Core 2 Duo E6750 review

Evan Gower October 7, 2009

£101

Price when reviewed

Intel has given its Core 2 range a mid-life makeover, with new CPUs receiving a hike in clock speeds and bus frequencies. It’s likely to be the last refresh at the top end before the release of the new Penryn-microarchitecture parts towards the end of the year. Penryn-based processors will be the first produced with a 45nm fabrication process; the new CPUs remain on 65nm. In theory, they’ll work in 965- and 975-chipset motherboards, but there’s actually no benefit in doing this – the headline feature is higher FSB (front side bus) speeds, which will require the new P35 chipset to reap the rewards.

New Extreme Edition

The fastest new release is at the very top end of Intel’s enthusiast CPU line-up: the Core 2 Extreme Edition QX6850, replacing the QX6800 at the pinnacle of the food chain. A quad-core processor, it’s the third Core 2 Extreme model to be released since the original quad-core QX6700 at the end of last year (web ID: 100680). The new part’s basic specifications haven’t leapt massively, starting from 2.66GHz in the QX6700 and now hitting the 3GHz mark. The architecture remains identical, with two dual-core dies in one package, each with 4MB of L2 cache, making for a total of 8MB.

The major difference with the new release is its support for a 1,333MHz FSB frequency, up from a previous maximum of 1,066MHz. In theory, this represents a substantial 25% increase in bandwidth between the CPU and the rest of the system, which in particular should help memory performance with DDR3 RAM. Aside from its sheer speed, the QX6850 – like all other Extreme Edition Intel CPUs before it – is clock unlocked, allowing for direct clock-multiplier overclocking (standard Core 2 processors can be overclocked only indirectly by hiking FSB speeds).

As a general rule we’ve found you can overclock an Extreme Edition by one “speed bin” – our review sample ran perfectly happily with the clock multiplier increased from 9x to 10x, boosting it to 3.33GHz. And, in fact, a little more FSB tweaking got it up to 3.4GHz, running at 10x clock multiplier and 340MHz FSB (remembering that FSB speed is quad-pumped, giving an effective 1,360MHz), all with a stock cooler, giving us the fastest benchmark result ever of 2.04. At its standard clock, it still managed a mightily impressive result of 1.84.

New top-end dual core

Released at the same time as the QX6850 is a new standard Core 2 dual-core chip, the E6750. This is the new top-end standard part and comes in enormously cheaper than a QX6850 – not far from a sixth of the price, in fact. Even given the fact that you’re getting two fewer cores, it’s a hefty difference and betrays Intel’s determination to beat AMD on price when it comes to the volume end of the market. The new part runs with a 1,333MHz FSB, with a clock speed of 2.66GHz (the same as the existing E6700).

In comparison with the QX6850, the performance isn’t an open and shut case – with single-threaded code, the extra cores of the quad parts sometimes have little effect. With the majority of games, for instance, the E6750 will be within a whisker of the QX6850’s performance. In our application benchmarks, however, the overall difference is clear – a score of 1.43 for the E6750 against the QX6850’s 1.84 when fitted to the Gigabyte GA-P35T-DQ6 using 2GB of Kingston’s DDR3 HyperX RAM. But delving into the individual breakdowns shows that the gap closes with applications such as Office that aren’t extensively multithreaded; Office scores a figure of 1.42 against 1.53.
Given the ridiculously low price of this processor, you’d have to think long and hard before choosing a quad core instead – it’s incredibly good value for money.

DDR3: the untold story

At the grand age of three, DDR2 RAM is set to be replaced by DDR3 – not to be confused with GDDR3, a variant of DDR2 popular in graphics cards. DDR3 offers similar benefits, though: operating voltage is reduced from DDR2’s 1.8V to 1.5V, and while DDR2 officially supports a maximum I/O bus rate of 533MHz, DDR3 goes up to 800MHz – effectively 1,600MHz due to DDR’s double-pumped bus. The pre-fetch buffer is also doubled, from 4 bits to a whole byte.

These improvements are now offset by increased CAS latency, though. Most DDR2 DIMMs have a latency of 4 or 5 clocks before they can start to return the data stored at a given address. Current DDR3 modules, meanwhile, have a latency of 7-9 clock cycles, and while this may fall as the manufacturing process is refined, the standard dictates an absolute minimum CAS latency of 5 clock cycles for DDR3.

Intel’s P35 chipset supports both DDR2 and DDR3 memory, but the two technologies demand different interface hardware: DDR3 memory won’t even fit into a DDR2 slot. To take advantage of the new DIMMs you’ll therefore need a DDR3-capable motherboard. As happened with the move from DDR to DDR2, we expect manufacturers will offer motherboards with both types of slot to ease the transition, but we haven’t seen one yet. DDR3 is also, for now, expensive, with DDR3 modules costing more than five times as much as DDR2 DIMMs of the same capacity and rated at the same speed.

DDR3 vs DDR2

Is the performance boost worth the expense? To establish a baseline, we first ran our Real World Benchmarks on a P35-based system, based around a 2GHz Intel Core 2 Duo E6750 equipped with 2GB of 1,066MHz CAS5 DDR2 RAM in a dual-channel configuration.

We compared these results with the scores achieved using a range of DDR3 DIMMs, using the same processor and graphics card installed in the Gigabyte GA-P35T-DQ6 motherboard.

For our first DDR3 test, we used two 1GB Kingston Value RAM DIMMs, rated at 1,066MHz with a CAS latency of 7 clocks. This is a fairly specialist use of the term “Value RAM”, as a 2GB pack will set you back around £190 (£224 inc VAT), but that’s still £40 less than the Kingston HyperX modules, rated at 1,375MHz and again with a latency of 7 clocks, which we tested next. This unusual clock rate isn’t supported by many motherboards, so we tested it at 1,333MHz, but in the awareness that it could be run around 3% faster. Our third module was a Corsair XMS3 DHX DIMM at 1,333MHz with a CAS latency of 9 clocks.

After all that, our results indicate that DDR3 is currently no faster than high-performance DDR2 RAM, even where the DDR3 runs at a higher clock speed (and the difference in speed between the DDR3 DIMMs was marginal – there was no clear winner). In the multitasking test, DDR3 was actually slower; we suspect this is down to CAS latency, as in every other way DDR3 ought to be at least the equal of DDR2, and the multitasking test is the benchmark that’s most reliant on random memory access.

This isn’t as poor a result for DDR3 as it may appear. These early modules offer little advantage over mature DDR2 technology, but as latencies fall and clock speeds rise, DDR3 has the potential to keep up with ever-faster CPUs and chipsets long after DDR2 has reached its limits. For now, however, it’s impossible for most users to justify splashing out on DDR3 – unless a must-have motherboard demands it.