This started as a simple question in the forum, what CPU is better? The 3200+ Clawhammer or the 3200+ Newcastle? Since I have one of each flavor on hand I figured I'd do a little testing to answer that question.

The Newcastle I have is a true 3200+ but the Clawhammer I have is a preproduction model that was called the 3100+. The 3100+ never made it to retail, but it has the same core as the 3200+ Clawhammer and has the 1024KB of cache on board. The only difference is that it is clocked at 1800MHz instead of 2000MHz. I would have liked to up the timing to run it at 2000MHz, but unfortunately that CPU is locked up tighter than Tekzoid's liquor cabinet so for testing I did the next best thing, I clocked down my Newcastle from 2200MHz to 2000MHz. That keeps the playing field exactly the same.

The gear I used for testing is of course the two CPU's, an Asus K8VSE Deluxe mobo, a GB of Ultra Products PC4000 RAM, and an old Ti-4200 Gainward Golden Sample vid card. I already had my mind made up on which CPU is the better buy before testing, but ya'll make up your own mind based upon my tests. Here they are:

First off, a little Sandra action.

Sandra CPU

ALU

Clawhammer: 7785
Newcastle: 8653
Winner: Newcastle by 10%

FPU

Clawhammer: 2843
Newcastle: 3160
Winner: Newcastle by 10%

Sandra MM

Integer

Clawhammer: 17156
Newcastle: 19067
Winner: Newcastle by 10%

Floating Point

Clawhammer: 18444
Newcastle: 20502
Winner: Newcastle by 10%

Now for a little Crystalmark testing.

ALU

Clawhammer: 6830
Newcastle: 7565
Winner: Newcastle by 10%

FPU

Clawhammer: 8338
Newcastle: 9267
Winner: Newcastle by 10%

PCMark04 is always a good synthetic test for hardware. Here, I did the CPU test only.

PCMark04 CPU

Clawhammer: 3362
Newcastle: 3654
Winner: Newcastle by 8%

Prime95 is a little more real world when testing CPU speed.

Prime95 2048k FFT ms

Clawhammer: 159.995
Newcastle: 146.858
Winner: Newcastle by 8%

On to some gaming benchmarks where the cache on the Clawhammer CPU should make a difference.

AquaMark3

Clawhammer: 15993
Newcastle: 16013
Winner: Newcastle by less than 1%

Halo

Clawhammer: 34.73
Newcastle: 35.19
Winner: Newcastle by 1%

CodeCreatures

Clawhammer: 27.3
Newcastle: 27.8
Winner: Newcastle by 2%

Unreal Tournament 2004

BR Colossus

Clawhammer: 89.96
Newcastle: 89.04
Winner: Clawhammer by 1%

DM Rankin

Clawhammer: 75.56
Newcastle: 75.74
Winner: Newcastle by less than 1%

As you see, the Newcastle CPU won every benchmark except one. Granted, it was close for all of them, but it still won. In order to give you some more comparisons, I clocked the Newcastle down to 1800MHz so it would match the Clawhammer speed and benched it again. You can compare the results to those above:

Sandra CPU ALU: 7786
Sandra CPU FPU: 2843
Sandra MM Integer: 17157
Sandra MM Floating Point: 18448
Crystalmark ALU: 6832
Crystalmark FPU: 8356
PCMark04: 3328
Prime95: 160.832
AquaMark3: 15988
Halo: 34.88
Codecreatures: 27.2
UT2K4 BR Colossus: 87.21
UT2K4 DM Rankin: 74.24

As you see, when I match the clock speeds the CPUs are almost the same in performance. What most people don't realize is that the core isn't the same on each, the Newcastle has some internal enhancements that the Clawhammer doesn't. That's why even at the same clock speed and giving a 512MB cache advantage to the Clawhammer, the Newcastle can still hang with it.

Since both processors are the same price, in my opinion there is no reason not to go with the Newcastle CPU. Some folks say that the Clawhammer overclocks better, but I beg to differ. The X10 timing of the Clawhammer is not nearly as FSB friendly as the X11 timing of the Newcastle. And if you still don't think it overclocks, check out the attached thumbnail. That is on an open air bench system running average air cooling on a board that is known not to OC. If I had higher timings available I would have easily broke 2600.

