Utterly Dissappointed in CPU Progress lately...

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Basically in June of 2003, I upgraded from AXP1600+ (bought in 2001) to P4 2.6C w/HT. I overclocked my CPU to 3.2ghz. The CPU originally cost me $220, retail box. So I figured now that I am graduating from university soon, I might as well upgrade to something better. So what can I buy today, 2.5 years later for around $200? A64 3500+ Venice...Then I thought hmmm...how much of a performance am I going to get? Also I thought to make the comparison fair, I have to assume I'll overclock that CPU just like I overclocked the P4 2.6 when I got it. A fair target is about 2600-2700mhz. So then I decided to include FX55 2.6ghz Sandiego 1mb cache core (even faster than what 3500+ overclocked would give me due to higher cache). I also threw in AXP3200+ to see if the dominance of A64 over P4 is really overexagerrated or not. Found these benches and here is what I saw:

Test Setup: - Source

Athlon 64 FX-55 San Diego (2600 MHz 64+64/1024 kB) S939 (NF4 U)
Athlon 64 3500+ Venice (2200 MHz 64+64/512 kB) S939 (NF4 U)
Pentium 4 3.20 GHz Northwood (3200 MHz 12KµOps+8/512 kB) S478 (i875)
Athlon XP 3200+ Barton (2200 MHz 64+64/512 kB) S462 (NF2 U)
6800GT graphics card

GAMES
Wolfenstein 3D - Enemy Territory 1280x1024 - High Quality
FX55 - 208.2fps (+24.7%)
3500+ - 183.3fps (+9.8%)
3.2ghz - 167.0fps (100%)
XP3200+ - 140.4fps (-18.9%)

Quake 3 - 1280x1024 - High
FX55 - 280.7fps (+18.1%)
3500+ - 245.1fps (+3.2%)
3.2ghz - 237.6fps (100%)
XP3200+ - 193.3 (-22.9%)

Doom 3 - 1280x1024 - High
FX55 - 83.4fps (+8.5%)
3500+ - 82.0fps (+6.6%)
3.2ghz - 76.9 (100%)
XP3200+ - 72.3 (-6.4%)

Doom 3 (another level) - 1280x1024 - High
FX55 - 108.0fps (+19.2%)
3500+ - 102.6fps (+13.2%)
3.2ghz - 90.6 (100%)
XP3200+ - 81.6 (-11.0%)

UT2k4 - 1280x1024
FX55 - 178.8fps (+8.5%)
3500+ - 152.3fps (+6.6%)
3.2ghz - 90.6 (100%)
XP3200+ - 81.6 (-11.0%)

Far Cry - 1280x1024
FX55 - 207.1fps (+29.7%)
3500+ - 186.2fps (+16.6%)
3.2ghz - 159.7 (100%)
XP3200+ - 134.3 (-18.9%)

VIDEO
Xvid - converting from VOB to XVID (182mb)
FX55 - 317 seconds (+6.0%)
3500+ - 366 seconds (-8.9%)
3.2ghz - 336 seconds (100%)
XP3200+ - 423 seconds (-25.9%)

Microsoft Windows Media Encoder - Encoding AVI to WMV (450mb)
FX55 - 148 seconds (+5.4%)
3500+ - 175 seconds (-12.2%)
3.2ghz - 156 seconds (100%)
XP3200+ - 207 seconds (-32.7%)

Mainconcept Encoder - Encoding DV to MPEG2 (1.2gb)
FX55 - 147 seconds (+16.3%)
3500+ - 171 seconds (0.0%)
3.2ghz - 171 seconds (100%)
XP3200+ - 203 seconds (-18.7%)

DivX - Converting from VOB to MPEG4 (182mb)
FX55 - 387 seconds (+10.9%)
3500+ - 454 seconds (-5.8%)
3.2ghz - 429 seconds (100%)
XP3200+ - 552 seconds (-28.7%)

Pinnacle Studio 9 Plus - Encoding and Transition Rendering from DV to MPEG2
FX55 - 110 seconds (+4.5%)
3500+ - 128 seconds (-11.3%)
3.2ghz - 115 seconds (100%)
XP3200+ - 144 seconds (-25.2%)

