New Sandy Bridge Computer Help (Built - WOW!)
 

Sandy Bridge CPUs are now available at newegg! I want to build a new computer with this CPU. Goals: modest overclock and cheap. Intelligent input is needed - I don't follow all the latest hardware like I used to.

I made a first guess as to what I'll need below.

Questions:

1) What memory is best? There are hundreds of DDR3 choices with different CAS ratings, speed, voltages, etc.

2) Is the power supply good enough to power an Nvidia GPU?

3) Can I build it using the on-chip GPU and add the Nvidia GPU later (they might be cheaper by the time I'm done with AVX programming and ready to try GPU coding)?

4) Is the Intel heatsink junk? Is there a better heatsink choice?


[CODE]
OCZ 4GB (2 x 2GB) 240-Pin DDR3 SDRAM DDR3 1600 (PC3 12800) Desktop Memory Model OCZ3F1600LV4GK
Item #:N82E16820227507
Return Policy: Memory Standard Return Policy
$20.00 Mail-in Rebate Card
$64.99

HITACHI Deskstar 7K1000.C HDS721010CLA332 (0F10383) 1TB 7200 RPM SATA 3.0Gb/s 3.5" Internal Hard Drive -Bare Drive
Item #:N82E16822145304
Return Policy: Standard Return Policy
-$15.00 Instant
$69.99
$54.99

Antec EarthWatts EA650 650W Continuous Power ATX12V Ver.2.2 / EPS12V version 2.91 SLI Certified CrossFire Ready 80 PLUS Certified Active PFC "compatible with Core i7/Core i5" Power Supply
Item #:N82E16817371015
Return Policy: Standard Return Policy
-$30.00 Instant
$99.99
$69.99

Rosewill R102-P-BK 120mm Fan MicroATX Mid Tower Computer Case
Item #:N82E16811147111
Return Policy: Standard Return Policy
$29.99

SONY Black SATA DVD-ROM Drive Model DDU1681S-0B
Item #:N82E16827118031
Return Policy: Standard Return Policy
$16.99

Thermaltake Silent 1156 CLP0552 92mm CPU Cooler For Intel Socket LGA1156
Item #:N82E16835106139
Return Policy: Standard Return Policy
$25.87

GIGABYTE GA-P67A-UD3 LGA 1155 Intel P67 SATA 6Gb/s USB 3.0 ATX Intel Motherboard
Item #: N82E16813128460
Return Policy: Standard Return Policy

Intel Core i5-2500K Sandy Bridge 3.3GHz (3.7GHz Turbo Boost) LGA 1155 95W Quad-Core Desktop Processor BX80623I52500K
Item #: N82E16819115072
Return Policy: CPU Replacement Only Return Policy
-$20.00 Combo
$354.98
$334.98
Grand Total: $597.80

[/CODE]

[QUOTE=Prime95;245402]
[CODE]
Thermaltake Silent 1156 CLP0552 92mm CPU Cooler For Intel Socket LGA1156
Item #:N82E16835106139
Return Policy: Standard Return Policy
$25.87
[/CODE][/QUOTE]

I haven't kept up much with the hardware either, so I can't answer your questions, but I think I can help avoid a possible error. I think Sandy Bridge will use socket LGA1155. The cpu cooler you quoted above says it is for a socket LGA1156. This one probably won't work for your cpu. It doesn't look like newegg or tigerdirect have a cpu fan for socket LGA1155.

I watched youtube videos from newegg earlier today. My understanding is that LGA1156 HS&F fit onto LGA1155.

I think if you overclock (past turboboost max) you need to use an off-CPU GPU.

[URL="http://www.youtube.com/user/newegg?blend=1&ob=4#p/search/1/JGxMdwJ1Rjk"]http://www.youtube.com/user/newegg?blend=1&ob=4#p/search/1/JGxMdwJ1Rjk[/URL] (iirc!)

You can't use on-die GPU if you want to overclock:

[URL="http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/6"]http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/6[/URL]
[QUOTE]The H67 chipset has an FDI so you can use the on-die GPU, however it doesn’t support CPU overclocking—only memory. What about those users who don’t need a discrete GPU but still want to overclock their CPUs? With the chipsets that Intel is launching today, you’re effectively forced to buy a discrete GPU if you want to overclock your CPU. This is great for AMD/NVIDIA, but not so great for consumers who don’t need a discrete GPU and not the most sensible decision on Intel’s part.[/QUOTE]

Power supply will be fine for any single graphics card. Your proposed system is unlikely to draw more than 400W DC with a single graphics card.

Can't really help with memory, aside from general personal experience; I avoid all but Corsair and Crucial. I've given up on twiddling with RAM settings and simply buy modules that use max. 1.65V to run one speed grade above whatever the fastest officially supported speed of the MB (MB supports 1333, I buy 1600).

I strongly suggest acquiring the HDD locally, even if it means resorting to BestBuy. UPS/FedEx destroy drives that are not properly packed, and in my experience Newegg has been among the worst at packing HDDs.

LGA1156 coolers work with LGA1155.

Regarding power supply it depends on which Nvidia GPU you plan to get. If you are getting one of the newest GTX 570 or 580 you might want a 750W or 850W, since power supplys runs best at 50-70% of max capacity. Tomshardware uses a 1000W on all the newer graphic card tests, but that if probably not necessary. Quality is also important, I don't know the Antec brand, but I usually use Corsair.

