Intel: lol

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Intel: lol

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Update 2018: Patch your poo poo if you can! Meltdown and Spectre are here...

Update 08/31/2014: Haswell-E is here, pairing up to 8 Haswell CPU cores with the X99 chipset and quad-channel DDR4 memory. This platform is for people who really need 6+ CPU cores, >32GB of RAM, double the normal memory bandwidth, or 32+ PCIe lanes. Most people would be better served with a 4790K, especially for gaming.

Update 05/02/2014: The first motherboards with Intel Z97 chipsets are now available early. They continue to use the LGA-1150 socket and support existing Haswell CPUs as well as the upcoming Haswell Refresh/Devil's Canyon models and next year's 14nm Broadwell CPUs. Z97 boards typically sport SATA-Express and many include NGFF/M.2 slots for SSDs. We will know more once chipset reviews launch next week.

Update 06/01/2013: Haswell reviews are here, here is selected coverage from Anandtech:
Anandtech: The Haswell Review: Core i7 4770K and i5 4560K Tested
Anandtech: The Intel Iris Pro 5200 Graphics Review: Core i7 4950HQ Tested
Anandtech: Intel's Haswall Quad-core Desktop Lineup
Anandtech: Intel's Haswell Quad-core Mobile Lineup
Anandtech: Intel's Haswell: An HTPC Perspective

Update 10/05/2012: Anandtech has posted their Haswell architecture analysis.

Update 04/23/2012: Anandtech has posted their Ivy Bridge Core i7 3770K review.

Update 03/06/2012: Anandtech has posted their Ivy Bridge Core i7 3770K preview, also confirming the Ivy Bride launch has been pushed back until April 29th. 7-series motherboards will still be launching April 8th as scheduled.

Update 01/31/2011: A defect has been found in the SATA controller of the Intel 6-series chipset that will require a recall and replacement of Sandy Bridge motherboards. From the Intel statement:

Intel posted:

SANTA CLARA, Calif.--(BUSINESS WIRE)-- As part of ongoing quality assurance, Intel Corporation has discovered a design issue in a recently released support chip, the Intel� 6 Series, code-named Cougar Point, and has implemented a silicon fix. In some cases, the Serial-ATA (SATA) ports within the chipsets may degrade over time, potentially impacting the performance or functionality of SATA-linked devices such as hard disk drives and DVD-drives. The chipset is utilized in PCs with Intel�s latest Second Generation Intel Core processors, code-named Sandy Bridge. Intel has stopped shipment of the affected support chip from its factories. Intel has corrected the design issue, and has begun manufacturing a new version of the support chip which will resolve the issue. The Sandy Bridge microprocessor is unaffected and no other products are affected by this issue.

The company expects to begin delivering the updated version of the chipset to customers in late February and expects full volume recovery in April. Intel stands behind its products and is committed to product quality. For computer makers and other Intel customers that have bought potentially affected chipsets or systems, Intel will work with its OEM partners to accept the return of the affected chipsets, and plans to support modifications or replacements needed on motherboards or systems. The systems with the affected support chips have only been shipping since January 9th and the company believes that relatively few consumers are impacted by this issue. The only systems sold to an end customer potentially impacted are Second Generation Core i5 and Core i7 quad core based systems. Intel believes that consumers can continue to use their systems with confidence, while working with their computer manufacturer for a permanent solution. For further information consumers should contact Intel at https://www.intel.com on the support page or contact their OEM manufacturer.

Here's an article from Anandtech with more details.

Update 01/05/2011: The NDAs expired early. Here are some reviews:
Anandtech: The Sandy Bridge Review: Intel Core i7-2600K, i5-2500K and Core i3-2100 Tested
Anandtech: Intel�s Sandy Bridge: Upheaval in the Mobile Landscape
HardOCP: Intel Sandy Bridge 2600K & 2500K Processors Review

Here are the original preview articles:
Anandtech: Intel's Sandy Bridge Architecture Exposed
Anandtech: Performance Preview of Intel's Sandy Bridge

On January 9th, Intel will launch their new Sandy Bridge processors, and with them usher in some of the most significant changes to the computing landscape we've seen in years.