AUDIO
Ogg-Vorbis - Audio CD Terminator 2 SE, 74 min, Quality =5
FX55 - 184 seconds (+20.7%)
3500+ - 217 seconds (+2.3%)
3.2ghz - 222 seconds (100%)
XP3200+ - 233 seconds (-5.0%)

Lame - Audio CD Terminator 2 SE, 160kBit/s 74 min
FX55 - 203 seconds (+24.1%)
3500+ - 240 seconds (+5.0%)
3.2ghz - 252 seconds (100%)
XP3200+ - 261 seconds (-3.6%)

APPLICATION
WinRAR - 283mb, 246 files, compression 'best'
FX55 - 90 seconds (+23.3%)
3500+ - 105 seconds (+5.7%)
3.2ghz - 111 seconds (100%)
XP3200+ - 144 seconds (-29.7%)

CloneDVD - Transcoding DVD9 to 4.7 Terminator 2 SE (9gb)
FX55 - 559 seconds (+16.8%)
3500+ - 649 seconds (+0.6%)
3.2ghz - 653 seconds (100%)
XP3200+ - 727 seconds (-11.3%)

Multithread Benchmark (4 Threads) - WinRAR, Ogg-Vorbis, Windows Media Encoder, Lame Encoder
FX55 - 478 seconds (+29.7%)
3500+ - 596 seconds (+4.0%)
3.2ghz - 620 seconds (100%)
XP3200+ - 936 seconds (-51.0%)

Multithread Benchmark (2 Threads) - WinRAR, Ogg-Vorbis
FX55 - 162 seconds (+3.8%)
3500+ - 191 seconds (+22.4%)
3.2ghz - 156 seconds (100%)
XP3200+ - 259 seconds (-66.0%)

3D-Studio Max 7 - 1280x1024 - Rendering
FX55 - 132 seconds (+18.2%)
3500+ - 156 seconds (+0.0%)
3.2ghz - 156 seconds (100%)
XP3200+ - 176 seconds (-12.8%)

So what conclusions can I make?

- Before I start, i want to stress i am not saying A64 is bad or P4 is good or anything of that nature...just simply referring to progress.

When I upgraded from 2001 to 2003 from XP1600+ to 2.6@3.2ghz I saw 65-90% increases in performance for $220. This is also illustrated by the same website that did the benches I just listed P4 3.06 vs. AXP1600+

There was NO benchmark in which AXP1600+ beat the P4.

Now lets look at today:

1) A64 3500 + for $200 not only provides 0 incentive to switch over from my P4 3.2ghz since its performance on average is less than 10% greater, but it even looses across a variety of benchmarks not related to gaming. Yes I do want a processor for gaming, but I dont want it to be slower anywhere else 2.5 years after I upgrade! That is simply unacceptable.

2) To be fair, I would have had to compare 2.6 to xp1600+. So I'll compare if I upgraded from 3.2 to FX55 since I'll overclock 3500+ 2.2ghz to 2.6ghz or 2.7ghz (cache of 1mb makes up for 2700mhz 3500+ making it a fair comparison). At most, the FX55 will provide 30% increase. Again after 2.5 years this is DISSAPOINTING for anyone who wants to throw away the old AGP platform.

3) AXP3200+ was often slower to a much greater extent compared to P4 3.2ghz in contrast to P4 3.2 compared to FX55. 2 Years ago I would have never recommended anyone upgrade from xp3200+ to P4 3.2ghz since it simply was not worth it. Therefore, FX55 barely has a serious incentive to upgrade to. Again I am utterly dissappointed in CPU progress in general, be it AMD or Intel, when graphics cards double their speed practically every 12-15 months. Are engineers there much more creative? Essentially putting more pipelines is multitasking if you ask me! (Not exactly) but CPU companies should have long promoted dual- and mulit-core environments and prioritized more efficiency per clock cycle....alas...yet these CEOs get paid Millions to set corporate vision and set industry trends.

4) "Oh well, you are missing the point, you can buy dual core today." I have a problem with this statement. Again X2 processors won almost every single benchmark in the benches I've shown. Still to me there is little support for dual-core at the moment - and even they will not make a 65-90% performance boost upgrade for today's environment. Again, I am also dissatisfied knowing that 2x A64 3200+ cost less than A64 x2 3800+ if purchased separately. This highlights that dual core processors are still overpriced today.