[QUOTE]Rosewill R102-P-BK 120mm Fan [B]MicroATX[/B] Mid Tower Computer Case

GIGABYTE GA-P67A-UD3 LGA 1155 Intel P67 SATA 6Gb/s USB 3.0 [B]ATX[/B] Intel Motherboard
[/QUOTE]
Will that fit?

Is it worth paying extra for an i7-2600k with 8 threads?

Good catch on the case!
How about this one for $5 more: [URL="http://www.newegg.com/Product/Product.aspx?Item=N82E16811147073"]Rosewill ATX[/URL]

cooler
 

[QUOTE=Prime95;245402]
4) Is the Intel heatsink junk? Is there a better heatsink choice?
[/QUOTE]

From what I have read [URL="http://ht4u.net/reviews/2011/intel_sandy_bridge_sockel_1155_quadcore/index31.php"]here[/URL] it is still the same cooler I bought with my i5 750. It will work and has reserves for a modest overclock. And SB will be cooler than Lynnfield (hopefully, lower core voltages, TDP includes IGP, ...).
But it's like running a marathon with sprint speed. I was much happier with a better cooler: less noise, >15° cooler.
If you overclock with a 2-steps-forward-1-back-strategy I assume you might wanna try 4.5 Ghz to drop to 4 GHz at last.

Most gamers buy Noctua. I chose an European brand: Alpenföhn.

Of the 2x chipset choices at the moment:

P67 needs an additional video card, and allows you to overclock the cpu.

H67 doesn't need an additional video card, and doesn't allow you to overclock.

The 'Z' version of the chipset which is coming soon, will allow both overclocking the CPU and using the GPU.

So you'll need a video card with what you have there. If you have a spare pci card - I'd suggest run with that until you're ready to spend money on a gpu. Current recommendation is the nvidia GTX460 1GB. Good balance between $$ and performance. GTX570 if you're keen. :)

According to anand, the stock cooler can get you to 4.4GHz- "Both my Core i5-2500K and Core i7-2600K hit ~4.4GHz, fully stable, using the stock low-profile cooler."

[url]http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/3[/url]

As for memory, the cpu runs the memory at 2133, 1866, 1600, 1333, or 1066 MHz.

So buy what you can afford basically for memory. Buying 2133 will allow you to test multiple memory configurations to see if things like memory starvation occurs with different combinations of hardware.

As for PSU - 650W is plenty for a single GTX460, GTX560 (when they arrive) or even GTX570. According to nvidia's website 550W is enough for a 570GTX

I have a 650W with GTX460 on i7-920 and I run my PC 24x7 with prime95 on 3x cores, and mfaktc on the other core. It also has 3x SSDs and 1x HDD. So far everything seems sweet.

-- Craig

I would be suprised if Prime95 wanted to overclock while testing so I would suggest H67. This might also mean that he could program the internal GPU as well as NVIDIA cards with the same pc plus saving cost for now and enabling him to get a better GPU later. It would surprise me if most home users don't just use GPUs built into the CPU in a few years.

[QUOTE=ATH;245426]Regarding power supply it depends on which Nvidia GPU you plan to get. If you are getting one of the newest GTX 570 or 580 you might want a 750W or 850W, since power supplys runs best at 50-70% of max capacity. Tomshardware uses a 1000W on all the newer graphic card tests, but that if probably not necessary. Quality is also important, I don't know the Antec brand, but I usually use Corsair.[/QUOTE]

Systems with the newest nvidia cards draw ~500W peak from the wall, which means about 400W actual load on the PS assuming typical 80% efficiency. The 650W referenced in the OP is overkill in that context. It's still a bit of overkill if you're trying to keep the PS in the 50-75% max capacity range, but a bit less so.

Probably won't hurt anything and you'll only save a few dollars going to a more reasonable 500-550W PS. But going to a quality 750W+ will bump the price significantly. There's no need to do this unless you're looking at running multi-GPU setups.

[QUOTE=nucleon;245490]According to anand, the stock cooler can get you to 4.4GHz- "Both my Core i5-2500K and Core i7-2600K hit ~4.4GHz, fully stable, using the stock low-profile cooler."

[url]http://www.anandtech.com/show/4083/the-sandy-bridge-review-intel-core-i7-2600k-i5-2500k-core-i3-2100-tested/3[/url][/QUOTE]
Is this under full load? I would imagine that an "ordinary user" could stably overclock quite a bit on a stock cooler, but only because he's not running a program like Prime95 to keep the CPU at constant 100% (and consequently, maximum heat).

[QUOTE=Brain;245435]
Most gamers buy Noctua. I chose an European brand: Alpenföhn.[/QUOTE]

I found a review on the Thermaltake Silent 1156:

[url]http://www.overclockersclub.com/reviews/thermaltake_silent_1156_/4.htm[/url]

I'm switching to the superior Coolermaster Hyper Tx3 for the same price. I can't justify an extra $50 for a Noctua.

Based on what I've read about these new 32nm cores, I'd say that with any heatsink better than stock you're more likely to be voltage-limited than heat-limited. [url=http://www.tomshardware.com/reviews/phenom-ii-x2-555,2540-4.html]Do not go over 1.4V[/url], and I'd suggest staying under 1.35.

Added a Nvidia Fermi 460 card from Gigabyte:

[url]http://www.newegg.com/Product/Product.aspx?Item=N82E16814125334[/url]

Changed to Corsair 550W power supply. $20 rebate.

You really want to talk to frmky - about Fermis.
460 is cooler and quieter, he said, but its performance is lower.
There are [URL="http://www.newegg.com/Product/Product.aspx?Item=N82E16814162052"]some[/URL] [URL="http://www.newegg.com/Product/Product.aspx?Item=N82E16814130557"]465[/URL] cards for the same price, but with 465, you get the heat and the noise.