The CPU

Oddly enough for a new Intel product, the CPU cores are perhaps the least interesting aspect of Sandy Bridge. The upshot is that Sandy Bridge is about 10% faster clock-for-clock than current Lynnfield Core i5/i7 CPUs using current code, and uses about 10% less power. Sandy Bridge combines the SSE4.2 instructions from the Nehalem Core i7s, the AES-NI instructions from the Gulftown Core i7 hex-cores that drastically improve encryption and decryption performance, along with new Advanced Vector Extensions (AVX) that improve floating point performance. Intel has also introduced a new "Level 0 cache" that caches decoded instructions, improving power-efficiency by saving the CPU from having to duplicate work. Intel has also connected all parts of the CPU together using a high-speed ring bus, which is interesting because it physically sits on top of the L3 cache, like a highway overpass, allowing high-performance connectivity without taking up any room. Turbo Boost has also been enhanced in Sandy Bridge. Previously, Turbo Boost allowed the CPU to overclock itself if some cores were idle, or if all cores were in use but the processor hadn't reached its Thermal Design Power (TDP) yet. Turbo Boost is now allowed to exceed the rated TDP for up to 25 seconds at a time, in order to more quickly finish tasks and return to an idle condition. The Anandtech Sandy Bridge architecture article goes into more detail about these and other changes.

Graphics
Update 01/05/2011: It turns out that only the K-series overclocking edition desktop processors will feature the Intel HD Graphics 3000 GPU with 12 Execution Units. All other desktop processors will have Intel HD Graphics 2000 with 6 EUs. All laptop processors are Intel HD Graphics 3000.

For the first time ever, we have onboard graphics that that are faster than a dedicated graphics card. Sandy Bridge has a DirectX10 graphics core integrated into the processor that handily beats either a Radeon HD 5450 or Geforce G210, spelling the end of the line for low-end GPUs that aren't capable of decent gaming performance. Unlike the Clarkdale processors which had a separate graphics chip sitting next to the CPU on the same package, on Sandy Bridge the graphics core is integrated into the processor die, much like another CPU core. The graphics core also supports Turbo Boost, allowing it to be overclocked to improve gaming performance if the processor cores aren't fully using their power budgets. The graphics core runs at 850Mhz, and can turbo up to 1100-1350Mhz depending on model. There's some uncertainty on the graphics performance for desktop Sandy Bridge chips, as depending on the model, they come with a graphics core that either has 6 or 12 Execution Units, and it hasn't been confirmed how many units the chip Anandtech tested had. Laptop chips all have 12 EUs.

Media Engine
Update 01/05/2011: The media engine has been branded as Intel Quick Sync technology. Unfortunately, using this technology requires that the processor's on-die graphics be enabled and in-use. You can't use it if you have your own videocard. On the plus side, performance is very impressive, and video quality is almost as good as CPU-only encoding (from a crappy encoder, it's no x264). Quality is MUCH improved over GPU-based encoders.

One very disappointing aspect of the graphics integrated into Sandy Bridge is that Intel, being a CPU company, has chosen not to support any general purpose GPU acceleration features, such as OpenCL or DirectCompute. This isn't just a lack of driver or SDK support, the GPU is almost entirely fixed-function, with a bare minimum of programmable functionality. As an alternative, Intel has provided a media processor that performs pure hardware decoding of high definition MPEG2/VC-1 (WMV9)/AVC (H.264) video. In addition, it offers pure hardware encoding of AVC (H.264) video. The idea is that you'll be able to transcode video faster, and with lower power usage, on the CPU using the media engine than you would using a program that supports nVidia's CUDA or OpenCL on a dedicated GPU. This does require that you use a program that supports the Sandy Bridge media processor, but it's expected that support will be integrated into programs like Cyberlink's Media Show Espresso at launch, allowing performance comparisons with CUDA. We also don't know what kind of video quality it will produce, but probably likely that it won't be quite as good as dedicated software encoders like x264.

A New Kind of Overclocking
Update 01/05/2011: Overclocking requires that you have a motherboard based on the Intel P67 chipset. Motherboards that support the on-die graphics (using the H67 chipset) cannot support overclocking. In Q2 Intel will release the Z68 chipset that supports the processor graphics, overclocking, and an unannounced feature called SSD Caching. In addition to the K-series processors that are fully unlocked, the other CPUs that support Turbo mode can be overclocked to 400Mhz beyond their highest supported Turbo speed. Unfortunately if you buy a K-series CPU you have to give up on Intel VT-d virtualization technology (you still get VT-x) and Intel Trusted Execution Technology support, but essentially no one uses these features on the desktop so it's not really a big deal, just annoying. Overclocking your K-series CPU to 4.4Ghz+ on air seems to be attainable for nearly everyone.