5) To make matters worse, there is almost 0 incentive from a cpu perspective to step up from a modern cpu. Yet neither Nvidia nor ATI offer AGP based top of the line cards. So here we have thousands of AGP users like me who would not want to buy a new mobo + new cpu for MAX 30% PERFORMANCE increase just to plug in a new PCIe graphics card. 2.5 years ago if someone told you that in the next 2.5 years cpu performance would only increase by 30%, you probably would have written hate letters to both AMD and Intel for not being on the cutting edge and slacking off.

Sidetracking

6) wouldnt it then make sense for ATI or Nvidia to release AGP based high end cards given hardly any evolution in cpu performance (always unjustifiably blamed for bottlenecking)?

"The recent Diamond X1800 AGP PR fiasco actually shows a clear opportunity for ATI - with so many AGP-based systems around, high-end ones included (who won't change the whole system overnight just to fix in a PCI-E graphics card), it would be quite an opportunity to address this real and existing market.

Even the current X1800XT and X1800XL with an added Rialto PCIE-AGP bridge to fit into an AGP slot, would probably take in all the current AGP-based high-end GPU demand, as there is NO competition from Nvidia there. If you want to play H.264 HD video on an AGP system, this would be the only solution, full stop. So, while fighting the high-end battle, ATI (or any of the specialist ATI card vendors like HIS or Sapphire) could take this market."

Yet neither ATI nor Nvidia see this as an opportunity. Then why did they spend all this money trying to get PCIe chips to work with AGP based boards last generation. I bet at least 30-40% of users who buy high end cards still own what today is still considered a high end system (ie. P4 3.0/AXP3000+/S754 A64 or faster).

Basically I am simply dissapointed with CPU progress as of late and its not looking like its improving in the next 12-month period either. I am also disappointed with the graphics card companies not realizing lost opportunities.

What do you guys think?
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
I think I can encode a DVD and play Quake 4 at the same time and get faster frames than you playing alone.

I also think you should try other sites besides THG. Bad idea to put any eggs in that basket.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Originally posted by: Zebo

I also think you should try other sites besides THG. Bad idea to put any eggs in that basket.

You make a good point. The reason why I did this was for simplicity purposes. But since AMD won majority of the benchmarks, I feel that in this case Tom's was not biased towards Intel systems and clearly illustrated that A64 is the better platform today.

My point is, 2.5 years later, 30% increase in performance....graphics cards increases are around 300+%. Are cpu companies getting more credit than they should?
 

akugami

Diamond Member
Feb 14, 2005
6,210
2,552
136
Yep, with the system you have, I'd just wait for next year when dual core CPU's should fall to the $200'ish range and upgrade them. Unless you are the type to absolutely want the latest and greatest, I'd just wait. I think that dual core CPU's and dual core/card GPU's will extend certain applications by a lot (such as 3d rendering, video & audio encoding), for most mundane tasks it won' t do a thing because stuff like word processing really isn't that taxing on today's computers. For something like games, I don't think dual cores will net you much of a frame rate increase but rather it'll show up where it's less tangible to benchmarks. This would be stuff like physics and enemy AI.
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: IdaGno
Any high-end rig built today is gonna have reeeeaaaal long legs.

http://overclockers.com/tips00869/

There's a chipmakers' conference going on, and one speaker told them that clock speeds would only double or triple over the next 10 to 14 years.

Good link thanks... A Northwood/A-XP setup is still fine. Lets not kid ourselves.. AMD has had a 2400Mhz AMD64 out for two years now. .. Sure AMD has added stuff like dual channel, more cache etc to give single digit percentage performance improvements but only moved up 400Mhz in that time (FX-57). A 3.4 Northwood, also two years old is equivalent to a 3.6 Pressy.. Intel has moved up 200Mhz to 670 (from a 3.4 nothwood which is really a 3.6 Pressy) in the same time frame.

The real future is dual core which OP left off for some reason.
 

