Clock for clock would my old tbred-b get embarassed by Atom?T

[dug777]

I plan to go from my 2.1Ghz tbred-b/1.5Gb DDR(single channel)/6800GS@U to a slightly newer Q6600/4GB DDR2/9600GT(or 4850) box shortly, but I'd love to see how the old goat stacked up against an Atom, and against a Q6600...

EDIT:

Clocked down to 10x 166 I got 1 min 8 secs.

Clocked down to stock, 12.5 x133, I got 1min 11 secs.

 

[PlasmaBomb]

I think your Tbred would do OK against an atom since it makes the celly 220 look fast.

 

[GuitarDaddy]

My guess is the Tbred would be a bit faster than the Atom, but would lose miserably to the Q6600 of course.

The Atom although a leap forward for mobile or PDA type cpus, is butt slow when compared to any fairly recent desktop CPU's. In a recent review I read about Atom it was basicly 1/3 the performance of a bottom of the line C2D E2XX desktop CPU

 

[dug777]

Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

 

[996GT2]

Originally posted by: dug777
Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

Haha sure.

I built Q6600 workstation for my parents last year, using a G0 stepping Q6600 (non overclocked since they want stability for work), 8 GB DDR2-800, GeForce 7600GT, and Gigabyte P35-DS3R motherboard.

Just did a quick SuperPi 1M run and got 21.3 seconds. That should be something to compare to.

 

[CTho9305]

Atom's performance is much worse than 32-bit Athlon performance, clock for clock. Atom's architecture is drastically simplified compared to other modern x86 processors, whereas the original Athlon's architecture is still very similar to the modern Athlon 64 architecture.

Originally posted by: 996GT2

Originally posted by: dug777
Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

Haha sure.

I built Q6600 workstation for my parents last year, using a G0 stepping Q6600 (non overclocked since they want stability for work), 8 GB DDR2-800, GeForce 7600GT, and Gigabyte P35-DS3R motherboard.

Just did a quick SuperPi 1M run and got 21.3 seconds. That should be something to compare to.

I don't think you can meaningfully compare SuperPi results between AMD and Intel processors. For whatever reason, one consistently trashes the other, even when real-world code shows a much smaller performance difference. SuperPi is just a bad benchmark. If you check ripping.org's database, a 4GHz Athlon 64 X2 takes about 20 seconds for 1M, but it'd trash a Q6600 at any application that isn't threaded and even some things that are threaded (i.e. virtually everything).

 

[996GT2]

Originally posted by: CTho9305
Atom's performance is much worse than 32-bit Athlon performance, clock for clock. Atom's architecture is drastically simplified compared to other modern x86 processors, whereas the original Athlon's architecture is still very similar to the modern Athlon 64 architecture.

Originally posted by: 996GT2

Originally posted by: dug777
Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

Haha sure.

I built Q6600 workstation for my parents last year, using a G0 stepping Q6600 (non overclocked since they want stability for work), 8 GB DDR2-800, GeForce 7600GT, and Gigabyte P35-DS3R motherboard.

Just did a quick SuperPi 1M run and got 21.3 seconds. That should be something to compare to.

I don't think you can meaningfully compare SuperPi results between AMD and Intel processors. For whatever reason, one consistently trashes the other, even when real-world code shows a much smaller performance difference. SuperPi is just a bad benchmark. If you check ripping.org's database, a 4GHz Athlon 64 X2 takes about 20 seconds for 1M, but it'd trash a Q6600 at any application that isn't threaded and even some things that are threaded (i.e. virtually everything).

Core 2 CPUs trash Athlon X2s in SuperPi because they have a huge, unified L2 cache that essentially allows SuperPi to run from cache.

Your comparison between stock Q6600 and 4 GHz X2 is very impractical because very few Athlon X2s have ever been clocked that high, and that was with LN2 cooling...hardly practical for daily use.

I just did a quick SuperPi 1M benchmark because it's a very common one to use and only took me about a minute to do on the other computer.

 

[CTho9305]

Originally posted by: 996GT2

Originally posted by: CTho9305
Atom's performance is much worse than 32-bit Athlon performance, clock for clock. Atom's architecture is drastically simplified compared to other modern x86 processors, whereas the original Athlon's architecture is still very similar to the modern Athlon 64 architecture.

Originally posted by: 996GT2

Originally posted by: dug777
Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

Haha sure.

I built Q6600 workstation for my parents last year, using a G0 stepping Q6600 (non overclocked since they want stability for work), 8 GB DDR2-800, GeForce 7600GT, and Gigabyte P35-DS3R motherboard.

Just did a quick SuperPi 1M run and got 21.3 seconds. That should be something to compare to.

