Change MoBo: feel "snappier"

[manu2b]

Hi all,

I put my 955BE in a m5a97 Pro (from a GA-790xt-USB3), and other than it needs 0.04vCore less to achieve 4.14GHz, the whole system feels much more responsive.

Is it in my mind?

Is there a way to "validate" that feeling?

 

[BeepBeep2]

Hi all,

I put my 955BE in a m5a97 Pro (from a GA-790xt-USB3), and other than it needs 0.04vCore less to achieve 4.14GHz, the whole system feels much more responsive.

Is it in my mind?

Is there a way to "validate" that feeling?

...
That is because the M5A97 Pro overvolts a little. I would imagine that the 790XT-USB3 undervolts a little too in reality.

Your system is not more responsive. In fact, it is slower than it was before. 9x0 series boards, especially ASUS ones, are heavily optimized for Bulldozer FX while almost not at all for Phenom II.

If you were to run SuperPi you would notice that it was ~5 seconds slower on the new board than the 790FX board with the exact same settings. The only thing I can think of would be that HDD performance is better on the ASUS board, older Gigabyte boards were a bit slow for some reason.

 

[manu2b]

...
That is because the M5A97 Pro overvolts a little. I would imagine that the 790XT-USB3 undervolts a little too in reality.

Your system is not more responsive. In fact, it is slower than it was before. 9x0 series boards, especially ASUS ones, are heavily optimized for Bulldozer FX while almost not at all for Phenom II.

If you were to run SuperPi you would notice that it was ~5 seconds slower on the new board than the 790FX board with the exact same settings. The only thing I can think of would be that HDD performance is better on the ASUS board, older Gigabyte boards were a bit slow for some reason.

Mmm, interesting. I am going to run some HDD (SSD) tests...

 

[manu2b]

...
That is because the M5A97 Pro overvolts a little. I would imagine that the 790XT-USB3 undervolts a little too in reality.

Your system is not more responsive. In fact, it is slower than it was before. 9x0 series boards, especially ASUS ones, are heavily optimized for Bulldozer FX while almost not at all for Phenom II.

If you were to run SuperPi you would notice that it was ~5 seconds slower on the new board than the 790FX board with the exact same settings. The only thing I can think of would be that HDD performance is better on the ASUS board, older Gigabyte boards were a bit slow for some reason.

So, I ran some tests and you were 100% right.

Got a (very) tiny improvement on random RW (like 1.5%, within the error margin).
And a (very) tiny loss in SuperPi (around 0.3% still within the error margin).

Result? Phenom II is back in its Gigabyte board, and the Asus is in the closet waiting for either a PD (if good enough...) or Ebay!

Thanks again BeepBeep2.

 

[trents]

Why would overvolting make any difference in performance? It doesn't change the frequency.

 

[manu2b]

We were talking about the cpuz displayed vcore difference between the 2 boards at an identical given frequency.

 

[Neuromancer]

Wow so you are going with a system that feels slower but benches better. Why not get an intel setup then??

It is obvious that you and the person that answered your querstion do not understand computing at all.

Did you do a fresh install on 990-FX? or at "fairest". restart 3 times?

I wouldn't suggest going to 990... unless pushing performance boundaries.. and financial ones.. and screw SLI.. who wants it who needs it.. leave SLI for the stupid people. yhe hybrid crossfire crowd.

for thuban though... its nice. I didnt have 890FX to compare it to. But I still run my 790GX board.I am on it now. Despite x79 siting next to me./

 

[manu2b]

^ A bit harsh

- The beepbeep answer seems logical to me: feel snappier because of a better HDD/SSD management, and bench lower because not optimized for Phenom II but BD's.
- I used to have a 2600K running@5GHz/24.7 (sold due to need of cash) and I miss it!
- Yes, fresh install with both Win7 Ultimate and Win8 Preview, both with a PhIIx4 and a FX-8120.
- I use 5830 xFire, and don't really have issues with it (can only remember of Rage actually...)

EDIT: the bench results difference was not between the FX-8120 and the PhII, but with the PhII in the 790 and then in the 970 boards. The large amount of beers that I had yesterday might have impacted the logic of my (kind of) reasoning!

 

[BeepBeep2]

Why would overvolting make any difference in performance? It doesn't change the frequency.

It doesn't. The memory subsystem is slower, secondary subtimings in BIOS that you can't set are looser (and ones you CAN set are looser by default). However Gigabyte boards were rather slow for some reason as far as SATA performance.

We were talking about the cpuz displayed vcore difference between the 2 boards at an identical given frequency.

For the most part, correct...
Really it has to do with you setting 1.45v on the gigabyte board and it coming out to less than 1.45v real while the ASUS board overvolts to about 1.47v real.

I am guessing about the Gigabyte board, but I've measured the M5A97 Pro myself with a DMM ...girlfriend's rig (doesn't live here, I am a minor, she is a college student) has one.

Wow so you are going with a system that feels slower but benches better. Why not get an intel setup then??

It is obvious that you and the person that answered your querstion do not understand computing at all.

Aren't you over at TechReaction? I would have figured someone of your type that is always talking out your (mouth) as a 'reviewer' would know a little better.

Did you do a fresh install on 990-FX? or at "fairest". restart 3 times?

As far as your comments on a fresh install, of course it was a fresh install on both fronts. I've also tried it with the same damned OS, Win XP SP3, Vista x86, and 7 x64 because drivers are universal from 7x0 to 9x0. I didn't test this with a 24/7 setup.

I wouldn't suggest going to 990... unless pushing performance boundaries.. and financial ones.. and screw SLI.. who wants it who needs it.. leave SLI for the stupid people. yhe hybrid crossfire crowd.

