Intel Dual-Core Waxes Intel Quad-Core (and I eat crow)

Pretty much the case - as my comparison, strange as it sounds, does illustrate.

A little improvement here and there adds up - and the entire path between Kentsfield and Haswell is full of little improvements. (Can anyone name so much as ONE major improvement along the way from Kentsfield to Haswell outside of overclockability?)

Besides, there WAS one other reason I made that particular comparison - what is the price gap between used Q6600 and new G3220 today?

Are you really forgetting Nehalem and Sandy bridge? both as itself were major improvements and architectural changes, Nehalem was better in any sense than the best C2Q yorkfield (which are already better than your kentsfield and btw I'm still using a QX9770 ^^ as backup ).. the only fact that Nehalem replaced FSB for QPI, the integrated memory controller, the native 4 cores in a single die and other stuff made it already a considerable upgrade over C2Q.. then came Sandy bridge which then again bring another architectural jump but not only that but also way higher clocks that made a Sandy i7 be in pair in multi threaded applications with gulftown (32 nm based Nehalem 6 cores chips) like the 980X.. then not only that but also ivy bridge and Haswell.. isn't Haswell vs kentsfield(which its still Penryn based) it's your kentsfield vs all of the above.. maybe could be more understandable if you tried to compare a high clocked Q9500 (just for example) with the pentium dual core haswell.. but trying to avoid Nehalem and Sandy bridge or trying to take it as minor upgrades, are a major mistake..
 
My point was an Intel dual-core stomped the same company's seminal original quad - and the dual-core doing the stomping is basically bottom-end.

This is NOT Intel vs. AMD - but Intel vs. Intel.

What it does is not JUST illustrate how much technology has improved, but how little software (even all too much of the software of today) utilizes multicore.

And I was comparing Q6600 to the P4 - not the other way around.

For what it's worth: The vast majority of software tasks are, and always will be, single-threaded.

Multi-threaded software isn't a magic bullet. You can't just apply parallelization to any code and magically get a performance boost. Making code handle a problem in a multi-threaded fashion often just isn't possible or worthwhile. When it is, a multi-threaded version of the code will be one or even two orders of magnitude more complex to write, debug, and maintain.

That's why multi-threaded applications are limited to only the most CPU-intensive and time-intensive programs. Rendering, video encoding, and other time-bottlenecked tasks are worth parallelizing.

Some games can do a bit of parallelization, but it's still not a night-and-day difference.

Realistically, most people would be better off with a very fast dual-core chip for most non-multitasking scenarios. Open up task manager and watch your total CPU load. It's rare that it will exceed 1 / (number of cores) because the vast majority of what you're doing is single-threaded.

Multi-core is great when you need it, but single-threaded still dominates most workloads. Unfortunately, processes have reached the point that CPU manufacturers can't scale clock frequency like they once did, so instead we're scaling core count.
 
They need to continue scaling the core size and IPC. eg the leap between P3, Core 1, 2, and now haswell sized cores and idealize around 2 or 4 cores that HT as needed.


But I do love me a good 8C/8T chip for VM's.
 
Open up task manager and watch your total CPU load. It's rare that it will exceed 1 / (number of cores) because the vast majority of what you're doing is single-threaded.

I always have task manager open and rarely do I ever see only 1 core pegged. It's nearly always balanced over the 4. The exception being some (very old) 3D games that will peg 1 core only.
 
I always have task manager open and rarely do I ever see only 1 core pegged. It's nearly always balanced over the 4. The exception being some (very old) 3D games that will peg 1 core only.

A mix between 4, sure, but all 4 are not pegged most of the time for the vast majority of people. If yours are, you are far from a typical user and even then I'd say you're probably exaggerating significantly.
 
Sorry, that was worded poorly. I meant that I have never seen all 4 of my cores near 100% during normal use aka gaming. Typically 50% max, across 4 cores.
 
Sorry, that was worded poorly. I meant that I have never seen all 4 of my cores near 100% during normal use aka gaming. Typically 50% max, across 4 cores.

you have to be careful with the thread load readings.. if you are using a monitoring app like afterburner or Core temp to check each individual core load.. if you see 1 core at 50% usage that can mean that 1 of the 2 core threads its at 100% usage.. those kind of things are rarely noticeable, but are common in semi old games (and even few new games) that run in one or two threads.. to see a single core at full 100% usage both threads have to be used at 100% and that's not gonna happen in games.. generally by the nature of the hyperthreading sharing resources you can see 2 threads of a core at 80-85% so the core will never show 100% usage..
 
Refresh my memory. What was better?

Yeah no kidding, what CPU did everyone miss? For a while the two best CPU choices were a Q6600 (quad) or an E8400 (dual). Most 6600s were good for 3.3 to 3.6GHz and the 8400s were hitting 4GHz with slightly better IPC. The 6600 was a bit slower to start however it proved to have longer legs compared to the dual core 8400 when games supported more cores.
 
