Lookin for GHZ only for older software

S

sappyse107

Limp Gawd
Joined
Apr 21, 2000
Messages
510
I do a lot of engine simulations for cam specs and need GHz. cores mean nothing, technology as a whole is useless for this software so i am looking to go with cheap and overclock with good cooling. water/peltier is even acceptable for this.

Was thinking FX4170 but its been over a decade since I overlocked anything, and my forum searches for all the cpus I was looking at have left me thinking AMD, minimal cores that still offer GHz.

Good ram and MB ill be used, so just need to know the best starting point for roungly $150 investment max on the CPU.
 
FX-4350 will do the job just fine. Should overclock fairly high (4.6ghz at least, 5.0ghz+ best case) with minimal effort.
 
i was just looking up passmark scores. The 3.2 HT is 459 and the q9300 I am running now is 3955. 3.2 HT is noticeably faster.

pricewatch didn't even list the 4350. Now I wonder what else im missing haha! Thanks, thats over 4 already and if I can get 5ghz with a peltier and water cooling later, sweet:) I think a water cooled setup would be just as cool as the car parts in the shop. Start with air at 4.1 GHz.

I clicked on the APU processors and my expectation was low level, but its looking like they have the same GHz. Do they overlock as well or perhaps they are locked? I know stuff from 2003 at best when it comes to computers. Was the FX line the right choice for overlocking?
 
Don't be fooled by Passmark scores if all you're using is a single core. That benchmark is heavily multithreaded.

If you need single-threaded performance, Intel stomps on AMD. And the Core i3-4130 is definitely within your price range!

http://www.bit-tech.net/hardware/2013/11/14/intel-core-i3-4130-haswell-review/3

A3lPggT.jpg


The Image Editing test is almost completely single-threaded, and you can see Intel walk all over AMD.

Overall, Intel's advantage in everything I have seen runs anywhere from 40 to 70% faster performance (depending on the use case)! You can't make up that gap overclocking AMD (only 20% increase without an exotic board and cooler). And if you want faster than the Core i3, you can always splurge on the Core i5 4670k and overclock! That yields performance up to twice the speed of the fastest overclocked AMD chips!
 
Last edited:
Are you saying that you are running a single threaded app, thus cores doesn't matter? If so Intel should have better single threaded application speed. I'd ask in the Intel section about which processor can overclock the best for a single threaded app. You could get one of their older quads for cheap on Ebay and overclock it.

Not trying to dissuade you from AMD as I'm sure it can handle the job. Just pointing out a cheaper route to what you may want. Also the new AMD quads are coming out this month and are supposed to be pretty fast. They have the graphics card built into the processor so that would save you even more money. I'd wait and see how they overclock and their price range before I'd commit to a purchase.
 
The P4 3.2 HT is faster than a Quad core 2.5 ghz core 2 cpu, so passmark means nothing at all to me. Even the efficiency doesn't matter because you would need to use a benchmark from the mid 90s most likely. Its GHz only as far as I can tell.
 
sappyse107 said:
The P4 3.2 HT is faster than a Quad core 2.5 ghz core 2 cpu, so passmark means nothing at all to me. Even the efficiency doesn't matter because you would need to use a benchmark from the mid 90s most likely. Its GHz only as far as I can tell.
Click to expand...

No, it's single-threaded performance, which can vary widely from one core design to another. You only THINK it's pure clock speed because you only have two data points.

The Core 2 is faster than the Pentium 4 clock-per-clock, but not THAT much faster. The P4 could hold it's own depending on the type of code it was executing.

That said, the per-clock performance of Intel's Haswell is about 50% higher than a Core 2, and it also runs at 3.4 GHz, which is also faster clocked than your Pentium 4. I would be surprised if it didn't run at twice the speed.

Still don't believe me? Check this comparison of a 3.6 GHz Pentium 4 versus a Core i3 at 3.3 GHz. Then Cinebench single-threaded numbers are ALMOST THREE TIMES FASTER despite the clock speed deficit!

http://www.anandtech.com/bench/product/677?vs=92
 
Last edited:
sappyse107 said:
The P4 3.2 HT is faster than a Quad core 2.5 ghz core 2 cpu, so passmark means nothing at all to me. Even the efficiency doesn't matter because you would need to use a benchmark from the mid 90s most likely. Its GHz only as far as I can tell.
Click to expand...