Was I unclear?

AGESA microcode in 990FX boards is optimized for AMD FX CPUs. 990FX boards with late BIOS are only good for running Phenom II at stock frequencies. I have to run a BIOS (beta 0051, does not POST with retail Bulldozer, to overclock Thuban. (ie. run anything over DDR3-1333, boot anything but stock CPU_NB)

Even then, it is 2-3 seconds slower than my 890FX M4A89TD Pro/USB3 board in SuperPi 32M on any OS at same or better sub-timings, with copy-waza, no copy-waza, maxmem or no maxmem. I really don't want to waste my time showing you this, and I no longer have a gigabyte board to show you those SATA throughput results either.

Take chew*/bingo13 (ASUS rep)'s word on 990FX if I'm really that clueless.
http://www.xtremesystems.org/forums...ir-V-Formula&p=4964464&viewfull=1#post4964464

Looks like it's preety artificial performance speaking.

4468 runs should have easily been sub 15's even on an untuned win 7 OS.

Any EFI above beta 0027 starts getting BD support and from that point forward it becomes more and more difficult to run Thuban or Deneb up high, sorry but the AMD AGESA code changes dictate most of this setup. After all, BD was supposed to be a June release when the boards came out. Starting with 0705 on the public release this is the first full BD optimized AGESA code so Thuban/Deneb overclocking capabilities take yet another step backwards

for thuban though... its nice. I didnt have 890FX to compare it to. But I still run my 790GX board.I am on it now. Despite x79 siting next to me./

That's nice. I wish I could have money and opportunities like you, while knowing what I was doing.

 

[trents]

Man, I'm utterly confused by this thread. Between the "beers" and the "mountain of snow" this thing got very hard to follow.

 

[manu2b]

Man, I'm utterly confused by this thread. Between the "beers" and the "mountain of snow" this thing got very hard to follow.

 

[ice445]

I'm going to chime in and say the latest AGESA code flat out doesn't even support C2 Phenom II's at all. After I flashed that BIOS, oh man. Instability heaven even at 100% bone stock.

 

[Neuromancer]

Man, I'm utterly confused by this thread. Between the "beers" and the "mountain of snow" this thing got very hard to follow.

I get old man crabby late at night. I should stay off the interwebs Friend of mine calls it sundowners syndrome. I think my natural asshatedness just glows brighter after midnight.

Anyway

thanks for the clarification beepbeep2. Good read over at XS too. Interesting results. I had no problems clocking thuban on 990FX but didn't compare it to a CH4 for performance results. Very interesting stuff if true.

 

[freeagent]

Wow so you are going with a system that feels slower but benches better. Why not get an intel setup then??

Now now thats not nice

Im down in the 45ns range, with tri channel and low volts, thats amd 939 territory right there

 

[Neuromancer]

yeah... you hit it on the head though. its ALL about the mem latency.

Sorry for crapping your thread up manub2.

I would use the system that "feels" snappier. But thats why I use AMD because it feels snappier. (well and costs less minus 990FX sli tax).

 

[manu2b]

Just sold the fx-8120 this morning, and destroyed (unloaded some lead on) the m5a97 Pro: Asus RMA/customer service sucks big time (at least here in Europe): I could not RMA the board due to "transfer of property" not done with previous owner...
So, happily back to my loyal PhII and Ga-790xt-USB3!

 

[freeagent]

You are right Rich, its all about the latency. Im on my office rig at work right now, a dell with a gig of ram and an a64 3800 single core...

Id love to go back to amd if I didnt have this as a referance

I learned something new thanks to this thread.. I would have takein me awhile to figger out how I lost 3s in spi

Glad to hear you got everything sorted out OP!

I kinda want to build an AMD rig now, just to fart around with..

 

[Randyman...]

Could also be related to Deferred Procedure Calls (DPC's) and specifically "DPC Latency". Some MoBo's (actually the associated drivers) can stomp all over an otherwise screaming fast PC. Run these on both different MoBo's and see if one has a high DPC latency:

http://www.thesycon.de/deu/latency_check.shtml

http://www.resplendence.com/latencymon

This is common knowledge over in Audio Production land as high DPC Latencies will bring a killer PC's "Realtime Audio Performance" to a crawl (basically makes it a paperweight). Killing a single device in the Device Manager or swapping drivers can sometimes make worlds of difference WRT DPC latency. This is often due to LAN/WiFi drivers and sometimes Video Drivers, but any driver is a possible culprit.

Anything over 200us or so is too high for realtime audio, but you aren't likely to "feel" this latency until it gets well above 1000us (1ms per Deferred Procedure Call!). My DAW PC's stay well below 50us and average around 20us.

Could be other things as previously mentioned, too

 

[BeepBeep2]

I know about that
One of my best friends screws with FL Studio and Reaper for a hobby.

However the drivers between these boards are all the same.

 

[Randyman...]

Asus's page for this MoBo is down - so I'm having to use less reputable pages to retrieve the specs from the Asus M5A97 PRO. But I'm actually seeing quite a few chips that are different (SB750 to SB950, different NIC revisions, etc etc). That's all it takes to find the actual driver (in the unified driver package) that is giving you DPC issues (and thus slightly detectable "snappiness" results)

Not saying it's for sure, but certainly possible unless you are swapping identical hardware. Let's not forget the BIOS itself - it has been a source of DPC issues in some MoBo's...

High DPC's have been the bane in many a DAW-PC owners existence. Been paying very close attention to DPC's for a good 5-6 years now as my DAW-PC DIY Prowess grows (PC#2 in my sig)