Tag: overclocking

AMD HD 7770 details leaked

A freshly leaked GPU-Z screenshot has enabled us to join all the dots and work out the specifications of the next AMD desktop card, the mid-range HD 7770.

The card is on the verge of release. It looks like it will be based around the “Cape Verde” GPU and will have a lesser number of better-configured Graphics CoreNext stream processors. This will decrease transistor counts and power draw while providing better performance.

The GPU-Z screenshot shows that there will be a GCN stream processor count of 640, 40 TMU’s, 16 ROPs and a 128-bit wide GDDR5 memory interface.

This will make the HD 7770′s core clock rate 1GHz out of the box unlesss it is factory overclocked.

The GDDR5 memory runs at 1125MHz or an effective 4500MHz and can give 72GB/s of memory bandwidth. A typical board rating would be 80W.

It is starting to look like the price will be around the $190 mark, where the HD 7750 will be somewhere around the $140 mark. These are pretty good prices for what you get, although the spec is a little bit of a yawn. 

Intel launches overclocker insurance

Intel must have noticed some of its geekier customers are burning their systems out through overclocking, because it has launched an insurance scheme for its enthusiast market.

Now, overclockers can buy a $20 plan which will extend beyond the standard three year warranty. In Intel’s own words: ” if it fails under normal usage, we will replace it under the standard warranty; if it fails while running outside of Intel’s specifications, we will replace it under the Performance Tuning Protection Plan.”

It appears to be a one-time replacement. In Intel’s FAQ, here, it claims that despite the offer, the company is not supporting or encouraging overclocking. Customers can’t buy multiple plans for one processor, either. 

Processors covered in the plan are all unlocked processors with an X or K suffix, and any processor in the LGA2011 socket, beginning with 2nd Generation Core products. 

Unfortunately for all the Intel overclocking enthusiasts in Syria, Burma, Sudan, Cuba, Iran and North Korea, the plan will not be available in those countries for legal reasons. 

Gigabyte declares X79 clock speed record

Gigabyte has announced some record speeds on Intel’s i7 3930K processor, becoming the first to reach a multiplier of 57x.

Gigabyte will be happy to win back its reputation after YouTube footage showed its X79 motherboards burning out, before a BIOS change was issued.

Now the mobo manufacturer is eager to show how it has been able to hit record CPU speeds, with the help of renowned overclocker Hicookie.

Using the recently released F7 BIOS, and armed with a load of liquid nitrogen for cooling, Gigabyte hoped to “erase any scepticism regarding the performance and overclocking capabilities”.

According to a statement made by the firm, rather than promptly bursting into flames, clock speeds of 5.6GHz were reached on the i7 processor, using a 57x CPU multiplier configuration.   Using the GA-X79-UD3 motherboard record scores were managed for Super Pi 1M and 32M and PiFast benchmarks.

The X79-UD3, designed for mainstream enthusiast users with i7 chips is based on a 3-way Digital Engine, and features 4-way SLI and Crossfire X graphics, PCIe Gen 3.0 support and 4 channel DDR3 memory. The firm noted that F7 BIOS updates are available for its entire X79 series of motherboards too.

AMD was let down by GlobalFoundries

Fabless chip outfit AMD has had to lower its sales guidance because its chum GlobalFoundries (GloFo) can’t cut cut the mustard with its next-generation, 32-nanometer Llano processors.

Word on the street is that Globalfoundries has not got its 32nm yields high enough and will not do so until after the end of the Mayan calender in 2012.

For AMD this is the sort of headache where you have to shut yourself in a darkened room and hope the kid in the next flat stopped hitting his toy truck against the wall and screaming.

Supplies of AMD’s 32nm accelerated processor units (APUs) were supposed to be flowing like the toilets in the Rat and Handgun until the company’s Trinity platform appears in the first half of 2012.

Unfolding AMD’s product roadmap, Trinity APUs will succeed the Llano generation’s A8 series of performance desktop products, perhaps as early as the first quarter of 2012. Then Weatherford and Richland will come later and replace Llano chips aimed at the mainstream segment.

Those three platforms will come with Piledriver, which uses AMD’s post-K10 Bulldozer architecture and was supposed to tackle piles once and for all .

AMD is not saying much other than there are lower-than-expected volume production of AMD’s 32nm APUs.