[QUOTE=Batalov;245629]460 is cooler and quieter, he said, but its performance is lower.
There are some 465 cards for the same price, but with 465, you get the heat and the noise.[/QUOTE]

I didn't see many 465 cards at newegg - at least not in the $150 price area. Heat and noise aren't a major concern, bang-for-the-buck is. Would the rumored 560 be a better choice? I don't know if I have and spare PCI video cards. I might have to shutdown my Core i7 waiting for the 560 to come out.

[QUOTE=Batalov;245629]
460 is cooler and quieter, he said, but its performance is lower.
but with 465, you get the heat and the noise.[/QUOTE]

OK, I'm confused. I found your 465s. You get a piddly 16 more cores, but they run 200 MHz (shader clock) slower. How does that result in better performance?

P.S. I shied away from the Galaxy brand - I've never heard of them.

[QUOTE=Prime95;245623]Added a Nvidia Fermi 460 card from Gigabyte[/QUOTE]

Is this a hint that George is about to get serious about GPUs?

As a tangent, a favourite quote of mine is "And then God got REALLY serious about kerning....

I read about that particular Galaxy card (I think) - it was very noisy in the review; scratch it (editing time for old message expired); those reviews must be the reason that they are now dumping them at discount.
But EVGA (will lifetime warranty, too) could be good. Or [URL="http://www.newegg.com/Product/Product.aspx?Item=N82E16814125343"]GA[/URL] (a few bucks more).

Re: cores. The difference is all about DP, as far as I understand, - which is of importance to most applications of out collective interest.
The 500s are in another price class - don't know where the sweet spot is.

I wouldn't recommend a 465 for a home system. The big question is whether you want a card with 32 or 48 SP's per SM. All the higher-end cards (465, 470, 480, 570, 580) use 32 SP's per SM, while the 460 and presumably the 560 use 48 SP's per SM. Each SM has two schedulers, each of which can feed 16 SPs. For the 32-SP cards, this is perfect. On the 48-SP cards, nVidia uses superscalar execution to keep the additional 16 SP's busy if instructions aren't dependent. On the older 3.0 compiler, this wasn't very common for most code so the 460 ran more like a 224-SP card (with 112 SP's idling) than a 336-SP card. With CUDA 3.2, instruction scheduling for ILP is noticeably better but still not great.

For code development, a 48-SP card might be better since it will allow you to optimize for ILP and measure the results. This won't affect the higher-end 32-SP/SM cards.

The 560 will probably be about 15% faster than the 460 for about the same price point. There are no performance surprises expected, and it should be out sometime near the end of this month. So the bottom line ... is it worth waiting 2-3 weeks for a 15% performance boost? :smile:

[QUOTE=Prime95;245613]I can't justify an extra $50 for a Noctua.[/QUOTE]

Do it. Noctuas are well worth the price, both more silent and better cooling.

Thanks for the help everybody. All but the video card is on order!

[QUOTE=Kaboom;245673]Do it. Noctuas are well worth the price, both more silent and better cooling.[/QUOTE]

I'll second that recommendation!

BTW Noctua is also European. An Austrian/Taiwanese collaboration.

[QUOTE=Prime95;245402]2) Is the power supply good enough to power an Nvidia GPU?

3) Can I build it using the on-chip GPU and add the Nvidia GPU later (they might be cheaper by the time I'm done with AVX programming and ready to try GPU coding)?[/QUOTE]

Finally jumping on the GPGPU bandwagaon, eh George?

[QUOTE=Prime95;245402]Sandy Bridge CPUs are now available ... I want to build a new computer with this CPU.[/QUOTE]
I am sooooo tempted to go down the same route. After all, the keyboard on this one [I]is[/I] a bit dodgy. :rolleyes:

Newegg was fast as usual. I built it today and it is FAST!

I'm using the Intel CPU cooler now and overclocking it from 3.3 to 3.7GHz. Temps are about 80C. I'll try overclocking more when the non-Intel CPU cooler comes in a week.

Benchmarking a 4M FFT takes just 40 ms. This is compared to 53 ms on my Core i7 running at 3.5 GHz. That's nearly a 25% boost in performance without any new AVX programming. I suspect this is all due to increased memory bandwidth, but much research is required before I know for sure.

[QUOTE=Prime95;246444]That's nearly a 25% boost in performance without any new AVX programming.[/QUOTE]

I look forward to the start of the new 8-core AMD 32 nm processor generation "Code Name Zambezi" in this year.

"The results varied depending on the tasks run, but, finally, the Bulldozer processor proved itself to be approximately 50% faster than the Core i7 950."

[URL]http://news.softpedia.com/news/AMD-Says-Bulldozer-is-50-Faster-Than-Core-i7-CPUs-177958.shtml[/URL]

[QUOTE=Prime95;246444]I suspect this is all due to increased memory bandwidth, but much research is required before I know for sure.[/QUOTE]
This could be (partially) a result of the improved L3 cache infrastructure (ring bus, cache "boxes", etc). The L3 cache latency dropped to 26 cycles from somwhere between 36 and 43 cycles (Core i7 965 and Core i7 980X).

[QUOTE=moebius;246458]I look forward to the start of the new 8-core AMD 32 nm processor generation "Code Name Zambezi" in this year.

"The results varied depending on the tasks run, but, finally, the Bulldozer processor proved itself to be approximately 50% faster than the Core i7 950."

