|
Alereon posted:Go with a good air cooler, only the crazy massive three-fan watercooling setups actually perform any better
|
# ? Apr 28, 2012 20:05 |
|
|
# ? May 8, 2024 07:13 |
|
Alereon posted:Yes, remember that Ivy Bridge is a price cut from Sandy Bridge (though that's absorbed by the inventory-clearing sales), is more power-efficient, faster clock-for-clock, and has much better graphics performance. The only downside is the higher temperatures at high power levels due to using the TIM instead of solder, and that really only comes into play when you start pushing the voltage to pull the last few mhz out of the core you can. Thanks. I don't care about the graphics, but with my mid-level silent fan I suppose IB + mild-high overclocking is the way to go.
|
# ? Apr 28, 2012 20:11 |
|
Is the 212 Evo still the recommended?
|
# ? Apr 28, 2012 21:16 |
|
Still great performance for the price, yes.
|
# ? Apr 28, 2012 21:20 |
|
The Thermalright HR-02 Macho for $39.99 is also a good option for higher performance and lower noise without getting into crazy expensive supercooler territory.
|
# ? Apr 28, 2012 21:54 |
|
I guess I expected a bit too much for the local TigerDirect/CompUSA to have Day 1 availability of Ivy Bridge. But good thing that they have 6 3930Ks waiting for a new home.
|
# ? Apr 28, 2012 22:21 |
|
Alereon posted:The Thermalright HR-02 Macho for $39.99 is also a good option for higher performance and lower noise without getting into crazy expensive supercooler territory. Xigmatek Gaia and Arctic Cooling Freezer aren't bad either. Most heatpipe tower types with a big quiet fan will do a good job, if you want to push the overclocking look for one with direct touch heatpipes that makes it easy to mount a second fan.
|
# ? Apr 28, 2012 22:31 |
|
pienipple posted:Xigmatek Gaia and Arctic Cooling Freezer aren't bad either. Most heatpipe tower types with a big quiet fan will do a good job, if you want to push the overclocking look for one with direct touch heatpipes that makes it easy to mount a second fan. I have a Freezer 7 Pro Rev. 2 which I picked up for like £10 recently. I've only ever used these coolers before. I've got no idea if it'll be good enough for Ivy though.
|
# ? Apr 28, 2012 22:54 |
|
Is there a summary of Ivy Bridge info somewhere in the thread? The OP is still about SB.
|
# ? Apr 28, 2012 23:03 |
|
aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa read the last few pages E: Oops, I was having flashbacks to the system building sticky thread for the past few weeks. There. Or read reviews.
|
# ? Apr 28, 2012 23:28 |
|
Fruit Smoothies posted:I have a Freezer 7 Pro Rev. 2 which I picked up for like £10 recently. I've only ever used these coolers before. I've got no idea if it'll be good enough for Ivy though.
|
# ? Apr 28, 2012 23:30 |
|
grumperfish posted:Not really the best choice for IB/SB. The design is pretty ancient and inefficient at this point, especially compared to the price/performance you get from the 212+ coolers. Yeah, it was one of the earliest and thus now one of the most deprecated heat pipe + radiator coolers. The smaller, faster fan means it's louder for less cooling. I was glad to replace the heat sink on my Q9550 with a Hyper 212+, dropped about 20ºC from the Freezer Pro 7. However, that's an extremely anomalous result, and mainly because the TIM had dessicated over the three years in which the Freezer Pro had been installed. As cheap as it is, it still performs pretty well for the money, it's just not a good buy as it is very outdated tech for cooling. I've been thinking more about the TIM vs. solder thing, and honestly I'm kind of impressed that we have a very "mystery solved!" moment here... It's not nearly as negative as "unfortunately Ivy Bridge has poor thermal performance," it's just that it's got a much less efficient heat transfer method between the CPU and the heat spreader. Which honestly seems like a (perhaps necessary, but...) show of faith in the lower thermal requirements for the 22nm/tri-gate lithography, justified in all but extreme usage scenarios. It's still unfortunate for overclockers, but it seems to bottleneck people thermally about performance-wise where many people were bottlenecked with voltage on Sandy Bridge, apples to apples (IPC improvements taken into account). Perhaps there's an issue with the integrity of the transistors being able to withstand the soldering process employed. It's difficult to say why exactly they went with what we all know is a less efficient method, but I don't think it's because they just decided it'd be cheaper to do this and screw everyone. I would imagine it has to do with process, not pocketbook... But I don't know that for sure. Just seems understandable.
