Tom's CPU Architecture Shootout: 16 CPUs, One Core Each, And 3 GHz
Table of contents
- 1. A Real (Theoretical) Performance Shootout
- 2. Six-Core CPUs: AMD Thuban And Intel Gulftown
- 3. Modern Quad-Core CPUs: AMD Deneb And Intel Sandy Bridge
- 4. Modern Dual-Core CPUs: AMD Regor And Intel Clarkdale
- 5. Older Dual-Core Designs: AMD Brisbane, Intel Conroe, And Intel Wolfdale
- 6. Outdated Dual-Core Designs: AMD Windsor And Intel Prescott
- 7. Platforms: LGA 1366, 1156, 1155, 775, Socket AM2+, And AM3
The CPU landscape is really complex. Both AMD and Intel offer tons of different models. But how would today’s processors perform if they didn't have multiple cores? We take 16 different CPUs and compare them all using a single core running at 3 GHz.
Ever since AMD and Intel started cramming more processing cores into their CPUs, potential performance has grown faster than it did back when single-core CPUs were king thanks to parallelization. Back then, pushing higher frequencies and improving performance per clock were the only ways to speed things up. Now, a developer has to optimize his application to take advantage of multiple cores. But doing so enables scaling that simply wasn't possible before.
We all know that more advanced manufacturing technologies are paving the way for more cores per CPU, and that clock rates are slowly creeping up as well. But how have AMD and Intel improved the performance each core is capable of delivering at a given frequency? Are today’s CPUs any faster than a five-year-old Core 2 if you compare a single core at the same speed? We grabbed 16 different processors from both companies and ran our latest benchmark suite with each operating at 3 GHz. This story examines how they all do in what most folks would consider a very experimental shootout between artificially-created single-core CPUs introduced in the past five years.
Prerequisites and Processors
In preparation for this article, we had to look at the processors available to us for benchmarking. Of course, we wanted to include AMD’s and Intel’s latest products sporting four and six cores. We also thought it'd be important to include a larger number of dual-core products. After all, much has changed since the Athlon 64 X2 and Pentium 4 ruled the jungle. Our selection includes first- and second-generation Core processors with two, four, and six cores, plus Phenom II, Athlon II, and Athlon 64 X2 models.
It's really a challenge to find motherboards for a job like this if you really want to limit core usage in the BIOS. We had to try a lot of potential subjects before finding one for each CPU interface that'd let us modify the number of active cores. And since we couldn't be sure that disabling processing cores in the BIOS would physically turn them off, we refrained from measuring power consumption.
The 3 GHz Battle
This comparison wouldn't have made much sense if we ran each processor at its default clock rate. In addition to limiting the number of active cores to one, we also locked down the frequency of each chip to 3 GHz. We also switched off all power-saving mechanisms like Cool'n'Quiet and SpeedStep, as well as performance-enhancers like Turbo Core and Turbo Boost. As a result, we can be sure that the conditions for testing each CPU are identical. There are only two exceptions. Our Clarkdale- and Lynnfield-based chips run at 2.93 GHz. We could have hit 3 GHz by tweaking base clocks and multipliers, but that would have altered the results. Losing 66 MHz is measurable, sure, but it shouldn't alter the bigger picture.
Latest CPU News
- 24/05 – Nvidia to Push Tablet Prices to $199 With Kai Platform
- 23/05 – AMD Backing Out of CPU Speed Wars Against Intel
- 22/05 – AMD Hitting The Casino Floor With Embedded R-Series
- 22/05 – Does Your AMD FX Platform BSOD with Steam? Read This.
- 20/05 – Intel To Support Cheap Ultrabooks With Low-Cost Celerons
Latest CPU reviews
- 24/05 – Overclocking Core i7-3770K: Learning To Live With Compromise
- 22/05 – Core i5-3570K, -3550, -3550S, And -3570T: Ivy Bridge Efficiency
- 15/05 – AMD A10-4600M Review: Mobile Trinity Gets Tested
- 14/05 – Core i7-3720QM: Ivy Bridge Makes Its Mark On Mobility
- 23/04 – Intel Core i7-3770K Review: A Small Step Up From Sandy Bridge
The page selector doesn't work... can you please use standard html dropdown boxes?
He missed 2006 with putting the real single-core
On the 4th page, you call the CPU an i5-530 in bold. I thought that was wrong and read down to see it's an i3-530. Apparently most people are asleep right now, so it's an excellent time to fix typos.