A CPU's clock speed is its Front Side Bus (FSB) speed multiplied by a value conveniently called the clock multiplier. On Intel CPUs since the Pentium II, this multiplier has been fixed by Intel at the factory (except on the most expensive Extreme Edition processors). As a result, overclocking has been accomplished by raising the FSB speed, which is generated by a clock generator chip on the motherboard. On the Sandy Bridge processors, this clock generator is now part of the chipset, and its frequency is fixed by Intel and cannot be changed by more than about 5%. This means that overclocking as we know it is now impossible. However, realizing the value of overclocking to enthusiasts, Intel will now be releasing some processor models with an unlocked multiplier. These models will have a letter "K" at the end of the model number, and will come at about a 10% price premium. Additionally, current indications are that those processors which support Turbo Mode (the Core i5 and i7 series) can be forced to run at any of their available turbo bins. While not an end to overclocking, this does mean the days of buying the cheapest processor model and overclocking the heck out of it are now behind us.

Other Platform Features
Update 01/05/2011: The Intel Z68 chipset coming in Q2 supports the processor graphics, overclocking, and an unannounced feature called SSD Caching. I don't know if it supports PCI-Express port bifurcation (splitting the PCI-E x16 into two x8s), but I would expect it does.

Sandy Bridge will use a new socket, LGA-1155, that is not compatible at all with existing motherboards and CPUs. The new motherboards feature Intel 6-series chipsets that support the new SATA600 interface on two of the 6 SATA ports (the other four are SATA300). Like the current Lynnfield processors, the CPU provides a PCI-Express 2.0 x16 connection that can be divided into two x8 connections for Crossfire/SLI, though like with Clarkdale motherboards based on the H6x-series chipsets that support the onboard graphics cannot divide this connection, so are limited to one graphics card. USB3.0 will not be supported by the chipsets, but Intel is considering making a third-party USB3.0 controller part of the reference motherboard design. Intel has upgraded the PCI-E 1.0 x4 "DMI" bus that connects the chipset to the CPU to PCI-E 2.0, doubling interface bandwidth to 2.0GB/sec in each direction. This should improve performance for SATA600 and USB3.0 controller chips on the motherboard, without requiring them to do bizarre things like cannibalize PCI-E 2.0 lanes from the graphics card to get acceptable performance.

AMD's Response

Anandtech: AMD Discloses Bulldozer and Bobcat Architectures at Hot Chips 2010

The Bulldozer architecture is AMD's answer to Intel's Sandy Bridge. Bulldozer is based on "modules" of two integer cores with shared floating point and other hardware. Intel's Hyperthreading Technology allows two threads to share a core for a very small cost in die area, and a performance improvement of about 10% on average. AMD's Bulldozer architecture actually allows two threads to execute simultaneously, for about a 50% increase in die area but with a potential 100% performance increase in the best-case scenario. The upshot is that while Sandy Bridge will be a quad-core processor that appears to Windows as 8 cores and performs like 4.4 cores, Bulldozer will be a quad-module processor that appears to Windows as 8 cores, takes up the space of 6 cores, and performs somewhere between 4 and 8 cores depending on the workload. While it's still too early to draw meaningful conclusions, it seems likely that Intel's solution will be faster and more power-efficient overall, but that AMD's will at least narrow the current performance gap, possibly take the lead in some heavily-threaded workloads, and will target a good price/performance ratio. Bulldozer will also feature an on-die AMD GPU that will likely be significantly faster and more capable than Sandy Bridge's graphics, rumors are that it will be a Radeon HD 5570-class GPU with 400 Stream Processors.

Some Useful Replies and Info:

movax found this sweet table comparing Asus's Sandy Brige boards. Here's his post with more info.
movax also made this useful post about what VT-d, VT-x, and TXT are.
PC LOAD LETTER found this newb-friendly Sandy Bridge overclocking guide.

Somebody fucked around with this message at 03:35 on Jan 9, 2018

# ¿ Sep 14, 2010 14:05

Adbot: ADBOT LOVES YOU

# ¿ Apr 27, 2024 06:35

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Raptop posted:

I hope that isn't a quote from IDF, because it definately isn't true. SandyBridge cache slices sit opposite cpu cores on the ring; i.e. the core and the cache slice straddle the ring interface. There's additional wires to let a core talk to this nearest slice directly without having to emit a ring transaction. The thing I always found interesting was that the 3d graphics unit could also use the L3 cache.

