Friday, July 10th 2009
AMD Staring at 140W Barrier with Phenom II X4 965?
Two of AMD's biggest setbacks with the 65 nm Phenom X4 series were 1. the TLB erratum fiasco with the B2 revision of the chip, and 2. the virtual TDP wall it hit with the 2.60 GHz Phenom X4 9950, at 140W. At that wattage, several motherboards were rendered incompatible with the processor because they lacked the power circuitry that could handle it. The company eventually worked out a lower-wattage 125W variant of the said chip, and went on to never release a higher-clocked processor based on the core.
MSI published the complete CPU support list of its a new BIOS for the 790GX-G65 motherboard a little early, revealing quite some about unreleased AMD processors. At the bottom of the list its the Phenom II X4 965. This 3.40 GHz quad-core chip will succeed the Phenom II X4 955 as AMD next flagship desktop offering. Its TDP is an alarming 140W. Alarming, because this is a chip with a mere 2 unit bus multiplier increment over the Phenom II X4 940, the launch-vehicle for AMD's 45 nm client processor lineup. There are, however, two things to cheer about. RB-C2 is not going to be the only revision of this core, future revisions could bring TDP down, or at least make sure clock-speeds of future models keep escalating, while respecting the 140W mark. A future variant of Phenom II 965 could come with a reduced TDP rating. The list interestingly also goes on to reveal that AMD will have a 95W version of the 3.00 GHz Phenom II X4 945.
Source:
HardwareLuxx.de
MSI published the complete CPU support list of its a new BIOS for the 790GX-G65 motherboard a little early, revealing quite some about unreleased AMD processors. At the bottom of the list its the Phenom II X4 965. This 3.40 GHz quad-core chip will succeed the Phenom II X4 955 as AMD next flagship desktop offering. Its TDP is an alarming 140W. Alarming, because this is a chip with a mere 2 unit bus multiplier increment over the Phenom II X4 940, the launch-vehicle for AMD's 45 nm client processor lineup. There are, however, two things to cheer about. RB-C2 is not going to be the only revision of this core, future revisions could bring TDP down, or at least make sure clock-speeds of future models keep escalating, while respecting the 140W mark. A future variant of Phenom II 965 could come with a reduced TDP rating. The list interestingly also goes on to reveal that AMD will have a 95W version of the 3.00 GHz Phenom II X4 945.
184 Comments on AMD Staring at 140W Barrier with Phenom II X4 965?
And as far as SLI needing hardware, google it on X58 my friend. It's purely a BIOS string. No hardware needed, unless you want 16,16,16x Tri SLI instead of 16,16,8. None of the SLI boards, and I repeat, NONE of the SLI boards that support only 16,16,8 or lower have the NF chips on them.
The "hacks" only worked up to nForce 4 and 500 series. From that point on, something changed. Were these original hacked drivers merely a corporate-sponsored software version (basically like Crossfire) or was SLI purely software on these chipsets? After nForce 500 series, something changed.
X58 already has the necessary hardware to run SLI (x16, x8, x8). The BIOS string determines whether or not the functionality is reported. Ehm, just like how nForce 600 series chips is the hardware for SLI on those platforms, X58 is the hardware for SLI on its platform. NVIDIA licensed the technology so Intel could embed it.
Obviously, if the manufacturer wants more lanes than that, they'll have to supplement the X58 chip.
Maybe if a virtual-machine software ever emulates a Core i7 + qualified X58 motherboard, we'll get to see that first hand on any machine? It's a big maybe.
Anyway, we're going south of the topic. Nice discussion, can resume it elsewhere.
And, I'm not sure if you remember, but when X48 first release, I believe either Falcon Northwest or VooDooPC released a system with SLI on an X48 board(and then later Crossfire on a 790i board). The only thing that was custom, was the BIOS and drivers. There were no hardware changes at all.
Edit: I Googled "SLI patent" and came up with this:I doubt they are talking solely about the SLI bridge there.
It's all gray area...intentionally. :(
Nevermind, that's on a thousand sites, word for word, as part of an NVIDIA statement when it launched.
Edit: There is no filed patents. :(
But the skyscraper large fact that they f-k-d us in favor of cutting down server chip -Istanbul core R&D they really did. They produce some Deneb core that consumes almost or even more than 50% of all that power (ACP) caused with enormous L3 cache thats nothing more than marketing hog .... Deneb==Nehalem marketing fables. and we see how great it should perform with just 2M of L3 cache or maybe 3M only and power consumption would be 100W instead 140W.They couldnt lie too much cause power VRM circutry couldnt handle too much more than 130W in those days. Even p4D need special boards for that double precotts (smithhhyfields iirc) just to power up. And canceled Tejas should be a real power hog with 150W on idle 2.80GHz :D :DPlease dont be such a cry baby. And self explanatory rocket scientist in the same post.
Current CPUs are made on Silicon with all that Ge/Hf and so on. And yes pure silicon could easily run on 150°C Tj. But not the processing units. Most of old CPUs (P6,K7,P4-prescott) work well on 95°C and it wasn't some sweating sauna for them. Just other elements on mobo sufferd (like capacitors that were overheated by overburdened VRM and poorly cooled CPU)
Most modern GPUs nv7800->gtx280 works great @100C ntm all that new ATI rv670/rv770/rv790/rv740 that in most cases if power supply on card is properly designed could work 120°C for a few hours of FurMark stress test. Yep they'll consume alot but they wont fail while power circuitry can :cool: And the best thing is that GPUs are based on pure silicon process that they call Bulk and not so fancy SiGe, streched-strained, with HighK gates, or just poor old SiO base.OMG. Unfortunatly they don't reach anynthing with current architecture. They reach their thermal envelope @140W and ~3.6GHz with old F3 revision on last Athlon FX core. Since then they just buying time and delaying real R&D in the desktop/workstation segment. And buying themselves a lot of time. But at least they stamping more and more cores with advanced processing for server cores like Istanbul. Not that helps us at all. But we get TWKR nuclear fusion hog. yupiiii. we really need to worship AMD golden bull now.