VOGONS

Search

Common searches

Performance Discrepancy Between Frequency and L1 Cache

Topic actions

First post, by brassicGamer

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

I really couldn't think of a better title than that, so apologies if it doesn't make a huge amount of sense. The question is pretty simple, though.

66MHz is ~33% faster than 50MHz. I did some tests using CACHECHK 4.0 and measured 52MB/s for an Intel 486DX/50 and 66MB/s for an Intel 486DX2/66. That's a difference of 28%. At this level of internal CPU transaction, why is it not the full 33%? I understand about wait states coming into play for L2 and RAM, but internally shouldn't it be more 'pure'? Curious to know what it is I'm missing, as I'm not an expert on CPU architecture. Thanks in advance.

Check out my blog and YouTube channel for thoughts, articles, system profiles, and tips.

Reply 1 of 4, by red-ray

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie

Without knowing the measured the CPU speeds it's impossible to comment. Maybe the 50 MHz CPU is really running @ 52 MHz or the 66 MHz CPU is @ 64 MHz.

On the DX2/66 does the L1 cache run at 66 or 33 MHz?

Reply 2 of 4, by brassicGamer

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie
red-ray wrote on 2023-03-11, 12:31:

Without knowing the measured the CPU speeds it's impossible to comment. Maybe the 50 MHz CPU is really running @ 52 MHz or the 66 MHz CPU is @ 64 MHz.

On the DX2/66 does the L1 cache run at 66 or 33 MHz?

I checked the measurements from CHKCPU - the DX2 runs at 65.4 on my system, while the DX runs at 50.3. That goes some way to explaining it, as the difference is 30%. The internal cache definitely runs at 66MHz as it's on-die.

Check out my blog and YouTube channel for thoughts, articles, system profiles, and tips.

Reply 4 of 4, by brassicGamer

Posted on 
User metadata
Rank Oldbie
Rank
Oldbie
red-ray wrote on 2023-03-11, 15:35:

I suspect the DX2/66 L1 cache may have higher latencies than the DX/50, what are their latencies? There is also write-back vs. write-through to consider.

If that's the case only an analysis of the die itself would reveal its secrets. Comparing datasheets might produce something, but unlikely or it would be better known.

Both of these chips are non-enhanced 486 models i.e. WT cache.

Check out my blog and YouTube channel for thoughts, articles, system profiles, and tips.

Go to top of page Go to top of page
Back to General Old Hardware
Close window
Close window