GuitarDaddy

Lifer
Nov 9, 2004
11,465
1
0
1. The first set of gaming benches show tremendous improvement IMO. Gaming FPS at 1280 x 1024 is primarily a GPU test and the CPU provides only a small percentage of the performance. So the increases you report represent a much larger increase in raw CPU power. If you rerun the test at the lowest resolution possible like 480x600(cpu limited) you will see much more significant increases

2. Your second set of benchs(video/multi) are all multithreaded(HT aware) and your comparing an HT capable CPU with 2 virtual cores to current single core chips. The fact that the higher end AMD chip still squeaks out a victory with this obvious disadvantage is pretty impressive

3. You conveniently fail to mention dual core X2's. Put an x2 4800+ in your benchs and see how your 3.2ghz chip holds up.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,250
16,108
136
Originally posted by: GuitarDaddy
1. The first set of gaming benches show tremendous improvement IMO. Gaming FPS at 1280 x 1024 is primarily a GPU test and the CPU provides only a small percentage of the performance. So the increases you report represent a much larger increase in raw CPU power. If you rerun the test at the lowest resolution possible like 480x600(cpu limited) you will see much more significant increases

2. Your second set of benchs(video/multi) are all multithreaded(HT aware) and your comparing an HT capable CPU with 2 virtual cores to current single core chips. The fact that the higher end AMD chip still squeaks out a victory with this obvious disadvantage is pretty impressive

3. You conveniently fail to mention dual core X2's. Put an x2 4800+ in your benchs and see how your 3.2ghz chip holds up.

Ditto. No choices like this above, so I did not vote.
 

RussianSensation

Elite Member
Sep 5, 2003
19,458
765
126
Originally posted by: Zebo

The real future is dual core which OP left off for some reason.

Originally posted by: GuitarDaddy

3. You conveniently fail to mention dual core X2's. Put an x2 4800+ in your benchs and see how your 3.2ghz chip holds up.

No i did not leave it out "conveniently". I even mentioned that X2 won all the benchmarks and I clearly admit that A64 is a far better platform. I am saying that X2 costs 300+. 2.5 years later, I cannot buy something for $220 to give me 65-90% performance boost I got when I upgraded from 2001 to 2003. Also, even if you upgrade to dual-core, you will hardly get this boost at gaming, or the majority of tasks that do not revolve around serious multi-tasking. Again, i believe that for majority of users there has been little incentive to upgrade systems. You could not make the same comment 2-5 years ago when upgrading meant a fairly linear and guaranteed performance increase across the board.

Also, you probably didn't read point 4) "Again, I am also dissatisfied knowing that 2x A64 3200+ cost less than A64 x2 3800+ if purchased separately. This highlights that dual core processors are still overpriced today." Plus like I said, that does not fit into my criteria of upgrading for around the same amount ($220, not $320) to get a sufficient boost in performance. But now I see that the only useful upgrade will be dual-core...so I just have to wait 5-6 more months for prices to fall ....
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: RussianSensation
Originally posted by: Zebo

I also think you should try other sites besides THG. Bad idea to put any eggs in that basket.

You make a good point. The reason why I did this was for simplicity purposes. But since AMD won majority of the benchmarks, I feel that in this case Tom's was not biased towards Intel systems and clearly illustrated that A64 is the better platform today.

My point is, 2.5 years later, 30% increase in performance....graphics cards increases are around 300+%. Are cpu companies getting more credit than they should?

I agree with your point until you look deeper.. like the overall snappiness reviewers talk about with A64's extremely low latency due to mem controller. Like sustained FPS over 30 in the extremetech article I linked to you awile back instead of averages which can be sporatic.. Like anything dual processor powered such as Cad, 3d apps, modeling... Like other sites who will show you huge gains on real apps. http://www.behardware.com/medias/photos_news/00/13/IMG0013518.gif
http://images.anandtech.com/graphs/amd%...hlon%2064%20fx57_061905100617/7574.png
 

dmens

Platinum Member
Mar 18, 2005
2,275
965
136
All the low hanging fruit in the architecture tree have been picked clean. In addition, multicore platforms have far stricter thermal specs per core than single cores, so the options for uarch enhancements have become very limited. Moreover, from a circuit pov, speedy custom design methodologies tend to draw far more power than more conventional techniques, so that too is probably a dead end.