The P4 3.2GHz running a single-thread program will be about as fast as 1/2 of one core of the C2Q 2.5GHz.
Hz means nothing unless you are comparing two identical processors, otherwise, you are comparing apples to oranges.


Also, check this out to see what everyone is telling you:
http://www.cpubenchmark.net/singleThread.html
 
According to that chart, the quad core is faster at this software than the P4 3.2. That is not the case, and is what I was trying to say before regarding benchmarks.
 
sappyse107 said:
According to that chart, the quad core is faster at this software than the P4 3.2. That is not the case, and is what I was trying to say before regarding benchmarks.
Click to expand...

That is why I said quite clearly: the P4 could be faster in some isolated circumstances. In much the same way, you shouldn't place such blind faith ojn pure GHz - there's more to a processor than a number.

That said, REGARDLESS OF THE SPECIAL CASE YOU HAVE HERE,the modern Haswell processors will run circles around that Core 2 Quad, let-alone your lowly Pentium 4. Give it a shot - I guarantee you won't be disappointed :D
 
GHZ rating is NOT the only thing that makes a single threaded app fast. It has a lot more to do with the instruction sets and the CPU logic than it does the clock cycles.

Exactly like was stated before: http://www.cpubenchmark.net/singleThread.html

There's a lot more to CPUs than just how fast it performs a single operation. Especially when each operation or cycle does not equate one "calculation".

If I can do something more efficiently and/or smarter than the next guy then I can be slower per operation than him and still come out on top. If it takes someone else 12 steps to do a problem but I do it in 6 I can be 50% slower per operation and still match their output.
 
Last edited:
Ah it was only a 4 year difference between my Prescott and the C2Q. I was thinking it was longer. Well, I would have thought that would be enough to make 2.5 more efficient than 3.2. Maybe not? Software is 1998 based.
 
Here's a good example with the AES instruction set:
A performance analysis using the Crypto++ security library showed an increase in throughput from approximately 28.0 cycles per byte to 3.5 cycles per byte with AES/GCM versus a Pentium 4 with no acceleration.

Obviously the AES instruction set probably has nothing to do with your software but this is a prime example of how the CPU logic itself reduced the cycles required to perform a task.

http://en.wikipedia.org/wiki/AES_instruction_set#Performance
 
sappyse107 said:
Ah it was only a 4 year difference between my Prescott and the C2Q. I was thinking it was longer. Well, I would have thought that would be enough to make 2.5 more efficient than 3.2. Maybe not? Software is 1998 based.
Click to expand...

I have a similar situation to you - stuck at work with a cross-compiler from 2002 that is single-threaded only. Moving from Pentium 4 HT to a Core 2 Duo of roughly the same speed increased performance by over 50%.

But the Core 2 is good at juggling lots of data with that huge L2 cache and 4-wide instruction decode, which probably makes a compiler feel right at home. Your processing load may be something completely different, which is why I believe you could see a small performance advantage in your older core (remember your P4 has a 30% clock speed advantage there).

Trust me when I tell you that Intel's Haswell will run miles around your Pentium 4 (and AMD's Piledriver). The only advantage AMD has is the large caches, and those are obviously not a factor for you (or else the Core 2 would own the P4 running your software).
 
Last edited:
Here's another set of Cinebench single-threaded scores for a bunch of generations of CPUs:

http://techreport.com/news/24886/haswell-compared-to-everything

They include a Pentium EE 840, which is the dual-core version of the Prescott; that scored an 0.45 on single-threaded Cinebench (ignore the multi-threaded scores). While the latest Haswell 4770K scores a 1.74 for around $330, there's also an i3-2100 that scores 1.25; that costs $136 on Amazon, but the i3-3220 which is a bit faster than a 2100 costs only $124 on Amazon.

The i3-3220 might be your best bang-for-the-buck and likely be around 3 times as fast as your Prescott for single-thread performance.

Edit: oh, and the i3-4130 is $128 on Amazon, and is a bit faster again than the i3-3220, so I would agree with defaultluser.
 
How would a Pentium G2120 (dual-core ivy-bridge) work out. Might even be able to get mobo+cpu for a little over $150.