I don't think you can meaningfully compare SuperPi results between AMD and Intel processors. For whatever reason, one consistently trashes the other, even when real-world code shows a much smaller performance difference. SuperPi is just a bad benchmark. If you check ripping.org's database, a 4GHz Athlon 64 X2 takes about 20 seconds for 1M, but it'd trash a Q6600 at any application that isn't threaded and even some things that are threaded (i.e. virtually everything).

Core 2 CPUs trash Athlon X2s in SuperPi because they have a huge, unified L2 cache that essentially allows SuperPi to run from cache.

Your comparison between stock Q6600 and 4 GHz X2 is very impractical because no Athlon X2 has ever been overclocked to 4 GHz, even on liquid nitrogen cooling. The highest clock speed for any AMD Athlon IIRC was 4 Ghz from a single core FX-57, and that was with LN2 cooling...hardly practical for daily use.

I just did a quick SuperPi 1M benchmark because it's a very common one to use and only took me about a minute to do on the other computer.

Oh really? What are these if not Athlon 64 X2 processors at >4GHz? I'm not claiming it's practical. I'm just saying that there's something horribly wrong with a single-thread benchmark that's 2.5x faster on a CPU that's 60% of the raw clock speed, and maybe 110-120% of the performance per-clock on real applications.

 

[CTho9305]

edit: double post

 

[dug777]

I'll do a superpi now

 

[996GT2]

Originally posted by: CTho9305

Originally posted by: 996GT2

Originally posted by: CTho9305
Atom's performance is much worse than 32-bit Athlon performance, clock for clock. Atom's architecture is drastically simplified compared to other modern x86 processors, whereas the original Athlon's architecture is still very similar to the modern Athlon 64 architecture.

Originally posted by: 996GT2

Originally posted by: dug777
Cool, no comparable benchies for either Atom or the Q6600 out there you know of?

Any I can run against someone with a Q6600 here?

Haha sure.

I built Q6600 workstation for my parents last year, using a G0 stepping Q6600 (non overclocked since they want stability for work), 8 GB DDR2-800, GeForce 7600GT, and Gigabyte P35-DS3R motherboard.

Just did a quick SuperPi 1M run and got 21.3 seconds. That should be something to compare to.

I don't think you can meaningfully compare SuperPi results between AMD and Intel processors. For whatever reason, one consistently trashes the other, even when real-world code shows a much smaller performance difference. SuperPi is just a bad benchmark. If you check ripping.org's database, a 4GHz Athlon 64 X2 takes about 20 seconds for 1M, but it'd trash a Q6600 at any application that isn't threaded and even some things that are threaded (i.e. virtually everything).

Core 2 CPUs trash Athlon X2s in SuperPi because they have a huge, unified L2 cache that essentially allows SuperPi to run from cache.

Your comparison between stock Q6600 and 4 GHz X2 is very impractical because no Athlon X2 has ever been overclocked to 4 GHz, even on liquid nitrogen cooling. The highest clock speed for any AMD Athlon IIRC was 4 Ghz from a single core FX-57, and that was with LN2 cooling...hardly practical for daily use.

I just did a quick SuperPi 1M benchmark because it's a very common one to use and only took me about a minute to do on the other computer.

Oh really? What are these if not Athlon 64 X2 processors at >4GHz?

I just read your link. I haven't really checked up on AX2 overclocking with the newer Windsor and Brisbane cores, so my info is pretty out of date. However, my original post still stands. How many cherrypicked Athlon X2s have ever been clocked beyond 4 GHz? Not very many, obviously, and the few that get OC'd that high are done on elaborate LN2 cascades or dry ice cooling...so saying that a 4.2 GHz Athlon X2 would rape a stock Q6600 is kinda irrelevant because 1) you're comparing OC'd to stock, 2) so few Athlon X2s are even capable of that clock speed even with huge voltage (1.8 wtf?) and elaborate LN2 cascades, and 3) a G0 Q6600 OC'd to 3.5 ish on air cooling would completely dominate the X2 in any benchmark. And unlike in the case of the X2, many people have actually achieved 3.5-4 GHz out of a Q6600 on air cooling.

Besides, dug just wanted something to quickly benchmark with...I don't think he really cares THAT much about Intel/AMD discrepancies on his T-Bird

 

[996GT2]

Originally posted by: dug777
I'll do a superpi now

haha ok looking forward to the numbers XD

 

[CTho9305]

Originally posted by: 996GT2
Besides, dug just wanted something to quickly benchmark with...I don't think he really cares THAT much about Intel/AMD discrepancies on his T-Bird

If you really want to do something you know will make one look artificially better/worse than the other, go for it. If you want something that's remotely useful information, you're just confusing the issue. If we say Atom is 1/3rd the speed of Core2 (based on various reviews), and Core2 is 2.5X faster than K8 at 60% of the clock speed based on SuperPI (so ~4X faster clock-per-clock?), you're going to end up concluding that Atom is faster than K8, let alone K7. That's a pretty dumb conclusion.