Also, comparisons made between CPU's and GPU's don't make much sense given how different their tasks are... like if all x86 code behaved like raytrace, lol.

Maybe it is time to shift some of the burden back to the software side... ;)
 

biostud

Lifer
Feb 27, 2003
19,912
7,016
136
Without the introduction of dualcore I would be disapointed, but the CPU is rarely the limit in games today. But it would be nice to see dualcore CPU's with a large chunk of the die specialized for physics and AI. Like a 65 nm quadcore where two of the cores are normal, and two are build for parrallel calculating.
 

RampantAndroid

Diamond Member
Jun 27, 2004
6,591
3
81
I wouldn't make much sense for any AGP cards to come out.....ATI has released a X1800 something videocard in AGP form, but look at the 6600 GT from nVidia...


It has a special chip on it to transform it from its native PCIx to AGP....and the performance drop was noticable...both in overall FPS, and certain lagginess....a friend compared both and found this to be true...

So look at these more bandwidth hungry videocards like the 7800 X1800....they really won't like being converted to AGP.

PLUS....thats an extra chip to add that needs cooling...on boards that are already heavy with lots of copper!
 

n7

Elite Member
Jan 4, 2004
21,281
4
81
I would somewhat agree with the OP, since really, CPUs haven't improved all that drastically in the last few years.

I voted for 3, since it's ridiculous that ATi/nVidia are not willing to offer those of us with high end AGP systems a card to upgrade to.

However, what i don't understand is how this is news to the OP.

Sure, it's disappointing, but it's rather obvious that CPU improvements have slowed down a lot, unless you consider dual core, which really needs more multithreaded software to back up its cause.
 

TuxDave

Lifer
Oct 8, 2002
10,571
3
71
Originally posted by: dmens
All the low hanging fruit in the architecture tree have been picked clean. In addition, multicore platforms have far stricter thermal specs per core than single cores, so the options for uarch enhancements have become very limited. Moreover, from a circuit pov, speedy custom design methodologies tend to draw far more power than more conventional techniques, so that too is probably a dead end.

Also, comparisons made between CPU's and GPU's don't make much sense given how different their tasks are... like if all x86 code behaved like raytrace, lol.

Maybe it is time to shift some of the burden back to the software side... ;)

:thumbsup::thumbsup::thumbsup:
 

ribbon13

Diamond Member
Feb 1, 2005
9,343
0
0
Originally posted by: TuxDave
Originally posted by: dmens
All the low hanging fruit in the architecture tree have been picked clean. In addition, multicore platforms have far stricter thermal specs per core than single cores, so the options for uarch enhancements have become very limited. Moreover, from a circuit pov, speedy custom design methodologies tend to draw far more power than more conventional techniques, so that too is probably a dead end.

Also, comparisons made between CPU's and GPU's don't make much sense given how different their tasks are... like if all x86 code behaved like raytrace, lol.

Maybe it is time to shift some of the burden back to the software side... ;)

:thumbsup::thumbsup::thumbsup:

This is definately an area of concern. Bloated code makes bloatware. Morrowind was a great example of this a couple years ago.
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
LOL Hey I gave MS another try with thier "update" scheme last night. Ram usage with just OS went from 62MB to almost 400MB. I sure as hell hope they don't include all this crap in SP3
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
I stayed on 2000 until this year.. I'll stay XPSP2 for four more years probably.. firewall and hosts files are my freinds.
 

Cooler

Diamond Member
Mar 31, 2005
3,835
0
0
With Intel I am they should have been at ~6-7 ghz by now with netburst and now that they know its a failer they should not have reacted pressler they should just when to next gen.
 

Zebo

Elite Member
Jul 29, 2001
39,398
19
81
Originally posted by: Markfw900
I'm still on w2k, all 12 boxes !

I would have stayed but there was some app which would'nt run... I remember.. some AMD benchmark tool. So I switched....probably not so wise. I had to take about 6 hours to make perform and look like 2000.. You know disabling services.. adjusting all these properties in various windows to get lean and mean.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
27,250
16,108
136
dual-boot, I have that on 3 boxes. Hardly ever use it though