Single core vs Dual Core
Single core vs Dual Core
(OP)
I was just curious, having just received my new Hypersonic Aviator FX7 (laptop) with the x2 4800+, so I'm comparing it against my Boxx with FX57.
vitals:
Hypersonic Aviator FX7(Clevo laptop chassis)
AMD 64 X2 4800+ (dual core)
via mobo / chipset
nVidia Quadro FXGo 1400 PCIe
95 Gb 7200 RPM USATA
2 Gb RAM
Boxx 3200
AMD 64 FX57 (single core)
nForce4 SLi chipset
nVidia Quadro FX 3450 PCIe, SLi capable
74 Gb 10000 RPM SATA
3 Gb RAM
The RAM difference shouldn't be a factor, since none of the models made the computers page out. The hdd speeds shouldn't be a factor either since I wasn't comparing open or save speeds, and there was no paging.
The shipinabottle time for the Boxx was 18.7, Hypersonic 27.7 Single runs were Boxx .30 and Hypersonic .59.
I thought this might have been due to the difference in graphics cards, so I enabled Software OGL and tried it again, this time Boxx = 19.375, Hypersonic = 30.1. That settled that. It's mostly processor.
**advantage Boxx
So I opened a part with some swoops and ctrl Qed. 224 features, many surface and fillet features. It was interesting to watch. The X2 processors sometimes pegged 100%, sometimes just 50%, and if you watched the percentage indicator on the SW lower right status bar, the Boxx seemed to be going faster, but the Hypersonic definitely finished first.
rebuild time: Hypersonic 670 seconds, Boxx 2171. Something strange up with that. This was an old model, so updating was a factor, as well as a couple of features erroring out. The big difference was a single surface knit feature, which didn't error out.
**advantage Hypersonic
Rendering, basically the one on the front page of my website, rendered to screen. The Hypersonic screen is 1900x1200, the Boxx is 1600x1200. Hypersonic was done in 7 minutes, Boxx took 8.
**advantage Hypersonic
A part with several inserted parts, combines, delete faces, etc. 24 features. Hypersonic = 123 s, Boxx = 124 s. What is interesting is that the feature order in the Feature Statistics was different. The combine features took longer on the Boxx, and the Delete Face took longer on the Hypersonic.
**tie
Another part with 312 features. Lots of draft and fillets. Hypersonic = 165, Boxx = 164. Again, though, the Boxx takes longer on Combine features (115s vs 125s), and made up for it on draft and fillet features.
**tie
Same part with verification turned on. Hypersonic = 453, Boxx = 424. This time Hypersonic was much slower on combine (323s vs 274s). Hyper sonic much faster on Move Face features, where I noticed it was pegging both processors (66s vs 100s).
**advantage Boxx
I don't have any big assemblies to compare. My guess is the dual core would smoke the single with opening and rebuilding assemblies.
Verdict? The first verdict is that the ship in a bottle benchmark is not very representative for the kind of modeling I do. In real world modeling, the two machines were about even. That one anomalous rebuild time with the knit surface worries me, but it really happened, so I don't know what to say about it. Other than in the shipinbottle, the dual core never lagged behind. If you put too much stock in the shipinabottle, you'd go around thinking one machine was twice as fast as the other, but that obviously isn't true.
There was a substantial price difference between the machines, the Hypersonic laptop was ~$3300 and the Boxx tower was ~$4000. The FX57 was certainly a premium price item. I should also mention that there was about an 8 month interval between the two purchases, which affects the price on fairly recent cpus.
The FX57 is the fastest single core processor available. It doesn't really compare with the FX60 which is dual core and clocked somewhat slower. The FX60 seemed to be about 10% faster than the 4800+ on Toms Hardware cpu benchmarks, but it was probably twice the price.
For the price and the performance, I would recommend the X2 4800.
Other nice things about this is that now you can get real workstation processors in a portable format. Remember that this is a comparison between a nicely equipped (not to say "top of the line") tower, and a nicely equipped "laptop".
I at one point used an Intel version of the Clevo chassis built by Boxx, looks from the outside exactly the same as my Hypersonic. It blew very hot air on your left hand, and the fans ran constantly and were quite loud. The difference with the more power efficient AMD chips is like night and day. Very noticeable. This is the difference between a laptop you can use and a laptop you can use to fry eggs, but can't sit near.
Anyway, lots of typing. Hope this sheds light for anyone looking at new hardware.
Matt
vitals:
Hypersonic Aviator FX7(Clevo laptop chassis)
AMD 64 X2 4800+ (dual core)
via mobo / chipset
nVidia Quadro FXGo 1400 PCIe
95 Gb 7200 RPM USATA
2 Gb RAM
Boxx 3200
AMD 64 FX57 (single core)
nForce4 SLi chipset
nVidia Quadro FX 3450 PCIe, SLi capable
74 Gb 10000 RPM SATA
3 Gb RAM
The RAM difference shouldn't be a factor, since none of the models made the computers page out. The hdd speeds shouldn't be a factor either since I wasn't comparing open or save speeds, and there was no paging.