So, ya'll got my tests and opinion. Which is the better one to buy? That decision is still up to you.

Attached Thumbnails

 

BigA, nice comparison but I've got one nagging comment that seems to be the root of the pro-Clawhammer debate;

What happens when both are OC'ed to the max? your saying 2.53 is the Newcastles top speed but what about the Clawhammer? The game benchmarks are the ones I'm mainly concerned with as for many of us they represent the best 'real world' tests. Now seeing that a 200mhz differential in favor of the Newcastle has them at under a 3% difference in frames, what about when both are OC'ed; theoretically the Newcastle needs to be able to STAY 200mhz in front of the CLawhammer to keep the lead but I've gotten the general gist that when paired with a AGP/PCI locked mobo (not the Asus, the AN51R maybe) the Clawhammer will hit around 2.4ghz. Again, in theory, thats just over a 100mhz difference between the two, and in theory the Clawhammer should be faster than the Newcastle by a few frames.

My point being; You've given us stock results which are almost the same meaning that 512kb L2 = 200mhz more or less. Now what about when overl****ed? In theory the Clawhammer will win if it passes 2.33ghz.

Also with the overclocks are you using a smaller RAM divider and a smaller LTD so the HTT bus and the RAM speeds don't cap your overclock? In theory with the way the Hypertransport bus works the FSB doesn't really mean anything as long as the RAM divider keeps the RAM in its know range (at or under its top speed) and the Hypertransport bus stays at or under its top speed.

It's a nice comparison BA but it definatly seems that you have your mind made up and are not giving the Clawhammer a snowballs chance in hell; also I don't know if it being a pre-production sample has anything to do with it but I'd wonder if it had any adverse effects on its performance/

The 11 multi of the Newcastle over the 10 on the Clawhammer gives it more overclocking potential. Even with 11 multi, if you have a good mobo/ram /cooling both chips hit 300 FSB no problem. Cores don't change their potential, the multi does.

I really need to start reading up on the whole 64 bit thing going on, so this was pretty irrelevant to me, though, a very good article nonetheless.

A point other than the valid ones Ark raised, is there any way that the vid card is capping the procs from their true performance? I'd be very interested to see how the numbers turn out following Arkangyl's suggestions and a better vid card.

EDIT to below (prior to the gracious edit that pointed out my folly and crappiness of reading skills): Dang it, I did read that too. Just misunderterpreted it.

Guess I better get some better ram - because I can shut BA's overclock down, but my memory divider has to be set so low it's crazy.

My response to Arkangyl is: Yes, my mind was made up before the review. As a matter of fact, my mind was made up when the Newcastle was first released. The faster CPU is the better buy. If you have to overclock the Clawhammer in order to equal the performance of the Newcastle, why buy it? The Newcastle is quicker out of the box. Not everybody has the gear or knowledge to overclock a CPU. That's why at the end I said each individual needs to make up their mind for themselves.

The cache doesn't enhance the performance of the Clawhammer much at all. Just look at the numbers when I had both CPUs clocked at the same speed. If you are going to base value on overclocking, then my recommendation is to just stick with Socket A and buy a 1700+ XP chip.

As far as your suggestions go on the divider, look at the board I used then ask again. About the only overclocking option for that board is power on or power off. No, the preproduction CPU isn't any different from the production CPU except for 200MHz. The only reason that it wasn't sent to retail is that the marketing folks at AMD rightly surmised that nobody would buy the 1800MHz 3000+ CPU if they could get the same speed with larger cache. They were going to be closely priced.

I appreciate your comments and if you feel that I may have skewed the results, I look forward to your own member review.

I'll have to switch mobo's to do that. The Asus board has no overclocking options except FSB. I plan on switching them out later on this week. You guys need to understand that I do this between projects on my wife's honey-do list.

Quote: Originally Posted by BigAkita I'll have to switch mobo's to do that. The Asus board has no overclocking options except FSB. I plan on switching them out later on this week. You guys need to understand that I do this between projects on my wife's honey-do list.

Did you ever try to write a list for your wife?

Quote: Originally Posted by Crazy_J Did you ever try to write a list for your wife?

Yeah, she laughed and kicked my ass. It was OK though cause we were naked.