[URL]http://news.softpedia.com/news/AMD-Says-Bulldozer-is-50-Faster-Than-Core-i7-CPUs-177958.shtml[/URL][/QUOTE]
If they put 8 physical cores against 8 HT cores (4 physical + 4 virtual) that result is no surprise...

[QUOTE=Prime95;246444]Newegg was fast as usual. I built it today and it is FAST!

I'm using the Intel CPU cooler now and overclocking it from 3.3 to 3.7GHz. Temps are about 80C. I'll try overclocking more when the non-Intel CPU cooler comes in a week.

Benchmarking a 4M FFT takes just 40 ms. This is compared to 53 ms on my Core i7 running at 3.5 GHz. That's nearly a 25% boost in performance without any new AVX programming. I suspect this is all due to increased memory bandwidth, but much research is required before I know for sure.[/QUOTE]
Would a same clock speed comparison be possible? Both running at 3.5 with memory at the same speed etc

If overclocking is done purely by raising the multiplier, would there be any point in me getting fast DDR3? Is that overclockable too in some way?

[QUOTE=Flatlander;246819]If overclocking is done purely by raising the multiplier, would there be any point in me getting fast DDR3? Is that overclockable too in some way?[/QUOTE]

Yes, memory can be overclocked. I have no benchmarks indicating what difference that might make.

[QUOTE=Flatlander;246819]If overclocking is done purely by raising the multiplier, would there be any point in me getting fast DDR3? Is that overclockable too in some way?[/QUOTE]If you have an unlocked CPU, go for it. Most aren't.

Intel and AMD like to charge a premium for the top end chips because they have an unlocked CPU multiplier, also they usually come with an X or "Black" in the name. They allow more freedom when overclocking, failing that, you have to overclock the system bus that the other clocks (memory, QPI/HyperTransport and for Intel CPUs Uncore) are based on.

The good news is that the other multipliers are unlocked, so you can keep the other clocks at roughly the same speed while overclocking the CPU (or you can raise them all a little too).

Note: When I say CPU multipliers are locked, I mean you can't raise them (except with Turbo mode on Intel CPUs), but you can lower them. Sometimes it is advantageous to have a lower multiplier, either because you are overclocking other components or because the system is more stable, even at the same frequency (ie: 19 * 200 vs 20 * 190). Some people believe that on Intel CPUs the odd multipliers are more stable, whether or not that's true I don't know.

[QUOTE=lavalamp;246882]If you have an unlocked CPU, go for it. Most aren't.[/QUOTE]Note that SandyBridge has changed a large number of things (mostly in the direction of locking them down severely) for overclocking, so certainly read some overview reviews (e.g. [url=http://www.tomshardware.com/reviews/sandy-bridge-core-i7-2600k-core-i5-2500k,2833-9.html]like this[/url]) before purchasing, if that's your goal. The current generation of Intel processors with unlocked multipliers have a "K" suffix, as in i7-2600K, generally at around a 10% price premium to their locked equivalents (well worth it).

Someone recently submitted a SandyBridge benchmark to my site, so comparing a first-generation [url=http://mersenne-aries.sili.net/throughput.php?cpu=Intel%28R%29+Core%28TM%29+i7+CPU+920+%40+2.67GHz|256|8192&mhz=3400]i7-920 (overclocked to 3.4GHz)[/url] to a new [url=http://mersenne-aries.sili.net/throughput.php?cpu=Intel%28R%29+Core%28TM%29+i7-2600+CPU+%40+3.40GHz|256|8192&mhz=3400]i7-2600 at 3.4GHz[/url] you can see the performance increase: on FFT sizes 2M-8M, Sandy Bridge gets 5.0-5.5GHz-days/day compared to 3.5-4.0GHz-days/day on the older i7. That's a notable jump (note: both running at same clock speed).

Yes, I just found out earlier today that for the Sandy Bridge chips, the base clock sets everything else on the board too, SATA, PCIe, USB etc. So raising it even 2 or 3 MHz can cause widespread system instability.

I can see why they've done this, but at the same time, it absolutely sucks and I hope they change it back.

Can I assume that any Prime95 speed increase due to new Sandy Bridge instructions will filter down to LLR.exe for k*b^n+-1?

[QUOTE=Flatlander;247203]Can I assume that any Prime95 speed increase due to new Sandy Bridge instructions will filter down to LLR.exe for k*b^n+-1?[/QUOTE]

Yes. PFGW too.

[QUOTE=lavalamp;246987]I can see why they've done this, but at the same time, it absolutely sucks and I hope they change it back.[/QUOTE]

...why?

[QUOTE=CRGreathouse;247219]...why?[/QUOTE]Because it has basically killed the good old past time of buying a cheap CPU and overclocking the hell out of it.

Now you have to buy a more expensive CPU that already has a high clock, meaning you only squeeze a few more MHz out of it before it reaches the limit.

They have "K" versions of the CPUs that are multiplier-unlocked, so you can overclock them quite a bit. So far, the "K" versions aren't much more expensive than non-K versions ($20-30 at most). Personally, I think the ~$200 i5 2500K is a better bargain than the ~$300 i7 2600K. The main difference between them being that the higher-numbered chip has hyperthreading, which I imagine doesn't matter much except on Folding@Home.

The price is low because so far Intel have only released the budget chips. When the higher end 6 and 8 core Sandy Bridge chips hit, I expect the price of the K chips to be nothing short of astronomical.