The page selector doesn't work... can you please use standard html dropdown boxes?
Their page selector is finnicky, but as workaround, click the words (and scroll bar that appears), not the drop down arrow, and it will work (just barely though).
Great article, thanks!
More proof that an i5-2500k rapes everything AMD has. This isn't like "Well, AMD is $40 cheaper and is pretty close, and even beat Intel in some areas" - It's more like "Intel absolutely rapes everything AMD has to offer...and It's slightly more money".
Certain features like Quick sync can even make the price difference non-existent to some people, as the time they save is invaluable and well worth an extra X amount.
Any who, I'm NOT buying AMD until they lower their price, or maintain their price and increase performance. 6 cores, 8 cores, 5,000 cores....it doesn't matter to me - mostly because the software doesn't utilize that many cores and clock per clock, they can't get in the game with Intel.
Correct me if I'm wrong... but looks like the CPU-Z snapshots for AMD Windsor and Brisbane are interchanged...
I think the P4 results are pretty awesome. Damn, I'm so glad I never bought one of those. Given recent events, I sometimes forget how awesome the Athlon 64 was back in the day.
Great article. There are a few caveats of your methods that should be mentioned as well as another pro-AMD qualifier in your conclusion.
1) Disabling cores gives a Cache size advantage to whichever architecture initially had a larger cache. The point being the disabled i7-980X has extra cache to work with (but same 2GB/per core) as the i7-970. There may be some other examples as well. Overall--great implentation of evening the playing field.
2) AMD CPUs often come stock at higher clock speeds when compared to Core 2 architecture. This should be stated because a Core 2 Quad loses to a Phenom II x4 specifically because the Core 2 Quad often only OC's to 3.5GHz easily and the Phenom II x4's hit 4.0GHz (and more easily since most are Black Edition). I'd have been satisfied with a simple reminder like "one must consider clock speeds when comparing comparable architectures such as Core 2 Duo v. K10". (Phenoms also have that L3 cache).
But your bottom line that AMD really needs to bring something to the table with Zambezi remains true. There isn't a price point over $100 that AMD competes well at.
Ummm, also might it be pointed out that none of the Llano chips were used. This seems like a serious omission as these should be the most representative of what AMD currently has to offer. And the fact that you must have them lying about from your previous review.
Seriously undermines the value of the article as most of us would love to do a side by side comparison of the BD vs SB architectures.
Ummm, also might it be pointed out that none of the Llano chips were used. This seems like a serious omission as these should be the most representative of what AMD currently has to offer. And the fact that you must have them lying about from your previous review. Seriously undermines the value of the article as most of us would love to do a side by side comparison of the BD vs SB architectures.
How does it undermine anything? A Llano chip is just a Phenom II w/o the L3 cache, but a doubled L2 cache that mostly makes up for it. Performance wise, the only real difference is clock speed (and integrated graphics, but graphics aren't tested here).
Llano doesn't do anything impressive or special as a CPU--it's the fact that it's integrated graphics do everything an average user wants very well in a single cost efficient package that's special.
@dalauder
It is just that it appears that the reviewer must have conciously chosen NOT to use AMD's most recent hardware while choosing to use INTEL's latest and greatest.
It just indicates a possible biase. I am aware that the llano chips aren't greatly different in architecture, they have however received a process shrink. I am just annoyed that they chose not to include it as otherwise it is a very interesting well thought out article.
It is a glaring oversight. 'Nuff said.
@ dalauder
I would have to agree with 'americanbrian', as this piece as an Intel vs Amd comparison should really include the latest chips from both companies, as well as older architectures. The smaller SOC's are keeping the market competitive.
Although I would have to agree that currently AMD per clock is simply not competing and regardless of 'Zambezi' processing power, an Intel die shrink will inevitably give them the lead again in the chip race.
I wonder what the smallest process shrink will be? Graphene cpu's would be amazing.
I really want to see a GPU oriented test like this one. And to compare a little the total performance increase trough the years.
I think the P4 results are pretty awesome. Damn, I'm so glad I never bought one of those. Given recent events, I sometimes forget how awesome the Athlon 64 was back in the day.
I like to see that the difference in performance between an Athlon X2 6000+ does not change vs the P4
check this out : http://www.tomshardware.com/charts [...] -2,39.html
and see the same performace difference for the X2 6000+ with all core enabled
I don't think that disabling cores in the bios makes any big difference