Yeah no kidding, what CPU did everyone miss? For a while the two best CPU choices were a Q6600 (quad) or an E8400 (dual). Most 6600s were good for 3.3 to 3.6GHz and the 8400s were hitting 4GHz with slightly better IPC. The 6600 was a bit slower to start however it proved to have longer legs compared to the dual core 8400 when games supported more cores.

hmm... you are mixing a lot there.... by the time the E8400 was out in the market, the Q9XXX series were also in the market and those 45nm Yorkfield were much better than the Q6XXX series specially with overclocking involved as they ran cooler and achieved higher OC.. and by itself out the box were better due to higher clock and more L2 cache..
 
hmm... you are mixing a lot there.... by the time the E8400 was out in the market, the Q9XXX series were also in the market and those 45nm Yorkfield were much better than the Q6XXX series specially with overclocking involved as they ran cooler and achieved higher OC.. and by itself out the box were better due to higher clock and more L2 cache..

Correct, though the comparison was often times made between the Q6600 and E8400 because they were the same price.

That said, I'd still like to know what was better than the Q6600 before Intel came out with it's successor as was suggested by the post I quoted a few posts up.
 
6600 was cheaper and around the same price point as the 8400. The 45nm quads were another $100 or so I believe.
 
That said, I'd still like to know what was better than the Q6600 before Intel came out with it's successor as was suggested by the post I quoted a few posts up.

nothing... before the Q6600 all was the Conroe Based Core 2 Duo and Core 2 Extreme...
 
nothing... before the Q6600 all was the Conroe Based Core 2 Duo and Core 2 Extreme...

AMD also didn't have anything remotely close either. Phenom's original release was terrible, the phenom II was worth it though, that didn't happen till much later.
 
The fact is, for those of us that live the [H] way, the motherboard was primary. Choosing the right board allowed me to start out with a Q6600, overclock it to 3.2+, then later when the price was right buy a Q9550 that I overclocked as well. This vastly extending the useful live of my PC, all the while upgrading the GPU right up to the recent GTX 680, new power supply and a new hard drive to a SSD. It was a long and fruitful ride that is now over, EOL. I will soon be replacing the system with a X99 platform motherboard, I'll be able to use the GPU and SDD and power supply and case, basically replacing the CPU/MB/RAM. Why is now the time to upgrade? Because the i7-5820 is exactly the CPU I've been waiting for to start the life of a new platform. Eventually, one day, I'll likely upgrade the CPU from a 6 core to 8+ core (depending on what the 'best' cpu ends up supported by my platform), max out the RAM, and continue to upgrade the GPU and other parts as needed. I expect that eventually that platform as well should EOL, but not before it has many many years on it.

The OPs post on the other hands, shows clearly what the rest of the world experiences when they are at the mercy of another person's build choices.
 
on the other hand, I can't believe the legs this G3258 is showing me

I'm eventually going to toss something beefier in my MB but for now it's holding steady until something comes along I can't live without.
 
For most tasks, a fast single core is plenty. It's only a handful of tasks that can use more then one core effectively. That's the nature of how software is designed. And most of the things that scale really well do so almost infinitely, in which case we're offloading to the GPU instead.

Hence why I never think we'll move beyond quad core CPUs, because there simply isn't any need.

I'll be very interested to see how Pentium/i3s do when DX12 hits, since they should benefit greatly from the decreased CPU overhead. I wouldn't be shocked if top of the line i3's start winning in gaming benchmarks over slower i7s once DX12 comes out.
 
For most tasks, a fast single core is plenty. It's only a handful of tasks that can use more then one core effectively. That's the nature of how software is designed. And most of the things that scale really well do so almost infinitely, in which case we're offloading to the GPU instead.

Hence why I never think we'll move beyond quad core CPUs, because there simply isn't any need.

I'll be very interested to see how Pentium/i3s do when DX12 hits, since they should benefit greatly from the decreased CPU overhead. I wouldn't be shocked if top of the line i3's start winning in gaming benchmarks over slower i7s once DX12 comes out.

Decreased overhead does not necessarily mean decreased usage. I'm pretty sure that developers will find good ways to put those extra CPU resources to good use and build more complex games.

As far as single core and most tasks go, there's still a benefit to multi-core. Just take a look at your processes tab. I've got 124 of them running right now, so even if every single one can only use one core, multi core is still beneficial. You never know when one or more of those processes needs some CPU time while you're doing something CPU intensive. It happens all the time, we don't notice it thanks to modern CPUs
 
welcome to a thing called progress. :)

I have a just bought a £100 box with a 15W kabini APU that is probably faster than my £1500 speed machine from 2005 sporting an Athlon X2 3800+.
 
Back
Top