[QUOTE=Ken_g6;247248]They have "K" versions of the CPUs that are multiplier-unlocked, so you can overclock them quite a bit. So far, the "K" versions aren't much more expensive than non-K versions ($20-30 at most). Personally, I think the ~$200 i5 2500K is a better bargain than the ~$300 i7 2600K. The main difference between them being that the higher-numbered chip has hyperthreading, which I imagine doesn't matter much except on Folding@Home.[/QUOTE]

Can the i5-2500K chip be installed into a LGA1156 socket (now hosting an i5-750)?

Luigi

It looks like the notches in the side of the CPU are in a different place so they will not physically fit into the socket.

Presumably this is because if they did slot into place, they wouldn't work and could damage the chip or the motherboard.

Here is a source for the pinout of the CPUs, however it is from April 2010, so a newer source may be preferable in case things have changed:
[url]http://www.bit-tech.net/hardware/cpus/2010/04/21/intel-sandy-bridge-details-of-the-next-gen/1[/url]

[QUOTE=lavalamp;247223]Because it has basically killed the good old past time of buying a cheap CPU and overclocking the hell out of it.

Now you have to buy a more expensive CPU that already has a high clock, meaning you only squeeze a few more MHz out of it before it reaches the limit.[/QUOTE]

That doesn't seem to be the case; yes, you have to buy a $300 CPU which is sold at 3.3GHz, but most review sites report that they can get to around 4.0GHz with the standard Intel cooler and 5.0GHz with standard ridiculous overclockers' coolers. Is that not enough?

[QUOTE=ET_;247375]Can the i5-2500K chip be installed into a LGA1156 socket (now hosting an i5-750)?

Luigi[/QUOTE]
No. Sandy Bridge will require the new 1155 socket.
Good old Intel... You need large profit margins to be big.
At least they are using them to develop great products :bow:

[QUOTE=ET_;247375]Can the i5-2500K chip be installed into a LGA1156 socket (now hosting an i5-750)?[/QUOTE]No. Despite the similar pin count, the new LGA1155 and older LGA1156 chips/boards are [i]not[/i] interchangeable (for assorted power and frequency design related issues). However, CPU coolers [i]are[/i] cross-compatible between the two sockets.

No, those are pretty good overclocks and proportionally match the extra capacity of the i7 920 (ie: 50% over stock), assuming they are truly stable (a lot of people consider just 1 hour of Prime95 stable).

However, with hyperthreading enabled (as in the 2600K which inexplicably costs $105 more than the 2500K) you lose a couple of hundred MHz on the max OC. Additionally, when they increase the stock frequency and double the number of cores for the high end, the extra performance overclockers can eek out will be reduced still further.

If Intel still sell high-end CPUs at low prices this time next year, and they are capable of routinely hitting 50% overclocks from the stock frequency (with HT), I'll be happy, and I may even buy one.

I don't think the extra pricing for the i7-2/600 is inexplicable: for embarrassingly parallel jobs like NFS sieving, the hyperthreading is very useful, so certainly I'll get an i7-2/600 if I get a Sandy Bridge at all.

Indeed, hyperthreading is useful, I use it myself. You can get 25% extra clock-for-clock performance out of the chip with it for certain applications. But the cost of the CPU is nearly 50% greater.

You may say that if you have a $600 system, then switching to the 2600K increases the total cost by only 17.5%, and then the extra cost is worth it for the performance gain. I agree with that, however I maintain that 0.1 GHz and hyperthreading is only a small improvment, so a fairer bump would be closer to $50.

Of course the price difference isn't really inexplicable, that was just hyperbole. It's actually quite simple; Intel like money.

[QUOTE=lavalamp;247475]Intel like money.[/QUOTE]That applies for the Extreme Edition chips (always introduced with a fixed price of US$999), but not so much here. And it's nothing Intel-specific; NVIDIA does the same thing (the GTX580 is 150% the cost of the GTX570; is that an appropriate incremental price?

[QUOTE=Flatlander;247203]Can I assume that any Prime95 speed increase due to new Sandy Bridge instructions will filter down to LLR.exe for k*b^n+-1?[/QUOTE]

[QUOTE=rogue;247212]Yes. PFGW too.[/QUOTE]
Great!

If I have pizzas delivered to George do you think he will code through the night? :wink:

[QUOTE=Flatlander;247661]Great!

If I have pizzas delivered to George do you think he will code through the night? :wink:[/QUOTE]

I guess he actually won't need pizzas to code through the night... :smile:

Luigi

He may not NEED the pizzas, but the question is if he will code better if he gets some...grin...

Thumbing through a magazine at a shop earlier,
It said that if you disable the on-die GPU you are also disabling the new instructions.

So is it [I]really [/I]overclock OR new-instructions, or did I mis-read?

edit:
Let me rephrase that, it said if you are using a discrete graphics card.

[QUOTE=Flatlander;248001]Thumbing through a magazine at a shop earlier,
It said that if you disable the on-die GPU you are also disabling the new instructions.

So is it [I]really [/I]overclock OR new-instructions, or did I mis-read?

edit:
Let me rephrase that, it said if you are using a discrete graphics card.[/QUOTE]


Yes that is correct the P67 Chipset requires Discrete Graphics aka a PCIe Video Card which fully disables the new Quick Sync Video Transcoding (and all of the on-CPU Video).

For H67/Z68 which has video outputs on the motherboard, I've been told that on-CPU Video still works when you have a PCIe Video Card plugged in as well making this great for 3-4 monitor setups.

AVX the 256-bit FP instructions which offer a possible 50-100% performance increase for Prime95 are not affected.

