We took our prior plateau of 4.4GHz at 1.465V as a starting point for our renewed overclocking attempts. We were soon able to reach 4.5GHz and then 4.6GHz at 1.525V, which was encouraging. However, getting the chip stable at 4.7GHz required more voltage, forcing us to ratchet things up to 1.55V, the peak value exposed in AMD's Overdrive utility—and a heckuva lotta juice for a 32-nm processor. Once at that voltage limit, we tried for 4.8GHz, but we quickly saw errors in Overdrive's stability test.
So, 4.7GHz was it. That's 1.1GHz above the FX-8150's base clock, but only 500MHz beyond its peak Turbo Core frequency. Still, it's not far from the projection in AMD's press literature, which says AMD's internal attempts with water cooling topped out at 4.9GHz (presumably with more than one chip on hand). 4.7GHz is also a couple of hundred megahertz higher than what we've seen from Intel's Sandy Bridge and Gulftown, in our limited, air-cooled overclocking exploits with those chips.
Even at 1.55V, we weren't pushing the FX cooler past it limits. During stability testing, CPU temperatures topped out at around 54.5° C, with an ambient room temperature of about 74° F/23° C. Yeah, the thing was incredibly loud—I'd give you a decibel number, but we're currently having the roof replaced, and I don't want to harm the workers' ears. Er, I mean, all of the hammering would throw off the measurements. Still, the cooler itself could have taken more heat, had we needed it.
At those overclocked settings, our FX-8150's power draw rose considerably from its stock levels. We did a quick measurement, in fact, and it came back like so:
That, my friends, is why AMD didn't push any higher than it did on FX clock speeds. Cranking up the CPU voltage does bad things for power consumption. Although our motherboard, PSU, and cooler could apparently handle it reasonably well, an FX processor at these speeds goes well beyond the top established PC power envelope of 125W. Even resurrecting the old 140W power window probably wouldn't have bought AMD much more in terms of frequency. There is headroom in this chip, but you pay for it dearly in wattage.
Oddly enough, the benchmarks we selected months ago for our overclocking performance tests seem to be pretty well suited to the Bulldozer architecture. Thus, turning up the clock frequency allows the FX-8150 to put up some really nice numbers, tying or beating a Core i7-2600K overclocked to 4.5GHz in several cases. There are some pain points here, such as the difference in single-threaded Cinebench performance between the FX-8150 at 4.7GHz and the Core i5-2500K at stock (scores of 1.16 vs. 1.48, respectively). Still, had Bulldozer landed at frequencies north of 4.5GHz within conventional power envelopes, the competitive landscape might look rather different. Indeed, if GlobalFoundries can manage to refine its 32-nm fabrication process to allow such speeds in the coming months, who knows?
For now, thanks to a formidable bundled cooler, those folks who bleed AMD green (or is it red now?) will have an option for achieving bragging-rights-type performance in some cases, so long as they're willing to pay for it in the form of added heat, noise, and power draw.
147 comments — Last by Draphius at 12:45 AM on 11/01/11
|Intel boosts the high-end desktop with its Broadwell-E CPUs More cores, more threads, more money||89|
|Intel's Core i7-6700K 'Skylake' processor reviewedEnthusiasts get the first taste of a new architecture||489|
|AMD's Carrizo brings power savings to mainstream laptopsExcavator and GCN combine at 15W||83|
|Intel's Broadwell goes broad with new desktop, mobile, server variants14-nm chips for everyone||167|
|The TR Podcast bonus video: AMD, Zen, Fiji, and moreWith special guest David Kanter||54|
|AMD: Zen chips headed to desktops, servers in 2016Details of its new x86 CPU and plans revealed||250|
|Inside ARM's Cortex-A72 microarchitectureThe next-gen CPU core for mobile devices and servers||42|
|Semiconductors from idea to productThe story of how chips are made||56|
|Intel Computex keynote confirms Kaby Lake and Optane for 2016||27|
|Asus shows off Avalon modular case and GX800 liquid-cooled laptop||6|
|Samsung designs minuscule single-package NVMe SSD||21|
|Thermaltake shows off The Tower and more at Computex||9|
|Adata shows NVMe and TLC SSDs at Computex||1|
|Corsair@Computex 2016: fans that levitate, fans that illuminate||8|
|Patriot adds 2TB model to Ignite SSD lineup||13|
|Intel boosts the high-end desktop with its Broadwell-E CPUs||89|
|EVGA@Computex 2016: Custom Pascal cards, new PSUs, and more||9|
|Everyone from Asus to Zotac has announced a non-reference GTX 1080. I see what you did there!||+46|