VOGONS


First post, by 386SX

User metadata
Rank l33t
Rank
l33t

Hi,

for the usual question, this time I'd like to read your opinion on this fight, the Atom D2500 Dual Core 1,86Ghz/1MB/DDR3 vs a P4 Prescott 3,4Ghz/1MB/478/DDR1. Beside the first being a dual core cpu and the much different power request, would they be similar?
Thank

Reply 1 of 11, by Koltoroc

User metadata
Rank Member
Rank
Member
386SX wrote:

Hi,

for the usual question, this time I'd like to read your opinion on this fight, the Atom D2500 Dual Core 1,86Ghz/1MB/DDR3 vs a P4 Prescott 3,4Ghz/1MB/478/DDR1. Beside the first being a dual core cpu and the much different power request, would they be similar?
Thank

no, they would not be similar. In general use (from a modren perspective) the extra core will let the atom run circles around the P4 no ifs or buts. In single threaded workloads it depends. If the application supports the modern extensions (SSE3, SSSE3) the atom will trounce the P4. If there are no optimization for modern instructions, the P4 will be faster.

When it comes to gaming or anything that makes heavy use of the GPU, it will depend if the Atom board has a PCIe slot, if you are stuck with the integrated GPU any halfway capable mid to late 2000s GPU will completely dominate the atoms GPU. They are *that* bad.

Reply 3 of 11, by 386SX

User metadata
Rank l33t
Rank
l33t
Koltoroc wrote:
386SX wrote:

Hi,

for the usual question, this time I'd like to read your opinion on this fight, the Atom D2500 Dual Core 1,86Ghz/1MB/DDR3 vs a P4 Prescott 3,4Ghz/1MB/478/DDR1. Beside the first being a dual core cpu and the much different power request, would they be similar?
Thank

no, they would not be similar. In general use (from a modren perspective) the extra core will let the atom run circles around the P4 no ifs or buts. In single threaded workloads it depends. If the application supports the modern extensions (SSE3, SSSE3) the atom will trounce the P4. If there are no optimization for modern instructions, the P4 will be faster.

When it comes to gaming or anything that makes heavy use of the GPU, it will depend if the Atom board has a PCIe slot, if you are stuck with the integrated GPU any halfway capable mid to late 2000s GPU will completely dominate the atoms GPU. They are *that* bad.

Thank, interesting. I wasn't expecting the GPU to be like that considering based on the Powervr subsystem. I've seen it still support pixel shading anyway and I read also about drivers problems with linux.
I found this Intel board D2500HN with one PCI slot and up to 4GB of DDR3 ram and I was thinking with a picopsu power supply it would run well as light pc. Also it seems it should be one of the few cpu not affected by newer bugs found on almost every cpu.

Reply 4 of 11, by 386SX

User metadata
Rank l33t
Rank
l33t
SPBHM wrote:
there was this test comparing the Atom D510 with a Pentium 4 3.4GHz https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.h […]
Show full quote

there was this test comparing the Atom D510 with a Pentium 4 3.4GHz
https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.html

single thread the Pentium 4 is way faster...

D2500 is similar to the D510 tested

Thank interesting. It looks like the D2500 have a different 32nm process but no Hyperthreading so 2 cores (64bit) and 2 threads for a 10W TDP.

Reply 5 of 11, by BinaryDemon

User metadata
Rank Oldbie
Rank
Oldbie
SPBHM wrote:

D2500 is similar to the D510 tested

Similar, except a generation newer (better IPC) and +200 mhz. A P4 Prescott will still win single threaded benchmarks but get beat in anything multithreaded or with SSE3 support. I think it really depends on your plans for the system. Gaming vs General use.

A PCI videocard wont get you very far. And yes, the powervr igpu was sort of a one off- intel doesnt support it very well. I have a D2500 and I can't use it for my DosBox Distro project (tinycore based) because XVesa has this 'ghosting' effect.

Reply 6 of 11, by Skyscraper

User metadata
Rank l33t
Rank
l33t
BinaryDemon wrote:
SPBHM wrote:

D2500 is similar to the D510 tested

Similar, except a generation newer (better IPC) and +200 mhz. A P4 Prescott will still win single threaded benchmarks but get beat in anything multithreaded or with SSE3 support. I think it really depends on your plans for the system. Gaming vs General use.

A PCI videocard wont get you very far. And yes, the powervr igpu was sort of a one off- intel doesnt support it very well. I have a D2500 and I can't use it for my DosBox Distro project (tinycore based) because XVesa has this 'ghosting' effect.

The Prescott P4 has SSE3.

New PC: i9 12900K @5GHz all cores @1.2v. MSI PRO Z690-A. 32GB DDR4 3600 CL14. 3070Ti.
Old PC: Dual Xeon X5690@4.6GHz, EVGA SR-2, 48GB DDR3R@2000MHz, Intel X25-M. GTX 980ti.
Older PC: K6-3+ 400@600MHz, PC-Chips M577, 256MB SDRAM, AWE64, Voodoo Banshee.

Reply 7 of 11, by BinaryDemon

User metadata
Rank Oldbie
Rank
Oldbie
Skyscraper wrote:

The Prescott P4 has SSE3.