Micro-OP cache might be helpful for your sims

Is your software hindered by a OS like XP as it is probably gonna get harder to run things on XP without driver support.
 
A socket 1155 Celeron or Pentium would do the job.

A $35 G540 out performs a (at launch) $700 Q6600 in single threaded apps.

For your user case a cheap Intel CPU makes sense.
 
Core i3 2120 was basically the same as the P4 3.2 HT. Single thread performance doesn't seem to favor a penitum 4 3.2 yet its doing alright. Will a 3rd or 4th gen intel really jump the maths up that much?
 
sappyse107 said:
Core i3 2120 was basically the same as the P4 3.2 HT. Single thread performance doesn't seem to favor a penitum 4 3.2 yet its doing alright. Will a 3rd or 4th gen intel really jump the maths up that much?
Click to expand...

We've all answered this question again and again.
Re-read our posts if you need to.
 
sappyse107 said:
Core i3 2120 was basically the same as the P4 3.2 HT. Single thread performance doesn't seem to favor a penitum 4 3.2 yet its doing alright. Will a 3rd or 4th gen intel really jump the maths up that much?
Click to expand...

Dude, stop yapping your jaw and buy a damn new processor. Unless you want to share this simulation program with me (so I can run it for you on my Core i3), the only way you're going to solve this mystery is to take the plunge!

Buy from Amazon and they will let you return it within 30 days. If it sucks, then send it back :D
 
sappyse107 said:
Core i3 2120 was basically the same as the P4 3.2 HT. Single thread performance doesn't seem to favor a penitum 4 3.2 yet its doing alright. Will a 3rd or 4th gen intel really jump the maths up that much?
Click to expand...

Again it should be roughly 3 times as fast. Single core performance is a function of IPC and frequency. The P4 had 1/3 the IPC of a modern day i3.
 
NaroonGTX said:
FX-4350 will do the job just fine. Should overclock fairly high (4.6ghz at least, 5.0ghz+ best case) with minimal effort.
Click to expand...

Thanks, now to find a motherboard and ram:) I was sad to see Abit quit making motherboards in 2008.
 
Is it just me, or does this feel like a Happy Hopping thread?
 
Red Falcon said:
Is it just me, or does this feel like a Happy Hopping thread?
Click to expand...

I do like how he ignored every single benchmark, the fact that I posted a link with the very same OVERCLOCKED FX-8350 getting destroyed by Intel, and also ignored my experience optimizing platforms for outdated software, along with my offer to try out his software oon my Core i3 :D

Yeah, in one ear and out the other. Have fun on my ignore list, sappyse107.
 
defaultluser said:
I do like how he ignored every single benchmark, the fact that I posted a link with the very same OVERCLOCKED FX-8350 getting destroyed by Intel, and also ignored my experience optimizing platforms for outdated software, along with my offer to try out his software oon my Core i3 :D

Yeah, in one ear and out the other. Have fun on my ignore list, sappyse107.
Click to expand...

Sounds good. Your benchmarks proved that benchmarks meant nothing for my application. Accept it, deny it, it means nothing to me.
 
The OPs observations don't seem unreasonable if you make a few assumptions about the code he's running. The P4 had some odd characteristics in that it had a very long pipeline and suffered badly from missed branch predictions. In other words, it was slow at running code with a lot of "if" statements. P4s did pretty well at grinding through big piles of simple math operations though. If you think back a bit you might recall that P4s were rather popular in HTPCs for a while. A modern system's solution to big piles of math operations with few conditionals are vector instructions that do more than one operation at a time and offloading the number crunching to the GPU. GPUs have hundreds or thousands of cores that are really bad at conditionals but decent at piles of math operations. If you assume the OP's app just does lots of math, is single threaded, has few conditionals, was optimized for a P4, and doesn't know how to use modern vector instructions or a GPU it's not at all unreasonable that a P4 3.2 could beat a C2Q.