Here's the slide from IDF discussing it. I used the highway overpass analogy not to indicate that it wasn't connected to the L3 cache, but to describe how it is placed on a layer over the top of the cache so as not to consume die area.

# ¿ Sep 14, 2010 15:29

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Mr VacBob posted:

Not only would it be not as good as x264, it won't even be faster. My slower CPU gets the same speed (180fps) for small video sizes using the x264 "faster" preset. A CPU is a much better video encoder than any kind of GPU for all kinds of great reasons; the only reason to avoid one is power budgeting.

It was actually clocked at ~400fps, and that includes decoding, scaling, and encoding simultaneously. Keep in mind also that since this is being done in dedicated fixed-function hardware, there won't be much of a performance impact on the system. You could have a video transcoding in the background while playing a game without the performance of either being impacted. Besides, the current generation of GPU-accelerated encoders look like crap, and most people don't really care if they're just putting videos on their mobile devices or uploading to Youtube.

# ¿ Sep 15, 2010 03:28

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Doc Block posted:

How is a GPU with no shader support anything other than useless? Even the ancient integrated GPU in my parents' computer can do shaders well enough to have Aero Glass turned on in Windows and run the handfull of old games they play.

Even the GPU in my iPhone can do shaders.

The GPU in Sandy Bridge has shader support, it's fully DirectX10 compliant, hence why it can run modern games with decent support, they just made everything that COULD be fixed-function, to reduce power usage and complexity. Here's the page of the Anandtech article with details. It actually comes very close to executing DirectX10 API instructions directly, which is an interesting approach.

Un-l337-Pork posted:

So, if I have an E8400, a Radeon 4870, a Vertex 2 SSD, and 8GB of RAM, what do I need to upgrade next? I play computer games and like to be able to play whatever I want at 1680x1050. I'm thinking that I should buy a 5850 or a 5870 when the price drops a bit?

I'd recommend taking this to the System Building, Upgrading, and Parts Picking Megathread.

# ¿ Sep 15, 2010 16:18

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Doc Block posted:

Kind of annoying then that the article you quoted says the hardware is fixed-function only when they actually just meant no GPGPU support but still supports DirectX/OpenGL shaders.

Anandtech posted:

SNB graphics is the anti-Larrabee. While Larrabee focused on extensive use of fully programmable hardware (with the exception of the texture hardware), SNB graphics (internally referred to as Gen 6 graphics) makes extensive use of fixed function hardware. The design mentality was anything that could be described by a fixed function should be implemented in fixed function hardware. The benefit is performance/power/die area efficiency, at the expense of flexibility. Keeping much of the GPU fixed function is in-line with Intel�s CPU centric view of the world. By contrast, taking the GPU as programmable as possible makes more sense for a GPU focused company like NVIDIA.

The point is that everything that could be fixed-function is fixed-function, but it still has the bare minimum of programmable functionality (the Execution Units).

# ¿ Sep 15, 2010 17:20

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Misogynist posted:

It seems obvious to me that this is targeted at netbooks and similar form-factors, and I don't know why people are putting so much effort into dissecting how horrible it is at replacing their GPGPUs and video encoders.

It's actually for regular desktops and laptops, though the Atom System-on-a-Chips that Intel has been debuting at IDF have the same Media Engine.

# ¿ Sep 15, 2010 17:24

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Ryokurin posted:

I'm interested in Sandy Bridge, but I'm more interested in Zacate as it's benching faster than a i5m chip. It's giving me hope that Llano will be nice hardware.

http://www.anandtech.com/show/3920/amd-benchmarks-zacate-apu-2x-faster-gpu-performance-than-core-i5

http://www.anandtech.com/show/3933/amds-zacate-apu-performance-update

I still think that at least for next year, most of us are still going to want discrete cards, so what I'm shopping for is for HTPCs, which sandy bridge may be a little bit overkill for. if Zacate can handle 1080i without the need for a discrete card I will be overjoyed. I have to ignore all the atoms or most onboard solutions nowadays because no one cares about 1080i.

Yeah Zacate looks to be pretty cool. I wouldn't say Sandy Bridge will be overkill for HTPC applications, as you need a reasonably fast GPU (R5570/GT 220) to do the vector adaptive deinterlacing that makes interlaced content look good. It looks like Zacate will bring that kind of performance to the table, but it remains to be seen what kind of video quality Sandy Bridge will offer. I do think Sandy Bridge will pair extremely well with nVidia's Optimus switchable graphics technology in notebooks, since it will let the dedicated GPU stay completely powered down except when gaming. For some strange reason Optimus implementations so far have tended to use crap GPUs, so having a decent integrated GPU will force nVidia to provide decent solutions to provide some contrast.

