent WU's? I'm thinking about benchmarking a couple of my AM3 chips by swapping them in and out of the same rig after a week or so (same motherboard, ram, running Ubuntu). Thanks! [Edit 1 times, last edit by GTCola2010 at Sep 30, 2011 11:14:41 PM] |
||
|
|
sk..
Master Cruncher http://s17.rimg.info/ccb5d62bd3e856cc0d1df9b0ee2f7f6a.gif Joined: Mar 22, 2007 Post Count: 2324 Status: Offline Project Badges:
|
I do, but different task types perform slightly differently, so elsewhere might not be as relevant.
I think HCC would do for here; the task run times are fairly even, they are short and validate quickly. This makes average credit a reasonable guideline, so long as you run for a day or two before taking values, use average values, compare like for like operating systems, use HT and matched frequencies. I would expect Ubuntu to be a bit faster than Windows, and x64 to be faster than x86. |
||
|
|
GTCola2010
Cruncher Joined: Jul 21, 2010 Post Count: 40 Status: Offline Project Badges:
|
thanks for the tips!
|
||
|
|
Former Member
Cruncher Joined: May 22, 2018 Post Count: 0 Status: Offline |
"It's an easy upgrade from an i7-920 to an i7-970, compared to a full system build. Overall, less expensive.
----------------------------------------1155 boards are only dual channel". . . "So basically what I am saying is that the CPU is not meant to be in a high end system." The i7 970 is a 1366 socket and yes you can go from i7 920 to i7 970 in the same socket. The 1155 are only dual channel and are dessigned as a lower performing chipset. The i7 970 was designed for a high end system as was the i7 920. The socket 1155 is paired with a lower performing chipset overall as it was intented by Intel to be. Given the choice between a i7 970 and a i7 2600k for getting the absolute maximum = number of results and points in BOINC - I think the answer is the i7 970 - no? Ultimately thats what this thread was about. So do we have a concensus? i7 970 beats i7 2600k clock for clock 4 cores vs. 6? BTW, it hasn't hit 97 in a while even around 12 days credit: 09/30/2011 0:011:12:23:46 45,189 73 09/29/2011 0:011:16:54:06 46,531 73 09/28/2011 0:011:01:57:53 44,573 73 09/27/2011 0:012:15:49:58 50,224 89 09/26/2011 0:012:03:39:30 49,559 79 09/25/2011 0:010:22:58:52 43,976 80 09/24/2011 0:009:21:42:57 40,289 64 09/23/2011 0:010:10:58:07 42,562 70 09/22/2011 0:009:09:03:39 37,779 60 09/21/2011 0:010:11:09:37 42,547 56 09/20/2011 0:002:22:45:03 11,330 12 09/19/2011 0:002:10:49:55 9,707 11 09/18/2011 0:006:02:18:10 25,878 40 09/17/2011 0:009:15:38:32 40,801 63 [Edit 3 times, last edit by Former Member at Oct 1, 2011 5:19:48 AM] |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
It's an easy upgrade from an i7-920 to an i7-970, compared to a full system build. Overall, less expensive. Hmm, just did a check, for the same price as only the i7-970, I can buy a i7-2600K + mainboard + 2x 2 GB memory + 320 GB HD + 500 W 80+ PSU... Yes, if you do need a new cabinet, the 2600K is more expensive. But, the combination of i7-920 + i7-2600K will easily outcrunch the i7-970 here at WCG, even at stock speeds against a heavily overclocked i7-970. The only place the i7-970 likely has an advantage is if you're running Folding@Home, due to their very high bonuses for the "big" "bigadv"-wu's... 1155 boards are only dual channel. Yes, this can be a disadvantage in some instances. How much disadvantage I don't know, since I've not seen anyone posting any benchmarks running example CPDN's Hadam3P-models... But, I do know that with my i7-920 the impact is large, so large that in reality the i7-920 performs as a 3.5x core system. Did also do a small test with single-channel memory, but the performance was so bad that didn't even finish the benchmark, even only one task was significantly worse than with triple-channel. So, yes, I'll expect the dual-channel of i7-2600 will hurt it somewhat, so maybe only 3x cores or something. (*) But, while i7-2600 will give a performance-hit