[QUOTE=abstractalgebra;248132]...
AVX the 256-bit FP instructions which offer a possible 50-100% performance increase for Prime95 are not affected.[/QUOTE]
Ah, that's where I was getting confused. Thinking AVX was part of Quick Sync.

Thanks. :smile:

On the topic of HT v non-HT: I find when running apps like prime95, the machine is more responsive with HT if the app is using 100% of the real cores. With the options in prime95 - it's quite easy to distribute the load across the real cores.

George/anyone else,

What's the noise level like with the stock cooler at full load?
(I'm considering getting the i7-2600k. I assume the cooler is the same as the i5 2500k.)

Thanks!

edit:
As this will be purely for prime stuff I am hoping to use Ubuntu rather than buy Windows 7. Is it likely to be able to install mobo drivers? (Gigabyte GA-P67A-UD3P, probably.)
I will be using a discrete gpu (2 yrs old) so am hopeful there will be drivers for that.

[QUOTE=Flatlander;249174]As this will be purely for prime stuff I am hoping to use Ubuntu rather than buy Windows 7. Is it likely to be able to install mobo drivers? (Gigabyte GA-P67A-UD3P, probably.)
I will be using a discrete gpu (2 yrs old) so am hopeful there will be drivers for that.[/QUOTE]
I don't think that Microsoft's idea of mobo drivers applies to *nix - it just works there. Then only moment I recall when I had to check my mobo chipset info was when I wanted to install tempereature monitoring software, and that was years ago and on gentoo. There's an extremely high level of certainty that you won't need to bother about mobo drivers on linux.
As for the gpu, do you imply using it just for 2d diplay stuff or you mean gpgpu computing ?

Thanks. I wasn't sure because I assume the P67 chipset came out after the last Ubuntu update. I don't speak *nix!

The gpu is just for the display, I might look into gpu computing later.

[QUOTE=Flatlander;249201]Thanks. I wasn't sure because I assume the P67 chipset came out after the last Ubuntu update. I don't speak *nix!

The gpu is just for the display, I might look into gpu computing later.[/QUOTE]

Quote from [URL="http://www.phoronix.com/scan.php?page=article&item=intel_sandybridge_linux&num=2"]here[/URL]:
[QUOTE]So if you find yourself with a new Intel P67 motherboard and the latest Core i3/i5/7 processor in January, you should be able to get the system up and running on other late Q3'2010 or Q4'2010 Linux distributions such as Ubuntu 10.10.[/QUOTE]
Apart from that, my understanding is that even if a given chipset's release date is younger than the one of an OS then within basic usage constraints all computers run the same and the OS will start, possibly using best match settings - with the details making difference only when some specific stuff (presumably like integrated graphics, energy saving,..) is being used.

As for the gpu then even if it was some exotic brand/model then it will be defaulted to some VGA-compatible modes, otherwise you shouldn't encounter major problems (after all for mprime you will only need a console with a command line).

Linux doesn't bite and I hope you will enjoy using it, hopefully soon noticing the [URL="http://widefox.pbworks.com/w/page/8042308/Kernel-Comparison-Linux-vs-Windows"]advantages[/URL].

Thanks for finding that.
I tried Linux once before but only for a while to do some 64bit seiving. Didn't really get into it in that time. A strange alien landscape!
Nothing to lose but my chains, right? :smile:

[QUOTE=Flatlander;249215]Nothing to lose but my chains, right? :smile:[/QUOTE]Eggheads of the world unite! You have nohting to lose but your yolks.

Seriously, I heartily recommnend using a number of different operating systems to become familiar with them. You don't need a number of different computers to do so. With a virtual machine package (VMware, VirtualBox, etc) you can even run all the different systems concurrently on the same hardware.


Paul

New build coming tomorrow. :smile:

Without Windows for the first time since 1993 :max:

I wish you all the best. I couldn't get any of the linux cuda code to compile for me so when the cuda windows binaries came out - the choice was a no brainer for me.

-- Craig

Hopefully it won't be too bad. I probably won't be attempting anything cuda until I've found my way around and llrCuda has matured. Then I'll need a cuda capable gpu.

There is still the possibility that I'll go running back to Windows. :redface:

[QUOTE=Flatlander;249379]There is still the possibility that I'll go running back to Windows. :redface:[/QUOTE]
You can have anything you want with a dual(multi)boot :)

Installed new video card and CPU cooler (the cheap CoolerMaster).

With stock cooler at 3.8 GHz temps reached as high as 87C. Now running at 4.1 GHz with temps reaching only 67C.

I haven't even tried to see how high this thing will overclock. It is stable at 4.1 GHz.

[QUOTE=Prime95;250499]Installed new video card and CPU cooler (the cheap CoolerMaster).

With stock cooler at 3.8 GHz temps reached as high as 87C. Now running at 4.1 GHz with temps reaching only 67C.

I haven't even tried to see how high this thing will overclock. It is stable at 4.1 GHz.[/QUOTE]

Is this a Windows build? I can't get a reading with lm-sensors in ubuntu. Same mobo.
I'm at 4hr-stable, 4.3GHz (i7 2600k) with a Noctua cooler. Just fits in the case!

[QUOTE=Flatlander;250507]Is this a Windows build?[/QUOTE]

I've loaded 64-bit Windows 7. I left 150GB available for Ubuntu, but I haven't loaded it yet.

[QUOTE=Flatlander;250507]I can't get a reading with lm-sensors in ubuntu.[/QUOTE]
What does it say as an error message ? Can you paste an output from dmesg as well ?