Oops your right, it's lacking Supplimental SSE3 (SSSE3 or SSE3S) tho.

Reply 8 of 11, by SPBHM

User metadata
Rank Oldbie
Rank
Oldbie
386SX wrote:
SPBHM wrote:
there was this test comparing the Atom D510 with a Pentium 4 3.4GHz https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.h […]
Show full quote

there was this test comparing the Atom D510 with a Pentium 4 3.4GHz
https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.html

single thread the Pentium 4 is way faster...

D2500 is similar to the D510 tested

Thank interesting. It looks like the D2500 have a different 32nm process but no Hyperthreading so 2 cores (64bit) and 2 threads for a 10W TDP.

right, I failed to notice that the d2500 had HT disabled, but the single thread performance would be the same if running at the same clock, it was just a die shrink without architecture changes on the CPU side from one to the other.

HT means the D510 is actually a bit faster for multi theading, but the D2500 shows the advantage of the higher CPU clock, still it wont change much how things look on the tomshardware test I don't think.
https://browser.geekbench.com/v4/cpu/compare/ … aseline=3801043

Reply 10 of 11, by 386SX

User metadata
Rank l33t
Rank
l33t
SPBHM wrote:
right, I failed to notice that the d2500 had HT disabled, but the single thread performance would be the same if running at the […]
Show full quote
386SX wrote:
SPBHM wrote:
there was this test comparing the Atom D510 with a Pentium 4 3.4GHz https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.h […]
Show full quote

there was this test comparing the Atom D510 with a Pentium 4 3.4GHz
https://www.tomshardware.com/reviews/atom-d51 … ettop,2649.html

single thread the Pentium 4 is way faster...

D2500 is similar to the D510 tested

Thank interesting. It looks like the D2500 have a different 32nm process but no Hyperthreading so 2 cores (64bit) and 2 threads for a 10W TDP.

right, I failed to notice that the d2500 had HT disabled, but the single thread performance would be the same if running at the same clock, it was just a die shrink without architecture changes on the CPU side from one to the other.

HT means the D510 is actually a bit faster for multi theading, but the D2500 shows the advantage of the higher CPU clock, still it wont change much how things look on the tomshardware test I don't think.
https://browser.geekbench.com/v4/cpu/compare/ … aseline=3801043

Thank. I always found Atom processor to be better than most thought even when there was only the N270 with the external chipset. Considering SSE3 features, low power request etc it wasn't a bad Raspberry light-desktop-replacement alternative. I was thinking testing it against the P4 Prescott cause it's the only desktop pc I've built right now but the P4 is quite a beast even now for such an old cpu anyway. 2GB DDR1 in Dual and most modern kernel and light GUI and it's impressive. I find interesting that such old socket got an SS3/HT cpu so fast. Too bad they didn't make a 65nm version.

Reply 11 of 11, by Koltoroc

User metadata
Rank Member
Rank
Member
386SX wrote:

Thank, interesting. I wasn't expecting the GPU to be like that considering based on the Powervr subsystem. I've seen it still support pixel shading anyway and I read also about drivers problems with linux.

Intel is notorious for abysmal GPU performance and support, the last time this showed was with the trainwreck that was the last win10 upgrade where the Intel GPU driver was the source of one set of issues. Considering their track record in that area I find their intention to make proper gaming GPUs quite hilarious.

386SX wrote:

I found this Intel board D2500HN with one PCI slot and up to 4GB of DDR3 ram and I was thinking with a picopsu power supply it would run well as light pc.

It should be fine for office work and light browsing. If you want to watch online videos, you better hope whatever GPU you can find will have decent hardware decoding, because the CPU won't be able to do it sufficiently

386SX wrote:

Also it seems it should be one of the few cpu not affected by newer bugs found on almost every cpu.

pretty much correct. It is an in order CPU without speculative execution, that means no spectre/meltdown vulnarability. Coincidentally that is also the reason why it is that slow.

386SX wrote:

Thank. I always found Atom processor to be better than most thought even when there was only the N270 with the external chipset.

I have a netbook with one of those and a friend had a 2000 series one. Funnily enough, you could actually moderately overclock the GPU in the separate chipset with a modded driver and doing so made the GPU significantly faster than that of the newer ones. Not that the overclock made it good, about FX5200 like performance IIRC

386SX wrote:

Considering SSE3 features, low power request etc it wasn't a bad Raspberry light-desktop-replacement alternative.

intel thought so too and made something like that with the Galileo platform in 2013. Not many people agreed however so Intel canned it in 2017

386SX wrote:

I was thinking testing it against the P4 Prescott cause it's the only desktop pc I've built right now but the P4 is quite a beast even now for such an old cpu anyway. Most modern kernel and light GUI and it's impressive. I find interesting that such old socket got an SS3/HT cpu so fast. Too bad they didn't make a 65nm version.

SSE3 was developed for the Prescott and it has a 65nm version in the cedar mill core. But I assume you are referring to the socket 478 and that one had to go IIRC because the P4 got so crazy with power requirements that they had to spread the power over more pins. Socket 478 simply couldn't reliably deliver enough power to get the higher performance out of the P4 cores that intel needed because at that time they were quite badly beaten by AMDs Athlon 64.