Some benchmarks will apply to the OP's application, but likely not a lot of the usual current generation benchmarks. Most application based benchmarks are too branchy and any sort of current game or media benchmark is going to utilize CPU features or GPU capabilities the OP's app probably can't use. An old scientific computing or multimedia benchmark would likely be more representative.
 
zandor said:
The OPs observations don't seem unreasonable if you make a few assumptions about the code he's running. The P4 had some odd characteristics in that it had a very long pipeline and suffered badly from missed branch predictions. In other words, it was slow at running code with a lot of "if" statements. P4s did pretty well at grinding through big piles of simple math operations though. If you think back a bit you might recall that P4s were rather popular in HTPCs for a while. A modern system's solution to big piles of math operations with few conditionals are vector instructions that do more than one operation at a time and offloading the number crunching to the GPU. GPUs have hundreds or thousands of cores that are really bad at conditionals but decent at piles of math operations. If you assume the OP's app just does lots of math, is single threaded, has few conditionals, was optimized for a P4, and doesn't know how to use modern vector instructions or a GPU it's not at all unreasonable that a P4 3.2 could beat a C2Q.

Some benchmarks will apply to the OP's application, but likely not a lot of the usual current generation benchmarks. Most application based benchmarks are too branchy and any sort of current game or media benchmark is going to utilize CPU features or GPU capabilities the OP's app probably can't use. An old scientific computing or multimedia benchmark would likely be more representative.
Click to expand...

Nailed it. I don't know that it has any conditionals. I give it lots of numbers and it gives me lots of numbers back.
 
something like an fx 4350 might be a better option than the 4170.

The FX processors are unlocked multipliers, so you would just need to adjust the voltage and set the mutliplier to whatever you want. Some of those CPUs hit 5 ghz, but 4.7 is probably as close to a sure thing as one can get when over clocking.
 
defaultluser said:
No, it's single-threaded performance, which can vary widely from one core design to another. You only THINK it's pure clock speed because you only have two data points.

The Core 2 is faster than the Pentium 4 clock-per-clock, but not THAT much faster. The P4 could hold it's own depending on the type of code it was executing.

That said, the per-clock performance of Intel's Haswell is about 50% higher than a Core 2, and it also runs at 3.4 GHz, which is also faster clocked than your Pentium 4. I would be surprised if it didn't run at twice the speed.

Still don't believe me? Check this comparison of a 3.6 GHz Pentium 4 versus a Core i3 at 3.3 GHz. Then Cinebench single-threaded numbers are ALMOST THREE TIMES FASTER despite the clock speed deficit!

http://www.anandtech.com/bench/product/677?vs=92
Click to expand...


A code branch prediction miss on the P4 stalled the whole chip for 70+ clock cycles. That was one of the design flaws of the P4. They tried to make up for it with more clock cycles.

I'd be looking at the highest clocked i3 chip you can find.
 
get a cheap i7 920 and OC it? Last I saw they were going for <$70

Alternatively you can get a haswell i3 with a high clock speed.

I'd stick with a dual core if possible so the application can hog a core without bogging down the machine
 
He's already getting an AMD FX processor, and has ignored everything everyone has posted in this thread.
Just look at his history, before this thread, he hasn't been on since 2009.

You probably won't get a response until 2019, if ever, so don't waste your breath on this guy. ;)
 
Red Falcon said:
He's already getting an AMD FX processor, and has ignored everything everyone has posted in this thread.
Just look at his history, before this thread, he hasn't been on since 2009.

You probably won't get a response until 2019, if ever, so don't waste your breath on this guy. ;)
Click to expand...

Tell em Steve Dave.
 
Check out the steep system specs for the new software update.

System Requirements:

  • A Windows-compatible PC with a CD-ROM drive (or you can request a package with a USB Thumb Drive for program installation).
    • A fast system processor (recommended 3GHz or faster) will improve processing speeds; especially helpful for Wave-Actionand Iterative testing. However, Dynomation will operate on any Windows system, regardless of processor speed.
  • A USB Port for the Security Key required to run Dynomation-6.
  • A minimum of 2GB of RAM (random access memory) for Windows Vista, Win7 and Win8 and Windows10.
  • Fully compatible with Windows Vista, 7, 8 and Windows10 (32- and 64-bit). Earlier versions of Windows not supported.
  • A video system capable of at least 1280 x 1024 or higher to optimize screen display of engine components and performance analysis graphics.
  • Mouse
  • Windows-Compatible Printer required for test reports

The i3-4130 worked out well. No help needed this time, just wanted to post because its 2019.
 
You must log in or register to reply here.
Back
Top