# ¿ Sep 15, 2010 17:42

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

One thing to keep in mind is that we're not going to see much development of interesting applications until Windows XP is dead and gone and everyone has switched to Windows Vista/7 64-bit. Without DirectX10 developers can't make applications that take advantage of hardware accelerated rendering via Direct2D, and without 64-bit support applications can't use more than 2GB of RAM, and the system can't have more than ~3GB of RAM total. Thankfully we've finally reached the point where any hardware old enough to not have 64-bit drivers is probably too old to be useful, and adoption of Windows 7 64-bit is growing at a pretty rapid rate. Once developers no longer have to target the lowest common denominator running Windows XP on a Pentium 4, we might see some cool stuff.

# ¿ Sep 16, 2010 05:25

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

MrBond posted:

I thought Handbrake and x264 shun 2-pass encodes in favor of the "quality based" ones now?

2-pass is used if you need to hit a target filesize, otherwise they can only do a moving average bitrate which makes more complex scenes look ugly. It's probably simpler when encoding for portable devices since they're extremely limited on maximum bitrate and available encoding settings anyway, and the resolution is low enough that the files will come out pretty tiny.

# ¿ Sep 16, 2010 19:56

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

I'm still skeptical of the value proposition for Light Peak and its chances for adoption. Costs are going to be astronomically higher than any existing solution, since you have to install an optical transceiver for each lane, rather than just connecting a wire to a pin on a chip. USB3.0 maintains backwards compatibility with an incredible base of installed USB2.0 devices, and is cheap enough to implement everywhere. It's at 5Gbps now, and has the proven ability to scale to 10Gbps, and can probably scale further if desired. A lot of noise is made about having a single connector type for every application, but it's not like we have a huge array of competing connectors now. You have one cable going to your display that can carry video, audio, Ethernet, control data, and soon USB. You have Ethernet for network connectivity (assuming you're not using Wireless), and USB for everything else. Ethernet isn't going away for obvious reasons, and USB3.0 is already fast enough to carry a 1080p60 uncompressed video stream. It kind of sucks that we have competing DisplayPort/HDMI standards, but you can convert DisplayPort to HDMI/DVI using a cheap passive adapter as long as the system has a TMDS connected to the port.

Apple wants Light Peak because it will let them instantly obsolete all of your peripherals, forcing you to buy new (probably Apple-branded) hardware or expensive adapters. Intel is happy to oblige them because they get a sweet royalty check every time someone makes a Light Peak device.

# ¿ Sep 17, 2010 06:12

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

BangersInMyKnickers posted:

I thought GPUs could access system ram as-needed since the AGP days?

At less than 8GB/sec, since that's the interface bandwidth of PCI-E 2.0 x16.

# ¿ Sep 19, 2010 15:43

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Intel is bringing a feature from mainframes to the desktop: software-upgradeable processors. For $50, you can buy a scratch-off code that can be used with the Intel Upgrade Application to enable HyperThreading and an additional 1MB of L3 cache on a Pentium G6951. This seems like a pretty lovely deal, since it basically takes a $100 processor and still doesn't make it as good as a $115 Core i3 processor. If it turned the CPU into a real Core i5 with Turbo and the regular GPU clock speeds, in addition to HT and the rest of the cache, that might be a worthwhile upgrade.

# ¿ Sep 19, 2010 19:44

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

The benchmarked chip also didn't have Turbo Mode enabled, which limits potential performance.

# ¿ Sep 20, 2010 01:00

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Modern games do scale pretty well with higher CPU clockspeeds though, and if you get a good cooler it can be near-silent at anything from stock to a 1Ghz overclock.

Regarding the move to the LGA-1155 socket, I remember reading it was mostly to enable additional monitors to be driven by the on-chip graphics over LGA-1156, but I'm not sure if this is correct.

# ¿ Sep 21, 2010 06:12

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Space Gopher posted:

Again, only if you're interested in making numbers go up rather than improving your subjective experience.

Game performance is almost always bottlenecked on the GPU. Take that bottleneck out of the picture by dropping to low resolution and visual settings, and CPU bottlenecks usually exist over 60fps. That means that, for the vast majority of users, the limiting factor is their video card or monitor. Running at 125fps with a 60hz monitor isn't really useful, unless you're playing Quake 3.