According to PC Mag,  GloFo was having a bad case of trouble with yields not improving at the pace both the foundry and AMD had wanted. It is also taking too long for GloFo to get its product to AMD.

AMD was hoping that it could be aggressive with Llano and push the 40nm Brazos APUs for ultrathin notebooks introduced around the same time. However, the move to the 32nm node hadn’t happened fast enough.

What is pulling AMD’s nadgers out of the fire is that it is getting good results out of its 28nm Bobcat-based APUs which are supposed to be heading for the ultra-portable market in the first half of 2012. Some of that production though is coming from TSMC.

AMD is probably kicking itself for not hanging on to TSMC and ask it not to close its 32nm process node in 2009 to focus on 28nm high-k metal gate bulk fabrication. 

AMD Trinity ready for first quarter 2012

A leaked AMD roadmap appears to show that Llano’s replacement, Trinity, is going to arrive in Q1 next year.

The roadmap tipped up on the 3DCentre forum and it shows that AMD is following the same Llano strategy with Trinity.

There will be four different main SKUs, although at the moment AMD is not telling anyone what the specific models are.

Trinity will be based on AMD’s new Pilderiver core which is derived from Bulldozer, and will be the same core that will be jacked under the bonnet of the AMD FX series.

Dubbed Comal, initially it’ll co-exist with the Sabine platform which is the current mobile Llano APUs combined with the A70M and A60M chipsets. Comal will use the same chipsets and socket.

Apparently there will be some improved graphics, which are codenamed London and the TDP will remain the same as for the current APUs.

Later there will be the release of Wichita and Krishna SKUs which are part of the Deccan platform. These will feature a Bobcat core, but the graphics core be the same as for Trinity.

AMD appears to have developed a new FCH for Wichita and Krishna called Yuba. Yuba runs two SATA 6Gbps ports, two USB 3.0 ports, 10 USB 2.0 ports, 4 PCI Express x1 lanes, under a SDHC controller and various other basic chipset features. Yuba is expected to kill off the A50M FCH in this product category. The roadmap suggest that the Deccan platform should launch sometime in Q2 next year.

In the “ultra low power” segment AMD is releasing Hondo, which is a next gen Desna APU which brings the power down from 5.9W to 4.5W.

This Hondo uses two Bobcat cores and Loveland graphics. The platform will be known as Brazos T and will probably end up in tablets. There is a new FCH called A55T. The A55T is a small and simplified FCH with a single SATA port, SDIO support and eight USB 2.0 ports.

Looking ahead to 2013 Trinity will be replaced by Kaveri which will feature Steamroller CPU cores and an unknown graphics core. Below that there will be the Kabini with its Jaguar cores and the tablet friendly Samara APU.

While we have heard rumours that AMD is speeding up its Trinity releases, it is not clear if it will manage this roadmap. This year it has been slow at meeting its targets and this leak might just be something someone shoved in the rubbish bin because it was unworkable. 

AMD announces Interlagos and Zambezi processors

It looks certain that AMD is going to launch server-class “Interlagos” and consumer-class “Zambezi” processors, based on new Bulldozer micro-architecture in two to three weeks.

Three “Zambezi” FX-Series CPUs surfaced for pre-order on online US stores, suggesting an imminent release of these models.

Turkish site Donanimhaber claims that it has found out that the AMD FX-Series family will launch on October 12. This would confirm previous rumours that the chip will be introduced in the middle of October, presumably on October 13.

Donanimhaber said that the prices of the first three FX-Series chips will be from $175 for FX-6100 and up to $245 for FX-8150. The price of a FX-8120 model will be $205.

However, it was not clear if these were retail or official prices for OEM processors in. The numbers appear to be a bit on the high side which suggested that these are retail prices, especially if they are being positioned against Intel Core i5 and i7 chips.

Quad-core FX-4100 series CPUs are expected to appear later in the year.

The range is interesting, not just because it is designed to give Intel’s i5 and i7s a run for their money, but because the FX currently holds the record for the world’s fastest CPU. The beast was overclocked to 8.429 GHz, winning AMD a place in the Guinness World Records. 

AMD overclocking record is marketing coup

AMD has gotten itself into the record books after overclocking a Bulldozer chip to make it the world’s fastest processor.