|
# ? Apr 29, 2012 02:38 |
|
Looks like the TIM / solder question might not matter so much... http://forums.anandtech.com/showthread.php?p=33354816#post33354816 Someone took off the IHS and placed a NH-D14 right on the core and did not see any temperature reduction (well, looks like 1 number went down 1 degree). While there are lots of comments about the heatsink not pressing hard enough on the core I think it might not matter as much as we though it would. I hope a few others test this when it officially releases tomorrow. At least it seems Intel didn't make a mistake in using TIM. It's probable they did various tests themselves and determined it wouldn't matter so much between TIM and solder and went with the cheaper method.
|
# ? Apr 29, 2012 03:36 |
|
I thought the 3570K was supposed to be priced at $212. Why does Newegg list it for $250?
|
# ? Apr 29, 2012 18:54 |
|
Because money. Micro Center's doing a $190 3570K doorbuster with an additional $50 off if you get a featured motherboard, too. Roughly the same deal as Sandy Bridge but $10 more.
|
# ? Apr 29, 2012 19:04 |
|
The nearest Microcenter to me is 6 hours away, and adding it to my cart on the website promptly removes it.
|
# ? Apr 29, 2012 19:07 |
|
My local Micro Center had zero 3570Ks in stock when they opened today. I picked up a 3770 non-K since they had that in stock
|
# ? Apr 29, 2012 22:30 |
|
So far only two z75 motherboards on newegg none by Intel or Asus. Checking on the z77 offerings from those two. The two Intel boards have 1394 but no dvi. And the two asus that are ~200 and lower with Intel lan chipsets have dvi but no 1394. I guess I could live without the 1394 but really would like the onboard dvi to drive my secondary monitor with the cpu gpu and my primary monitor with my video card. Currently looking at this board but will wait a few weeks. I really have no need for half the crap on that board. And really, a ps2 port still!? calcio fucked around with this message at 02:49 on Apr 30, 2012 |
# ? Apr 30, 2012 02:45 |
|
calcio posted:And really, a ps2 port still!? There are times when USB keyboards don't work quite right, and you would kill for a PS/2 port and keyboard.
|
# ? Apr 30, 2012 03:30 |
|
Factory Factory posted:There are times when USB keyboards don't work quite right, and you would kill for a PS/2 port and keyboard. A thousand times this. When I tried installing SP1 for Windows 7 it really messed up my USB drivers so nothing USB related worked. I was saved by my PS/2 ports. Just because it is older tech doesn't mean it is entirely useless.
|
# ? Apr 30, 2012 04:00 |
|
calcio posted:And really, a ps2 port still!? No PS2 port means enabled USB which means no secure computers and no DoD sales.
|
# ? Apr 30, 2012 05:24 |
|
hobbesmaster posted:No PS2 port means enabled USB which means no secure computers and no DoD sales.
|
# ? Apr 30, 2012 06:04 |
|
calcio posted:And really, a ps2 port still!?
|
# ? Apr 30, 2012 10:08 |
|
Any word on actual availability of Dell/HP 3820QM or 3720QMs for the EliteBook/Precision lines. I have an approved purchase order from work, but am beginning to really hate my Core2 P8600.
|
# ? Apr 30, 2012 13:57 |
|
incoherent posted:Kneecapping IVB so there is a big difference between IVB and IVB-E. Yeah IVB-E doubling the cores, using a bidirectional ring, completely rearchitecting I/O... all fluff. This internal heat transfer poo poo literally nobody sees except a couple overclockers, that's where there's some tricky segmentation happening.