by being dual-channel, I seriously doubt the i7-970 six real cores and 6 HT-cores will perform much better than the i7-920 four real cores and 4 HT-cores. Try finding a motherboard for an i7-2600 that can run 2 GPU's at X16. There is only 1 that I know of and it costs around £250. There are no 1155 boards that offer triple or quad SLI at X16. So basically what I am saying is that the CPU is not meant to be in a high end system. Even WRT the power, if you compare the TDP's relative to the performance, there is not that much in it (about 8% more efficient). As a platform for multi-GPU-crunching, the 1366-boards is decidedly better, and if you've got the money for multiple high-end GPU's the extra cost for an i7-970 or i7-980 will only be a fraction of the total system-costs. The benefits of a more expensive cpu is more dubious, since GPU-points will be atleast an order of magnitude higher than possible to get with the cpu... Power-wise, if don't use GPU, using 2 systems will likely be worse than a single system, even if takes into account the higher production of the two systems. How much higher is unknown, and if you'll need to cool the excess heat this can become expensive. (*): Anyone testing this should do this with turbo disabled. "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." |
||
|
|
Ingleside
Veteran Cruncher Norway Joined: Nov 19, 2005 Post Count: 974 Status: Offline Project Badges:
|
Skgiven , do you have in mind a Boinc project that has very consistent WU's? I'm thinking about benchmarking a couple of my AM3 chips by swapping them in and out of the same rig after a week or so (same motherboard, ram, running Ubuntu). Thanks! SIMAP is an excellent choise, with their short wu's (around 1 hour depending on cpu-speed), and with only a few distinct groups of wu-sizes (you'll see this based on various expected run-times before wu's has started). For benchmark-purposes only running for a couple hours should be enough to get a good indication of performance, by looking on cpu-times. The only disadvantage with SIMAP is they're normally only having work the start of each month, ocassionally not even enough for 2 full days of work. They've currently got work, and expects to have for atleast 1 week... If you're interested in testing the whole system, not just the cpu, your choise is CPDN. Depending on model-type, running multiple CPDN-models can heavily inpact performance of eachother, showing that your so-called 12-thread system really isn't performing any better than a 4x core system or something... CPDN-models takes around 4 days for the short Hadam3P-models you should use for benchmarking-purposes. FAMOUS and HADCM3N is significantly longer, but doesn't have same impact on performance so isn't as good as a benchmark. The thing to look for is seconds per TS, a CPDN-model will send a Trickle-message after 12-24 hours, and s/TS is displayed on web-page after you've uploaded a trickle for a model. CPDN has also problems with work-supply, but for the moment they've got some Hadam3P-EU-models available that should be good for benchmarking-purposes. Oh, and please note, you'll currently need to wait 1 hour between work-requests, so it can take some hours to fill-up a 12-core system with enough work for all cores.  Oh, and while a CPDN-model can take 4 days or something to run, the deadline is multiple months, so after getting some benchmark-results letting them continue in the background shouldn't be a problem. "I make so many mistakes. But then just think of all the mistakes I don't make, although I might." [Edit 1 times, last edit by Ingleside at Oct 1, 2011 1:22:54 PM] |
||
|
|
Malarik
Cruncher Joined: Jul 3, 2011 Post Count: 5 Status: Offline |
Hello fellow crunchers,