The shipinabottle time for the Boxx was 18.7, Hypersonic 27.7 Single runs were Boxx .30 and Hypersonic .59.
I thought this might have been due to the difference in graphics cards, so I enabled Software OGL and tried it again, this time Boxx = 19.375, Hypersonic = 30.1. That settled that. It's mostly processor.
**advantage Boxx
So I opened a part with some swoops and ctrl Qed. 224 features, many surface and fillet features. It was interesting to watch. The X2 processors sometimes pegged 100%, sometimes just 50%, and if you watched the percentage indicator on the SW lower right status bar, the Boxx seemed to be going faster, but the Hypersonic definitely finished first.
rebuild time: Hypersonic 670 seconds, Boxx 2171. Something strange up with that. This was an old model, so updating was a factor, as well as a couple of features erroring out. The big difference was a single surface knit feature, which didn't error out.
**advantage Hypersonic
Rendering, basically the one on the front page of my website, rendered to screen. The Hypersonic screen is 1900x1200, the Boxx is 1600x1200. Hypersonic was done in 7 minutes, Boxx took 8.
**advantage Hypersonic
A part with several inserted parts, combines, delete faces, etc. 24 features. Hypersonic = 123 s, Boxx = 124 s. What is interesting is that the feature order in the Feature Statistics was different. The combine features took longer on the Boxx, and the Delete Face took longer on the Hypersonic.
**tie
Another part with 312 features. Lots of draft and fillets. Hypersonic = 165, Boxx = 164. Again, though, the Boxx takes longer on Combine features (115s vs 125s), and made up for it on draft and fillet features.
**tie
Same part with verification turned on. Hypersonic = 453, Boxx = 424. This time Hypersonic was much slower on combine (323s vs 274s). Hyper sonic much faster on Move Face features, where I noticed it was pegging both processors (66s vs 100s).
**advantage Boxx
I don't have any big assemblies to compare. My guess is the dual core would smoke the single with opening and rebuilding assemblies.
Verdict? The first verdict is that the ship in a bottle benchmark is not very representative for the kind of modeling I do. In real world modeling, the two machines were about even. That one anomalous rebuild time with the knit surface worries me, but it really happened, so I don't know what to say about it. Other than in the shipinbottle, the dual core never lagged behind. If you put too much stock in the shipinabottle, you'd go around thinking one machine was twice as fast as the other, but that obviously isn't true.
There was a substantial price difference between the machines, the Hypersonic laptop was ~$3300 and the Boxx tower was ~$4000. The FX57 was certainly a premium price item. I should also mention that there was about an 8 month interval between the two purchases, which affects the price on fairly recent cpus.
The FX57 is the fastest single core processor available. It doesn't really compare with the FX60 which is dual core and clocked somewhat slower. The FX60 seemed to be about 10% faster than the 4800+ on Toms Hardware cpu benchmarks, but it was probably twice the price.
For the price and the performance, I would recommend the X2 4800.
Other nice things about this is that now you can get real workstation processors in a portable format. Remember that this is a comparison between a nicely equipped (not to say "top of the line") tower, and a nicely equipped "laptop".
I at one point used an Intel version of the Clevo chassis built by Boxx, looks from the outside exactly the same as my Hypersonic. It blew very hot air on your left hand, and the fans ran constantly and were quite loud. The difference with the more power efficient AMD chips is like night and day. Very noticeable. This is the difference between a laptop you can use and a laptop you can use to fry eggs, but can't sit near.
Anyway, lots of typing. Hope this sheds light for anyone looking at new hardware.
Matt






RE: Single core vs Dual Core
How about running the STAR 2.1 benchmark on the two systems and reporting back the results. To me, that seems a bit more realistic since it's simple modeling grinds--although your real-world model tests certainly are realistic to what I do every day. Ship-in-a-bottle seems to vary quite a lot from test to test anyway, plus seems to be influenced by graphics capacity more than raw processor power.
You're right--the double chips are cheaper than the double-core FX-60. My FX-60 was ~$1,000 through Xi. I've definitely never spent so much on a chip before. Strangely enough, I often hit 85% while crunching a model in SW--sometimes 100%. Strange for a single-thread application, but it does happen. I had a dual-chip NT system back in 1997 and the same thing happened. But you've got to love the extra core (or chip) for renderings--and especially for rendering animations. Very nice.
Jeff Mowry
www.industrialdesignhaus.com
Reason trumps all. And awe trumps reason.
RE: Single core vs Dual Core
time: 55.38
level: 5
rebuild: 5.16
Boxx:
time: 41.8
level: 5
rebuild: 3.73