|
# ? Apr 30, 2012 15:13 |
|
Local store has the 3570K for $189 boxed. Should I jump on that? I was planning to start building a new system anyway so I can set it aside for that. poo poo, it's $249 at Newegg.
|
# ? Apr 30, 2012 16:31 |
|
Powdered Toast Man posted:Local store has the 3570K for $189 boxed. Should I jump on that? I was planning to start building a new system anyway so I can set it aside for that. poo poo, it's $249 at Newegg. Yes, you will not find it for any cheaper.
|
# ? Apr 30, 2012 16:32 |
|
Just grabbed a 3770..no K . Hope I don't regret that in the future.
|
# ? Apr 30, 2012 16:35 |
|
The only chance for regretting that would be in the future, when other people might be putting off an upgrade for a generation or two by overclocking theirs by a freebie 1GHz/core or a little more effort for 1.2-1.3GHz/core and you're stuck at stock turbo implementation. Still, it's going to be a powerful CPU for a long time, and you don't have to dick with anything. But logically speaking, if you're going to regret it, it's gonna be in the future, yeah.
|
# ? Apr 30, 2012 17:14 |
|
Factory Factory posted:There are times when USB keyboards don't work quite right, and you would kill for a PS/2 port and keyboard. USB keyboards depend heavily on SMIs as well (at least at the pre-OS stage) whereas PS/2 keyboards are generally run off a KBC on a SuperIO which in turn is connected to the PCH over LPC.
|
# ? Apr 30, 2012 17:35 |
|
movax posted:USB keyboards depend heavily on SMIs as well (at least at the pre-OS stage) whereas PS/2 keyboards are generally run off a KBC on a SuperIO which in turn is connected to the PCH over LPC. Just look at that sentence, it's glorious with acronyms.
|
# ? Apr 30, 2012 18:40 |
|
I understood USB, PS/2, OS, PCH, and the IO part of SuperIO. I actually understood SuperIO. Seriously, though, that post is ripe for being linked like the first paragraph of a Wikipedia page.
|
# ? Apr 30, 2012 18:48 |
|
Sorry LPC is Low Pin Count, as the name suggests, it is a low pin-count, low speed bus designed to replace ISA. The physical layer is different, but software-wise, it is similar to ISA. On older boards, the FWH (Firmware Hub, the PLCC chip where BIOS would live) would be attached to this bus. Now, the trend is moving to SPI ROMs, which are connected via SPI (serial peripheral interface) to the chipset. I believe early Xbox modchips would sit right on the exposed LPC bus on the motherboard; Microsoft left a header for their EOL tooling. SuperIO is a catch-call term for a chip that integrates a whole bunch of low-speed devices/ports, generally considered legacy. Your average SuperIO could have KBC (keyboard controller), a few UARTs, a parallel port, and then various managment interfaces (SMBus mastering, PECI, etc). It's up to the mobo maker to see if they have the room/time/inkling to route those out. I really like mobos that still offer a little header for a COM port that you can breakout to a rear-panel mounted jack. SMI/SMM gets a bad rep, but it's System Management Interrupt. It's how you get into System Management Mode (around since the 386). This is a mode that's recently been exploited by rootkits (mostly proof-of-concept). Basically, the SMI# pin on the CPU gets asserted, and it goes into a mode where all normal system execution is completely suspended, an other code can be executed at extremely high privilege level. The intent was for it to be used for management style functions...thermal events, power management, sleep states, etc. It turns out it's also a great way to get a USB keyboard or mouse to emulate its PS/2 cousin. It also can murder the poo poo out of real-time performance, and anyone who does real-time software would have a heart attack if in the middle of their code, the CPU halted their stuff and went to service an ACPI PM event. It's mostly an open-secret (you could check with your DMM if you wanted I guess) that the physical pin on the CPU is tied to the PCH. You'd then write to a given I/O port to trigger a given SMI. By design, your OS can't do poo poo about this (aside from controlling the ways you can write to I/O memory space). The moral of the story is a lot of "odd" behaviour can be chocked up to backwards compatibility/legacy, I guess. On the IVB