Which of these old CPUs is faster?

[Chicken76]

CPU1: Intel Pentium D 820
CPU2: Intel Celeron D 347

What I'm interested in is single-threaded performance.
Celeron has a slightly faster core clock and is a refresh of the Pentium D (65nm vs. 90nm), but has less cache.

 

[mikeymikec]

To get a definitive answer, you'll have to track down reviews/benchmarks. Ideally you'll find both in the same review, but you'll probably have to settle for the same reviewer separating two similar processors with the same tests.

For what it's worth, IMO the Pentium will take the Celeron to the cleaners in most benchmarks. Having said, neither are from a good era in Intel's history. I remember reading a review for an Intel Celeron 2GHz and finding that a Pentium 3 would beat it in quite a few scenarios.

 

[Blastman]

They're both are essentially "P4" generation CPU's.

The 820 is a 2.8Ghz dual core on 90nm manufacture.
The Celeron 347 is a single core 3.06Ghz on 65nm manufacture.

Clock for clock they're quite close in speed. Larger L2 cache on 820 should provide a small advantage if we were comparing single cores. The fact that the 820 is dual core would make it close to 2x the speed of the single Celeron 347.

 

[NTMBK]

The Celeron also has a slower FSB than the Pentium D- on top of the much smaller cache, it could run into memory bandwidth problems.

 

[Maximilian]

CPU1: Intel Pentium D 820
CPU2: Intel Celeron D 347

What I'm interested in is single-threaded performance.
Celeron has a slightly faster core clock and is a refresh of the Pentium D (65nm vs. 90nm), but has less cache.

Purely single threaded the celeron is faster.

Real world single threaded the pentium D is likely faster. Dont forget all the additional crap the OS is doing in the background can be on one core, leaving you a single core entirely free of interference which is worth far more than the paltry extra mhz you get with the celeron.

Really real world, they're both garbage, get something better :thumbsup:

 

[Insert_Nickname]

CPU2: Intel Celeron D 347

Good god, not one of those. They're completely horrible due to their castrated L2 cache. Basically the 347 is incapable of running modern software. You'll get much more mileage out of the D 820. If you can't find anything Core based that is. Anything Core based is going to wipe the floor with Netburst architecture derivatives.

 

[Jimzz]

I vote 820. Faster FSB and more cache.

 

[PG]

Pentium D is dual core and Celeron is single so the Pentium will be better at everything. No contest here.

http://www.cpu-world.com/CPUs/Penti...22MN (BX80551PG2800FN - BX80551PG2800FT).html

http://www.cpu-world.com/CPUs/Celeron_D/Intel-Celeron D 347 - HH80552RE083512 (BX80552347).html

 

[VirtualLarry]

I vote 820. Faster FSB and more cache.

This. P4 Celerons were castrated, no matter which way you look at it.

If you were going to be overclocking, then things might get more interesting. Being 65nm, the Cedar Mill Celeron might OC higher.

 

[Chicken76]

Thanks guys. I appreciate your posts.

Let me throw into the equation the venerable Q6600. At only 2.4 GHz, is it going to be faster (in single-threaded workloads) than any of the other two?

And since we're talking about Core2 era, were there any notable IPC improvements between Conroe (65nm) and Wolfdale (45nm)?

 

[sm625]

It will be faster in every way. You could clock the Q6600 down to 1800MHz and it would still be faster in every way.

 

[Denithor]

Not sure what kind of computational tasks you're looking at, but check out the first chart on this page: http://www.anandtech.com/show/2045/12.

The e6600 (dual core version of the q6600) completely stomps all over the P-D 820. As in, 50% faster video encoding at 2.4GHz versus the faster clocked 820 @ 2.8GHz. And these are both dual core chips, so the results can be directly compared.

Which is why Maximilian above said to get something better.

 

[Magic Carpet]

And since we're talking about Core2 era, were there any notable IPC improvements between Conroe (65nm) and Wolfdale (45nm)?

Conroe > Wolfdale is what Sandy > Ivy Bridge

Personally, I haven't seen huge IPC improvements since Conroe. The biggest advantage of newer processors is pretty much higher clock speed and Turbo Boost (makes a difference in single threaded apps).

 

[SPBHM]

Conroe to Wolfdale gained mostly from l2 cache 4MB to 6MB.
lower power usage, oh, and they added SSE4.1 (but I don't think it's much of a factor)

 

[Magic Carpet]

Conroe to Wolfdale gained mostly from l2 cache 4MB to 6MB.
lower power usage, oh, and they added SSE4.1 (but I don't think it's much of a factor)

That is correct, but the raw computational speed hasn't increased that much. Let's see if AVX2 can turn things around and make people upgrade. All we need is a few killer apps, making exclusive use of those newer features.

 

[Concillian]

Consider this for a moment:
http://www.anandtech.com/bench/Product/93?vs=406

That EE is significantly faster than the Pentium D you have.

So if you get the Ivy equivalent of the G620 (G2020) for $65 CPU and a cheap <$50 H61 motherboard and $20 memory you can have SIGNIFICANTLY faster CPU that also is is also very very low power consumption.