[QUOTE]
No sensors found!
Make sure you loaded all the kernel drivers you need.
Try sensors-detect to find out which these are.
[/QUOTE]

Outputs from sensor-detect and dmesg attached.:smile:
I think the mobo isn't supported yet.

[QUOTE=Flatlander;250517]Outputs from sensor-detect and dmesg attached.:smile:
I think the mobo isn't supported yet.[/QUOTE]
Thanks, can you check this thread:
[url]http://www.spinics.net/lists/lm-sensors/msg30887.html[/url]

That worked. Thank you!
Cores are around 71 degrees @4.4Ghz. Stable for 2hrs :smile:

Looking at Chris's [URL="http://www.noprimeleftbehind.net/stats/index.php?content=port&server=GB&port=9000"]output[/URL] on NPLB's port 9000 server, I see he's cranking out 40 tests/hour, which is just about exactly twice what my Q6600 can do (16 tests/hour at stock and 20 tests/hour at 2.8GHz)! :shock:

Gary a.k.a. gd_barnes, also from NPLB, mentioned to me that his i7 (pre-Sandy Bridge, not sure of the exact model) running at stock speeds is worth about 5 cores of a stock Q6600 doing LLR on typical NPLB work (k*2^n-1 with n<1M). Since the pre-SB i7 can do about 20 tests/hour on similarly sized numbers, and Chris's SB can do 40 pairs/hour, that makes for a 100% speed gain! Granted, the SB is overclocked a fair amount while the pre-SB is not (and had a lower stock speed to boot IIRC), but this is still quite a significant boost.

What's interesting is that, based on George's finding that his SB gets about a 25% boost over a similarly-clocked pre-SB i7, small FFTs (64K in this case) seem to benefit the most.

Now I'm definitely reconsidering some earlier statements I made elsewhere on the forum that a Sandy Bridge i5 < an AMD Phenom II X6 in terms of total throughput. That may be the case for GIMPS (and other large-FFT work) but not so for NPLB and the like.

[QUOTE=mdettweiler;250554]
Now I'm definitely reconsidering some earlier statements I made elsewhere on the forum that a Sandy Bridge i5 < an AMD Phenom II X6 in terms of total throughput. That may be the case for GIMPS (and other large-FFT work) but not so for NPLB and the like.[/QUOTE]

I told you so...you don't believe me...lol The core i5 750@3.6GHz easily beats an overclocked X6. Paul with his overclocked X6 was making the same LLR time as my overclocked Q6600 at the same n range, the core i5@3.6 GHz is 40 % faster than the Q6600@2.8 GHz! This difference increases with n size.

This sucks: [URL="http://techreport.com/discussions.x/20326"]http://techreport.com/discussions.x/20326[/URL]

[QUOTE=frmky;250649]This sucks: [URL="http://techreport.com/discussions.x/20326"]http://techreport.com/discussions.x/20326[/URL][/QUOTE]
Yep. :cry:
[QUOTE]The two 6Gbps SATA ports aren't at risk.[/QUOTE]
Phew! My HD is running from one of them and has no other drives. Maybe one more in the future.

There's a reason they call it the "bleeding edge"....but, for us, it is a simple fix for the aware: use the 6GB SATA ports, and forget the 3GB SATA ports, done...we are usually not disk-bound, anyway. Cache/memory bus, yes, or CPU speed, yes, but not disk...

[QUOTE=Christenson;250737]we are usually not disk-bound, anyway. Cache/memory bus, yes, or CPU speed, yes, but not disk...[/QUOTE]Usually, but just recently I ran out of onboard SATA ports (there were 10). After a re-arrange and the purchase of a RAID card, I'm now using 12 SATA ports out of 20 total.

Flatlander, on some boards there are sometimes additional SATA ports put there by the manufacturer which use a separate controller hooked up to a PCIe lane or two. On my board for example, there are 6 ports run from the ICH10R in the chipset, and then 4 ports run from two extra controllers.

You should only use the two SATA6 ports (the others have the Intel Sandy bridge bug).

Back in the thread wondering which CPU was best, a Sandy Bridge machine was considered. Intel is clearly in full court fix mode, but suppose I wanted to duplicate my 6-core AMD machine...are there cheap, unfixed sandy bridge Mobos out there?

[QUOTE=Christenson;252223]Back in the thread wondering which CPU was best, a Sandy Bridge machine was considered. Intel is clearly in full court fix mode, but suppose I wanted to duplicate my 6-core AMD machine...are there cheap, unfixed sandy bridge Mobos out there?[/QUOTE]

No, working mainboards are to be expected in March. The unfixed ones obviously didn't get cheaper.

The 6core AMD is a nice piece of work (and for the price a lot of faster in several circumstances). The advantage of e.g. the 2500K is the wide overclocking ablilty (but only with P boards (without included graphic card)). AMD boards are known to be more priceworthy than the Intel ones.

I suggest if you want to buy one now you might check out one of the AMDs cpus.

I have 2600k, gigabyte ud7 motherboard, corsair h70 cooler.

I can get to 4.8GHz but no higher. I've tried heaps of voltage changes - but no luck.

4.8GHz is only about 30mins prime stable :( Temp reaches 83degC.

Currently trying 4.5GHz, temperature maxes at 66degC. So far 15mins and going....

-- Craig

[QUOTE=nucleon;252302]I have 2600k, gigabyte ud7 motherboard, corsair h70 cooler.

I can get to 4.8GHz but no higher. I've tried heaps of voltage changes - but no luck.