That just isn't true, take a look at these CPU scaling benchmarks for StarCraft II. MMOs like WoW usually scale similarly. Obviously old games are pretty much maxed out, but in newer games you can usually see very significant differences in achieved framerate. Remember also that even if a game averages 60fps, the minimum framerate will be much lower than that, and improving framerates during high-action scenes can noticeably improve how smooth the game feels.

# ¿ Sep 21, 2010 14:10

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

rscott posted:

AMD had better chipsets come out for socket 462 than they did for Slot A so there wasn't much impetus to for enthusiasts to keep their older mobos, and I can't think of any Slot A motherboards with DDR support off the top of my head. The 440BX was probably the best chipset ever made and everything from a Pentium II 233 up to a PIII 1.4 GHz would work with it.

440BX didn't have that much longevity. It was replaced within a year, and obsolete within 2 years as all of Intel's CPUs after that point required a 133Mhz FSB. The 1100Mhz Coppermine P3 was the last supported CPU, the Coppermine-T and Tualatin CPUs (as well as all previous 133Mhz FSB CPUs) required at least an Intel i810E chipset.

# ¿ Sep 21, 2010 19:37

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

PC LOAD LETTER posted:

_{old skool OC chat is awesome but anyone know how much those K series SB's will cost anyways?}

There will be K versions of the i7 2600 and i5 2500. Given this roadmap, I'd expect around $299 and $199 as the base prices, so figure ~10% above that for the K versions.

# ¿ Sep 23, 2010 00:27

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

mew force shoelace posted:

You know what would go a long way towards making computers seem better? A dedicated chip and memory just to run the OS. I'd be pretty tiny and it'd take out a ton of the apparent slowness of a computer.

That wouldn't really help. If you've got a quad-core with 4GB+ of RAM, you're not going to be running low on RAM for the OS or experiencing latency due to CPU load. Granted the Windows scheduler sucks and Explorer does retarded things like lock the interface while waiting for a drive to spin-up, but that's not something you can fix by throwing hardware at it.

JnnyThndrs posted:

Yeah, and there were factory 440BX boards with made expressly for a 133 FSB, with the right multipliers to keep AGP/PCI speeds correct. I had an MSI BX-Master like this - still have it, somewhere.

This is kinda a derail, but no board based on the BX chipset had proper dividers for 133Mhz. Those boards that advertised support were rated to overclock to 133Mhz, but you were still putting a 33% overclock on the PCI and AGP buses. Also, for the record, I didn't post that thing you quoted about Via chipsets.

# ¿ Sep 23, 2010 14:10

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

pseudointellectual posted:

My impression was that stuff like GPGPU was a big boon to low cost high performance computing applications like high frequency trading and scientific computing. What sort of market share do those applications represent and are they enough to sustain the discrete card makers?

The market is still pretty tiny from what I understand, because while there are a lot of theoretical applications, it still takes time and money to rewrite applications for CUDA, and for those applications to get adoption within the industry. It's definitely highly profitable, but there's just not that much volume yet.

quote:

Also does this mean by this time next year that the sweet spot build in the system building thread might not have a discrete card?

For a gaming system a discrete videocard will definitely be necessary, but for a non-gaming system it looks like the integrated graphics will be able to handle basic usage and HTPC applications with good performance. Basically, this will replace the lovely videocards used in low-end systems, but isn't good enough for someone that actually cares about gaming performance.

# ¿ Sep 23, 2010 17:32

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

A couple welcome updates:

Sandy Bridge's 6-series chipsets will have native USB3.0 support after all. Both notebook and desktop platforms will have it, no idea on how many ports unfortunately. Apparently Intel kept this on the down-low because they weren't sure if their native implementation would pass qualification, but it did.

Sandy Bridge graphics details, including branding. Graphics on Sandy Bridge will come in two flavors: the 6 Execution Unit version called Intel HD Graphics 100, and the 12 EU version called Intel HD Graphics 200. HD Graphics 200 will only be available on Core i5 2500 and i7 2600 desktop processors (and presumably higher), though according to previous information all mobile processors will have HD Graphics 200. This does appear to confirm that Anandtech tested the 12 EU version.

# ¿ Oct 7, 2010 12:19

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Remember that UEFI is required in order to boot from HDDs larger than 2TB, and since we currently have 3TB HDDs that are relegated to external storage applications, that's kind of important.