note, I'm excited to see the effects on the mobile market (battery life!). We all knew/were told it'd be much more important for mobile devices, and a "free" increase in battery life would be amazing. Solid benchmarks comparing say a 4.5GHz 2500K (average SNB overclock) to an average 3570K overclock would be nice to see. The slightly lower clocked IVB should be very close if not identical to a slightly faster SNB cousing, while sporting 22nm, PCIe 3.0 and a better IGP. I don't think getting a SNB CPU for a 7-series motherboard would be a clever idea, unless I'm missing something (which is entirely possible as I haven't read a huge amount lately) movax fucked around with this message at 19:27 on Apr 30, 2012 |
# ? Apr 30, 2012 19:24 |
|
movax posted:Solid benchmarks comparing say a 4.5GHz 2500K (average SNB overclock) to an average 3570K overclock would be nice to see. The slightly lower clocked IVB should be very close if not identical to a slightly faster SNB cousing, while sporting 22nm, PCIe 3.0 and a better IGP. [H] compared an i7-3770K to an i7-2600K, both at 4 GHz, to compare IPC. Results were all over, and I don't have a stats package handy, but the arithmetic mean was that IVB is 5% faster isoclock. link Intel made sure that 7-series boards support 95W TDPs, and there are no IVB-specific features that SNB can't do (except slower) besides DirectX 11 and OpenCL, so there's pretty much zero problem with matching a 7-series board and SNB. It works at least as well as a 6-series board.
|
# ? Apr 30, 2012 20:18 |
|
Factory Factory posted:[H] compared an i7-3770K to an i7-2600K, both at 4 GHz, to compare IPC. Results were all over, and I don't have a stats package handy, but the arithmetic mean was that IVB is 5% faster isoclock. link Ah awesome, I haven't read [H] recently, but I'll look through those graphs. quote:Intel made sure that 7-series boards support 95W TDPs, and there are no IVB-specific features that SNB can't do (except slower) besides DirectX 11 and OpenCL, so there's pretty much zero problem with matching a 7-series board and SNB. It works at least as well as a 6-series board. Yeah, compatibility isn't worrisome, I'd be more at a brand-new consumer (someone who isn't at least on Sandy Bridge) buying a new 7-series board + SNB CPU.
|
# ? Apr 30, 2012 20:43 |
|
While my current SB system has a PS/2 port, my old Core2 Quad system only had USB and I never had any issues with using USB keyboards in DOS or in the BIOS
|
# ? May 1, 2012 01:43 |
|
USB3 ports don't work for me before the OS starts (Sandy Bridge Z68). USB2 is fine though yeah.
|
# ? May 1, 2012 04:49 |
|
I used to love reading articles like that HardOCP review posted earlier. Now that I've done some of that kind of work I wonder why they don't go more in depth. Error bars please? Maybe some VTune analysis to tell us why there's even an observed 7% difference in some video encoder when we're expecting the microarchitectures to be the same in most compute-bound cases?
|
# ? May 1, 2012 05:41 |
|
1) There are no error bars because the benchmarks were run once each. Variations in end-user benchmarks occur because background tasks want to do their things at different times, and re-running an entire end-user OS identically down to the clock cycle is practically impossible. And even if you could, each end user will have different background tasks going, so the precision is wasted effort. 2) VTune is expensive, and [H] is a site made of dudes who enjoy burning up consumer hardware, not making cost-benefit analyses of business equipment. 3) We weren't expecting the microarchitectures to be exactly the same. Intel's ticks aren't just die-shrinks; they include minor revisions to the architecture. Improved IPC is the goal of such tweaks. The only reason I mentioned a stats package was to detail the variation in per-task performance differences rather than just report the mean. Factory Factory fucked around with this message at 06:38 on May 1, 2012 |
# ? May 1, 2012 06:36 |
|
|
# ? May 8, 2024 07:13 |
|
Ars has an article on DDR4 memory products becoming available, and it includes a bit of Intel roadmap data: DDR4 will be a part of Haswell-EX Xeons in 2014, and there is no roadmap yet for DDR4 in consumer devices.
|
# ? May 10, 2012 04:58 |