4.8GHz is only about 30mins prime stable :( Temp reaches 83degC.

Currently trying 4.5GHz, temperature maxes at 66degC. So far 15mins and going....

[/QUOTE]

@4.8GHz and 4.5GHz prime95 25.11 dies at 768k torture test consistently (about 45-50mins in), but only when run from the start with default settings. It ends in blue screen. If I run the 768k test straight off - no problems. Bizare.

@4.5GHz prime95 26.5.2 was stable for 2.5hrs then I ended the test. Temps max at 68degC.

Big difference in temps there. Going to keep it at 4.5GHz for a while. I might leave it go for 24hrs+

-- Craig

I've got 4.4GHz stable, Prime95, 8 threads, 1.325 volts in BIOS.
Max. temps are 64-68C with Noctua cooler depending on ambient.
4.5GHz is reachable with more heat/noise.
I haven't tried raising BCLK yet.

(i7-2600k on Gigabyte GA-P67A-UD3. RAM @ 1600MHz.)

[QUOTE=Flatlander;252358]I haven't tried raising BCLK yet.[/QUOTE]And you shouldn't...

I got to do some other things before I bring up yet another 'puter...idea had been to put the sandy bridge side-by-side the AMD 6 core and compare. I'll wait until march....

[QUOTE=Christenson;252401]idea had been to put the sandy bridge side-by-side the AMD 6 core and compare[/QUOTE]Not exactly the same, but gives you some idea: you can compare single-core performance of an [url=http://mersenne-aries.sili.net/throughput.php?scale=1.2795650924844&cpu1=AMD%20Phenom%28tm%29%20II%20X4%20940%20Processor|512|6144&mhz1=3000&cpu2=Intel%28R%29%20Core%28TM%29%20i7-2600K%20CPU%20%40%203.40GHz|256|8192&mhz2=4400]AMD Phenom X4-940 and an i7-2600K[/url] on my site. I don't have any X6 benchmarks, but I don't think they vary too far (comparing single cores) from the X4. Note the i7 is overclocked, but feel free to adjust clockspeeds on either system in the comparison.

I'm not above being a mersenne-airy myself. Therefore, if those other things get done, I think there's a sandy bridge in my future, wondering how hard it would be to set a GTX460 in it and get work out of it, too, probably under Xubuntu 10.10. I'd been hoping for a price break due to the SATA bug, but it looks like not.

[QUOTE=James Heinrich;252404]Not exactly the same, but gives you some idea: you can compare single-core performance of an [URL="http://mersenne-aries.sili.net/throughput.php?scale=1.2795650924844&cpu1=AMD%20Phenom%28tm%29%20II%20X4%20940%20Processor%7C512%7C6144&mhz1=3000&cpu2=Intel%28R%29%20Core%28TM%29%20i7-2600K%20CPU%20%40%203.40GHz%7C256%7C8192&mhz2=4400"]AMD Phenom X4-940 and an i7-2600K[/URL] on my site. I don't have any X6 benchmarks, but I don't think they vary too far (comparing single cores) from the X4. Note the i7 is overclocked, but feel free to adjust clockspeeds on either system in the comparison.[/QUOTE]
I have an AMD Phenom II X6 1045T, perhaps I can be of assistance somehow? It's not exactly a "standard" Phenom II, in that this particular model is only sold in pre-built PCs. But I'd imagine it's fairly comparable?

[QUOTE=James Heinrich;252404]Not exactly the same, but gives you some idea: you can compare single-core performance of an [url=http://mersenne-aries.sili.net/throughput.php?scale=1.2795650924844&cpu1=AMD%20Phenom%28tm%29%20II%20X4%20940%20Processor|512|6144&mhz1=3000&cpu2=Intel%28R%29%20Core%28TM%29%20i7-2600K%20CPU%20%40%203.40GHz|256|8192&mhz2=4400]AMD Phenom X4-940 and an i7-2600K[/url] on my site. I don't have any X6 benchmarks, but I don't think they vary too far (comparing single cores) from the X4. Note the i7 is overclocked, but feel free to adjust clockspeeds on either system in the comparison.[/QUOTE]
Here's one (oops it's actually a 1090T OC'd to 3400 [17x200] instead of 3200[16x200]).

Found out a problem with my machine.

During 25.11 testing, the voltage drops about 0.10V or so (hard to see exact figure as it reboots shortly after that). And hence the PC blue screens after that.

Similarly with the other test I did. (4.5GHz 26.5.2). Only difference is when.

Enabling Vdroop to level2, kept the voltage within a tighter bounds (0.03V). Voltage now increases with load and increases slightly. Without Vdroop the CPU voltage was decreasing due to load. (gradually and during the load, take away load - voltage went up)

The PC survived @4.8GHz with Vdroop-2 for at least 3hrs or so before blue screen.

Dropping to 4.5GHz the machine appears to be more stable. Now it's lasted 12+hrs (torture test started 7:30am, it's now 8:10pm). No reported errors so far.

From reading around overclocking forums, 4.5GHz seems pretty common as a 24/7 overclock.

So I think it's fairly stable now. Anyone got any ideas how long I should leave the torture test go for before putting this machine to work?

-- Craig

From what I remember of the previous i7s, enabling that option isn't recommended. Yes it means you get a higher voltage under load, but when the load is suddenly removed, it can't react fast enough and you get a large momentary spike in voltage.

Possibly things have changed since the last generation, but it's certainly something I'd recommend checking up on.

Is raising the voltage permanently not an option?