# ¿ Oct 9, 2010 01:09

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

4 Day Weekend posted:

Well right now all 2TB HDDs are 5400RPM, so not exactly something you'd want to have as a boot drive. Still, hopefully 1155/new AMD boards will make UEFI standard.

That's not true, there are 2TB and 3TB 7200rpm HDDs.

# ¿ Oct 11, 2010 07:44

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Triikan posted:

Is this the only reason 3tb drives aren't available except in external enclosures? I'm sure a bunch* of people would still buy them immediately.

*or at least some.

Yeah, there's no reason to make them internal because people would try to boot from them and be pissed when it didn't work.

# ¿ Oct 11, 2010 08:12

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

4 Day Weekend posted:

Really? Which ones?

The Western Digital Caviar Black 2TB for example, the Seagate FreeAgent GoFlex Desk 3TB also contains a 3TB Barracuda XT 7200rpm drive.

# ¿ Oct 11, 2010 13:40

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

According to a guy posting on a Chinese forum (yeah...), a Sandy Bridge Core i7 2600K has been overclocked to 5Ghz (from 3.4Ghz) using air cooling. This is one of the unlocked processors, and the overclock was accomplished by raising the multiplier, at a voltage of under 1.40V. Even if it wasn't stable or he used an unhealthy amount of voltage (1.4v is a lot for a 32nm CPU, especially for long-term stability), this bodes well for stable overclocks in the high-4Ghz range. Current dual-core Clarkdales hit ~4.15Ghz with stock cooling and 4.5Ghz with water, and the six-core Gulftowns have hit 4.13Ghz, so this isn't out of the range of whats possible.

# ¿ Oct 26, 2010 22:15

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Kerris posted:

How does AMD's Fusion compare to Sandy Bridge? Is Intel closer to delivering Sandy Bridge?

It appears likely that the graphics will be faster but the CPU will be slower. Sandy Bridge is a bit more than 2 months out, I think AMD's Bulldozer+APU products are further than that, they're launching the netbook/notebook products first.

# ¿ Oct 26, 2010 23:39

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

WhyteRyce posted:

It should make the netbook market interesting again at least, although I've never understood the fascination people have with wanting to play games on a netbook.

You need decent hardware acceleration to watch Youtube videos though, and it doesn't seem too unreasonable to want to play Flash games or Minecraft on a netbook, which may actually be possible with Ontario.

# ¿ Oct 27, 2010 01:10

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Cryolite posted:

How much of a decrease in boot times do you think we'll see from UEFI vs. BIOS?

In theory all of the time between when the first POST image appears on your monitor and the Windows logo can be eliminated, especially if you're using an SSD and don't have to wait for it to spin up. That's 10+ seconds on my system.

# ¿ Nov 4, 2010 05:40

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

dud root posted:

Are the Socket 2011 quad channel CPUs classed as "sandy bridge"?

Yes, but they won't have the on-die graphics.

# ¿ Nov 5, 2010 03:08

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

They're for multi-socket servers and high-end workstations, so not something you'd be considering for your system. That's what the LGA-1155 is for.

# ¿ Nov 5, 2010 03:41

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

LGA-1156 platform CPUs, such as the i5 700-series and i7 800-series, outperform the LGA-1366 i7s because they have higher-performance Turbo Modes (+533Mhz for dual-core operation, vs +133Mhz on the i7 920) and the platform is more efficient overall. I would expect an i5 2500 (3.3Ghz) to be around 25-40% faster than an i5 760, given the significant clockspeed and architectural advantages.

Your link seems to show DoWII NOT scaling significantly with cache size, so I'm not sure why they said it did. They didn't benchmark any CPUs that are comparable except for cache size, if you look at the numbers for the E6600 and the E2160 (1/4 the cache), there's around a 10% difference after you scale the clock speed, without accounting for the bus speed, which seems a pretty tiny difference for an app that likes cache and a 75% cache size reduction.

# ¿ Nov 7, 2010 13:09

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Fats posted:

Wait, what makes it more efficient? The overclocking benchmarks I've seen, with both chips at the same speed, generally put the X58 ahead (there are also the SLI/Crossfire benefits with X58, but that's for crazy people so it doesn't really count).

It has an on-die PCI-E controller and the memory controller supports higher clockspeeds. The LGA-1366 platform has an extra memory channel, but that's not really beneficial outside of data compression applications. The improved power efficiency also enables more aggressive Turbo Modes.

KingEup posted:

They compared the Core 2 Duo E8400 and Intel Core 2 Duo E6850. Same clock speed. Different cache size. 10% faster with 2MB more.