 

[dug777]

Originally posted by: 996GT2

Originally posted by: dug777
I'll do a superpi now

haha ok looking forward to the numbers XD

A powerhouse performance of 57 seconds to 1M

 

[996GT2]

Originally posted by: CTho9305

Originally posted by: 996GT2
Besides, dug just wanted something to quickly benchmark with...I don't think he really cares THAT much about Intel/AMD discrepancies on his T-Bird

If you really want to do something you know will make one look artificially better/worse than the other, go for it. If you want something that's remotely useful information, you're just confusing the issue. If we say Atom is 1/3rd the speed of Core2 (based on various reviews), and Core2 is 2.5X faster than K8 at 60% of the clock speed based on SuperPI (so ~4X faster clock-per-clock?), you're going to end up concluding that Atom is faster than K8, let alone K7. That's a pretty dumb conclusion.

This is definitely not true in the case of SuperPi. Atom does not have nearly the huge L2 cache of Core 2, which is one of the main reasons the Core 2 line have been so dominant in SuperPi. I stated this 2 posts ago btw.

EDIT: Here's an early benchmark of the Atom 230 (1.6 GHz). As you can see, its superpi times are not impressive at all, even compared to AMD CPUs. The 1M time is about 92 seconds, slower than any Athlon 64/X2 at that clock speed (A 2 GHz Sempron needs 42 seconds, so intrapolating we can estimate a 1.6 GHz Sempron at about 50 seconds).

The lack of L2 Cache and low-power architecture of the Atom makes a big difference in performance. It's a great mobile chip, but it's by no means a benchmark killer.

Link

 

[dug777]

I shall clock mine down to stock (1667) and see how it fares

 

[dug777]

Clocked down to 10x 166 I got 1 min 8 secs.

Clocked down to stock, 12.5 x133, I got 1min 11 secs.

 

[GuitarDaddy]

Sandra CPU tests would be a much more accurate comparison than SPI

 

[dug777]

dling now

 

[GuitarDaddy]

Here's some quad scores to compare

Q6600 @ 2.4ghz

Arithmetic CPU
Dhrystone ALU 43,599 mips
Whetstone 30,332 mflops

Multimedia CPU
Integer x8 SSE4 261,720
Float x4 SSE2 141,099


X3350 @ 3.2ghz (should be about equal to a Q6600 @ 3.5ghz, except the sse4 test)

Arithmetic CPU
Dhrystone ALU 54,688 mips
Whetsone 47,211 mflops

Multimedia CPU
Integer x8 SSE4 414,466
Float x4 SSE2 192,562

 

[dug777]

Originally posted by: GuitarDaddy
Here's some quad scores to compare

Q6600 @ 2.4ghz

Arithmetic CPU
Dhrystone ALU 43,599 mips
Whetstone 30,332 mflops

Multimedia CPU
Integer x8 SSE4 261,720
Float x4 SSE2 141,099


X3350 @ 3.2ghz (should be about equal to a Q6600 @ 3.5ghz, except the sse4 test)

Arithmetic CPU
Dhrystone ALU 54,688 mips
Whetsone 47,211 mflops

Multimedia CPU
Integer x8 SSE4 414,466
Float x4 SSE2 192,562

What did you run to get them?

 

[GuitarDaddy]

Sisoft Sandra Lite

Run the Arithmetic CPU and Multimedia CPU benchmarks

 

[dug777]

Arithmetic

Benchmark ResultsDhrystone ALU : 5748MIPS
Whetstone FPU : 3309MFLOPS

Performance vs. Speed
Dhrystone ALU : 2.76MIPS/MHz
Whetstone FPU : 1.59MFLOPS/MHz

Multimedia

Multi-Media Int x4 aEMMX/aSSE : 18853iit/s
Multi-Media Float x4 aSSE : 20766fit/s


Performance vs. Speed
Multi-Media Int x4 aEMMX/aSSE : 9.05iit/s/MHz
Multi-Media Float x4 aSSE : 9.97fit/s/MHz

 

[GuitarDaddy]

Yep, with a truly multi-threaded benchmark the newer quads absolutely destroy your old single core

the x3350 looks to be roughly 10x faster than the old tbred except in sse4 where its closer to 20x