I agree that the Hexacores (& my 980x) are real "CPU BEASTS"!  Since first joining the Grid, I have added computer after computer since July and have found that my older hexacore, the 12 thread i7- 980x, seems to do the best crunching overall. I did build a brand new i7- 2600k system from scratch with top the line stuff from Microcenter...but for some reason, the motherboard refuses to Overclock. Due to AC cooling concerns (I live in toasty Central Valley CA), the electric bill, over-heating issues, stability, and other concerns, I stopped OC all three of my PCs. The i7-980x > i7- 2600k > i7- 965x (overall, without OCing) I also have a i7 Imac, an older imac, and 2 minimacs running. The dual-cores are slow, but they do add up over time. Even though I started with just the i7 Mac in July, as I brought more and more systems "on-line", the hexacore exceed the Imac in a rather brief time. The 2600k system cost about $1.5k to build and it does work great, but the extra cores of the 980 (a Cyberpower build) make a big difference (but yes, it was a $1000 processor).  I stopped OCing all my systems to improve efficiency and to not roast me nor my computers.  I also found that replacing the CPU heatsink paste, after a year, really made a huge difference in cooling my 980. At OC, it was almost at the boiling point of water! I would recommend that many of you consider replacing your paste every 6months to a year. The water cooling (nor other cooling) works without a good interface. For better or worse, I just used the generic Radio Shack paste. I also have software that monitors the temps on all my systems, but they seem to do OK, even when my home office reaches 90 degrees in the air. It would stink to cook a $1k processor.  Between the 7 systems (plus the infrequent use of a Powerbook while away), I am doing about 30-40 days of CPU time a day. I have been devoting almost all that power to the HCC project, and get over 700 results per day for now (and am happy to have joined the "2000" result club, in just 4 months). I really like watching the i980 crunch away on 12 threads in almost "real-time".  In conclusion, the Hexacores such as the i7 970, 980, & 990 are real "beasts"! BTW- I am very new to this, but I am happy to join such a great bunch of people who are doing great things regardless of race, country, or processor. Anyway, Happy Crunching & Helping Mankind!!!! |
|||
|
||||
|
Malarik
Cruncher Joined: Jul 3, 2011 Post Count: 5 Status: Offline |
Oh, in addition, here is the device stats on each. After looking at this, without OCing, it sure seems that at least in my systems, the i7- 2600 is much more cost effective. Since I already bought the 980 when it first came on the market, and had it before crunching... the cost is not too much a factor in my case.
The 980 : 10/01/2011 0:011:23:31:21 43,692 213 09/30/2011 0:012:00:05:12 43,866 205 09/29/2011 0:011:02:18:08 40,151 198 09/28/2011 0:011:06:19:59 40,546 200 09/27/2011 0:010:15:51:59 38,805 192 09/26/2011 0:011:05:34:27 39,527 202 09/25/2011 0:010:21:51:55 39,304 198 09/24/2011 0:011:15:42:59 42,279 211 09/23/2011 0:011:13:43:53 41,392 208 09/22/2011 0:012:04:40:17 45,520 220 09/21/2011 0:010:14:00:38 38,792 191 09/20/2011 0:010:22:56:20 40,426 198 09/19/2011 0:012:00:18:02 43,622 217 The i2600k : 10/01/2011 0:008:19:08:58 44,620 202 09/30/2011 0:007:09:19:09 37,160 170 09/29/2011 0:007:06:08:42 36,886 168 09/28/2011 0:008:16:17:23 44,503 204 09/27/2011 0:007:12:24:50 38,348 176 09/26/2011 0:006:22:09:54 35,780 163 09/25/2011 0:008:10:12:00 42,892 197 09/24/2011 0:007:17:45:46 39,795 182 09/23/2011 0:008:15:15:35 43,617 202 09/22/2011 0:008:04:54:03 40,594 193 09/21/2011 0:011:07:25:14 53,184 261 09/20/2011 0:007:05:59:10 34,532 168 09/19/2011 0:007:12:21:23 37,741 176 BTW- If you have (or want to build) ordinary consumer systems like mine, get a Gold or Platinum Power Supply, good water cooling, a good roomy case with lots of good fans, etc... and you will save on the electricity. A "Kill-A-Watt" device (and real-time software temp monitoring) is also useful to see what you system "eats" with different components, level of OCing, or your advanced settings. A cool computer is a happy computer! |
|||
|
||||
|
|