Yeah but the E6860 is a 65nm Conroe CPU, the E8400 is a 45nm Wolfdale CPU. There's a lot more going on there than the change in cache size. If they had thrown in a Q9500 for example that would have been diagnostic.

# ¿ Nov 8, 2010 09:03

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

4 Day Weekend posted:

Do we have a solid release date? A friend of mine wants to buy a PC, trying to convince him to wait for it.

Probably early January, during the Consumer Electronics Show (Jan 7-11). Officially Intel is just saying "Q1" right now, though they've released forecasts expecting that Sandy Bridge will make up 20% of desktop CPU shipments in Q1, which indicates an early and heavy launch. Board manufacturers are demoing complete product lines of LGA-1155 boards now, so it can't be that far off.

# ¿ Nov 11, 2010 03:45

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Fudzilla is reporting that Apple will switch to AMD Fusion processors in their upcoming products (update with additional confirmation here). This is probably a direct response to Sandy Bridge's on-die graphics not supporting OpenCL, as Apple has committed to shipping every computer with support for GPGPU acceleration. On products without dedicated GPUs, they do this by pairing an older Core 2 Duo processor with an nVidia Geforce 320M chipset that provides CUDA and OpenCL support. Since the Core 2 Duo is getting old and nVidia won't be making any more chipsets, they pretty much have to switch to AMD unless they're willing to put a dedicated GPU in every product (including the Macbook Air) or abandon their GPU acceleration plans.

# ¿ Nov 13, 2010 05:15

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Good news! Based on leaked Sandy Bridge pricing from Sweden that SemiAccurate posted, the multiplier-unlocked K-series CPUs won't be significantly more expensive than the regular version. The Core i5 2500 is $257 while the i5 2500K is $270, a difference of only around 5%. For the Core i7 2600/2600K the price is $363, or $387 unlocked. These are Swedish prices not counting tax and converted to USD, so take that as you will.

# ¿ Nov 24, 2010 03:14

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

And here's an article with the supposed launch pricing in US$. It's consistent with the Swedish prices we saw earlier, but lower, and in-line with current Lynnfield pricing.

# ¿ Nov 26, 2010 19:27

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

demonachizer posted:

These are wholesale prices correct? One would assume they would be much more expensive on the retail level or no?

They are wholesale, but that's the price you pay for CPUs if you shop somewhere like Newegg that doesn't add a ridiculous markup.

# ¿ Dec 1, 2010 22:44

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Dr. Gaius Baltar posted:

I'm planning on buying a 2500K when it's released on January 5th, just in time for the big early January holiday sales season. I wonder what are the chances that I'll end up regretting not getting Bulldozer over this. Say, if Bulldozer is released in June 2011 and has 30%+ more price/performance.

I doubt you'll end up regretting Sandy Bridge over Bulldozer. Bulldozer brings 8 cores to the table, but it's unlikely that AMD can close its clock-for-clock, core-for-core performance gap with Intel, and Intel's Turbo Mode is extremely effective at making highly-threaded CPUs very fast with weakly threaded workloads. It's certainly possible that AMD could execute shockingly well and beat Sandy Bridge in terms of CPU performance, but I doubt it.

movax posted:

Should I buy DDR3 now or wait until January?

Wait, prices are still dropping pretty steadily and will be for the near future.

# ¿ Dec 8, 2010 03:04

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

Ryokurin posted:

Maybe I'm missing it, but where? I've seen limited 2gb pairs for $50 on sale in the past week or so, but I could have swore that this was the normal price this summer. Most I see is still in the $70-80 range for 4gb. Even the deal I got from newegg went up to $60 the next week.

The DDR3 DRAM market is in a steady price decline that isn't expected to end until the first or even second quarter of 2011. Granted wholesale DRAM price isn't the price you pay at Newegg, but they're related like the price of gas is to the price of oil. In the first half of the year the wholesale price of 2GB of DDR3 was $46.50, by the beginning of November it was down to $25, and the price is expected to fall to $20 by the end of the year.

# ¿ Dec 8, 2010 15:32

Adbot: ADBOT LOVES YOU

# ¿ Apr 27, 2024 06:35

Alereon: Feb 6, 2004; Dehumanize yourself and face to Trumpshed; College Slice

The coolers do share the same mounting holes so should fit.

# ¿ Dec 10, 2010 23:45

The Something Awful Forums > Discussion > Serious Hardware/Software Crap > Intel: lol

«‹›12 »