The marketing team at AMD will no doubt be rubbing their hands in glee after securing an entry into the Guinness Book of Records.The processor overclocked was the forthcoming eight core FX chip, and we are expecting shedloads of spin about this beastie around it’s release this Autumn.  The official title of “world’s fastest chip” will no doubt be used and abused in the coming months.

‘Team AMD FX’ got the chip to hit speeds of 8.429 GHz without bursting into flames or melting a hole through to China.  This beat the previous record of 8.308 GHz.

AMD used liquid helium to cool the chip to near absolute zero, a technique even the most avid overclockers may struggle to replicate in their own homes.

As a slight warning, spraying your processors with liquid helium at minus 180 degrees in order to reach similar speeds might just void your warranty.  It might also turn your cat into a statue and make your voice sound like Mickey Mouse.

Apparently the same team were able to hit speeds of around 5GHz using only a cheap water cooling system which cost only $100, a more realistic option for overclockers without access to liquid helium.


AMD's A8-3850 APU unlocked multiplier useless

AMD’s new A8-3850 APU has been shipping with an unlocked multiplier which is the sort of thing that gets overclockers moist.

However OCWorkbench  has been playing around with it and they say that it is absolutely useless. You can raise its default 29x value will not give you any performance gain.  One Chinese site reportedly has been posting a  CPU-Z screen shot of A8-3850 at a whopping speed of 5.6GHz (100×56)

In fact, if you want to look really cool, you can leaving the base clock (BClk) at the default 100 and push the multiplier to achieve some huge numbers.

It will make no difference to the chip which will chug along as well as it ever did. Some are wondering why AMD did not bother to lock it down.

We think that is an attempt at psychosomatic overclocking.

Overclockers are a bit like hi-fi experts who believe that they can tell the difference if they spend all afternoon making sure that their chip can go five percent faster.

AMD seems to have worked out that all you need to do is give them the software to think it is faster and they will convince themselves. 

Alienware caught overcharging for Intel overclocking

Dell’s gaming rig makers Alienware might have started thinking that its users are willing to part with a lot of cash for not much.

The outfit has started flogging the Intel Core i7-2600K ‘Sandy Bridge‘ processor and is offering to overclock it from 3.9GHz to a blistering 4.1GHz. The only downside is that this upgrade will cost you $250.

This means that those who have a need for such speed will have to write a big cheque. Well, according to Trusted Reviews  not really.

It says it was able to overclock the same processor to 4.7GHz and beyond with an Intel reference board.

While other enthusiast system builders like CyberPower and MAINGEAR also charge for overclocking, the amount they want is a lot less. CyberPower charges $19 for a 10 per cent overclock, $49 for 20 per cent and $99 for 30 per cent.

But all of them end up at a higher clock frequency than what Dell is offering on the same processor and a lot cheaper. MAINGEAR charges $49 to max out the system you bought to a level that they deem safe by checking system temperatures.

Quite why Alienware is charging so much for a job that many of its own enthusiasts could probably manage on their own is anyone’s guess. Maybe they have looked at the Apple business model and thought “I’ll be having some of that”.

IBM sets world clock-speed record

Biggish Blue boffins at the Hot Chips conference showed off what IBM claims was the world’s fastest CPU clock-speed.

IBM managed to get a quad-core z196 running at 5.2GHz. Currently the world’s clock-speed record was held by IBM with a 4.7GHz speed hit by the Power6 CPU.

The latest chip has 1.4 billion transistors on a 512mm2 board built using the 45nm manufacturing-process. The old Power6 CPUs had only 1.1 billion transistors.

IBM was telling the world+dog that each core had been given 1.5MB private L2 cache as well as access to 24MB shared L3 eDRAM cache.

The z196 had access to a fourth layer of cache off-chip. IBM said that six of these CPUs can be installed into a multi-chip module (MCM) and connected to each other through two controllers with a total of 192MB shared L4 cache via a 40GB per second link.

The extra memory is needed to make sure the data runs smoothly because the processors are so fast.

They also use a CISC-based z/Architecture which is not much like the more common RISC ISA.

Biggish Blue has a cunning plan to stick the z196 under the bonnet of its Z-series mainframes which should begin shipping next month. So this probably means that you will never get to try out a shoot-em-up on the beast. 

Hexus pointed out that the Chip has been announced since July, but this is the first time that Biggish Blue has provided any details about its clock-speed, or shown what it can do.