And it's also faster single threaded than the Q6600:
http://www.anandtech.com/bench/Product/53?vs=406

 

[Concillian]

Personally, I haven't seen huge IPC improvements since Conroe. The biggest advantage of newer processors is pretty much higher clock speed and Turbo Boost (makes a difference in single threaded apps).

What the hell are you smoking?
http://www.anandtech.com/bench/Product/54?vs=677
3.33 vs. 3.3 neither have turbo,so very fair comparison.

40-70% increase in IPC from wolfdale to IB at the same clocks. Back to Conroe is even more. Each step has been small, but the cumulative effect oveer the years of small improvements ends up being pretty large. up to a cumulative 100% improvement if you look from Conroe to now.

 

[VirtualLarry]

40-70% increase in IPC from wolfdale to IB at the same clocks. Back to Conroe is even more. Each step has been small, but the cumulative effect oveer the years of small improvements ends up being pretty large. up to a cumulative 100% improvement if you look from Conroe to now.

This. It's actually VERY impressive, how much CPU power is on tap, with even the lowliest IB dual-core, the G1610. Clock-speed wise, it's not really much better than the E5200, but when you consider the IPC increases, even with overclocking of the E5200 thrown in, the IB is probably better.

 

[Concillian]

This. It's actually VERY impressive, how much CPU power is on tap, with even the lowliest IB dual-core, the G1610. Clock-speed wise, it's not really much better than the E5200, but when you consider the IPC increases, even with overclocking of the E5200 thrown in, the IB is probably better.

The E5xxx were also crippled compared to the E8xxx too, so it's IPC was lower. The overclocking allowed them to get "good enough" for a cheap CPU. The current gen low end CPUs are much less crippled in terms of IPC. Intel is crippling them by disabling HT and preventing overclocking instead of crippling single core performance.

So it wouldn't surprise me at all if Ivy low end single core performance was double that of the E5xxx Wolfdales at the same clock.

Todays low end chips really do provide pretty awesome performance for the price. That's why they'll never come out with an i3 K-series chip.

 

[Magic Carpet]

CPU1: Intel Pentium D 820
CPU2: Intel Celeron D 347

What I'm interested in is single-threaded performance.
Celeron has a slightly faster core clock and is a refresh of the Pentium D (65nm vs. 90nm), but has less cache.

What the hell are you smoking?
http://www.anandtech.com/bench/Product/54?vs=677
3.33 vs. 3.3 neither have turbo,so very fair comparison.

40-70% increase in IPC from wolfdale to IB at the same clocks. Back to Conroe is even more. Each step has been small, but the cumulative effect oveer the years of small improvements ends up being pretty large. up to a cumulative 100% improvement if you look from Conroe to now.

A 40% IPC improvement, isn't 400%; and even then, there is no point to have superior hardware, if software can't take advantage of that. The awful fact is, Conroe-class or above CPU is good enough for general computing things for the vast majority of people. And with a simple GPU upgrade you can have your CPU idling around most of the time (Windows 8). So yeah, even a late Pentium 4/D should cut it, provided, you can find a cheap video card upgrade with WDDM 1.2 driver model.

And people wonder, why conventional PC sales have been stagnant?

LMAO, people would rather spend money on a new iPhone or Kindle or something else

Only a die-hard enthusiast/fanatic would keep on upgrading his/her PC gear in these market conditions. Most people aren't like that. Most people follow the market trends. And the devs behind the apps, follow their market. And guess what? The market is always right.

So, unless you're a developer or a hard-core gaming enthusiast, spend your money on something else. You are not missing out anything, that is "life essential".

we need a few killer apps, making exclusive use of those newer features.

Until then, there is no point to buy for the future, because you never know, what the future might hold for you. Buy it when you need it. Value your time and your dollar.

 

[Concillian]

A 40% IPC improvement, isn't 400%;

When is a 5 year period that we've seen 400% IPC improvement. Which 5 years ever? If you include the end of netburst to Ivy Bridge, you still aren't at 400% IPC and that's way more than 5 years.

How can you expect a 400% IPC improvement from wolfdale to Ivy?
How can you say "The biggest advantage of newer processors is pretty much higher clock speed and Turbo Boost" when a dual core i3 with hyperthreading is almost equivalent to an overclocked Q6600 running the same speed?
We've gotten to the point where 2 modern cores are about as good as 4 old conroe cores... that's a bigger advantage than current cores being <4GHz vs. the 2.4 GHz of a stock Q6600? No, the IPC improvement is MORE advantage than the clock speed. No matter how you try to spin it, your statement that clocks and turbos are a bigger gain than IPCs since Conroe is not true. IB i3 is approx equivalent to a Conroe quad (Kentsfield, Q6600) clock for clock... roughly 100% improvement, but Clocks have gone from 2.4 up to turbo of 3.9... only ~65% increase.

You have your head in the clouds. On that, I'm sorry. Your initial statement is flat out incorrect. IPC improvements have been huge and account for a larger part of Haswell and Ivy CPU improvement than clocks when you compare back to Conroe.

---

On your other rantings... Software exists to push the CPUs. People don't want it. Performance isn't going to make people want desktop CPUs. Software isn't going to make people want desktop CPUs. The only thing that's constant is change, and the desktop performance people like us are just going to have to adapt to the changes that are obviously upon us. Nothing is going to change that desktop CPUs are a shrinking market.