AMD FX "Vishera" reviews info (again, after mod mistake)

Hey wez...
Our chips seem to have come from the same batch (maybe even the same wafer :p:). Mine is a 1237 PGN, Serial #0126.

I've been pushing the limits of stability quite a bit.
Although it seems daily stable @ 4.95/1.50v (a # of different stress tests), I've since had a few quirks when encoding some HD video...

Even though temps are in check my chip gets a little squirley whenever my Vcore is >1.485 (ambient water).
I assume it's electron-migration... Just wondered if you've noticed the same thing with your chip?
I've rolled my OC back to 4.83/1.48125 and the encoding stability problems seem to have went away... :cool:

Don't get me wrong!
I can push it up to 5.3+ Ghz with voltages >1.525 and pass a lot of benchmarks, but real stability fades ~1.485v.
Maybe I just didn't get the best sample... Wouldn't be the first time... :rolleyes:

Quote:

---

Originally Posted by Daveburt714

Hey wez...
Our chips seem to have come from the same batch (maybe even the same wafer :p:). Mine is a 1237 PGN, Serial #0126.

I've been pushing the limits of stability quite a bit.
Although it seems daily stable @ 4.95/1.50v (a # of different stress tests), I've since had a few quirks when encoding some HD video...

Even though temps are in check my chip gets a little squirley whenever my Vcore is >1.485 (ambient water).
I assume it's electron-migration... Just wondered if you've noticed the same thing with your chip?
I've rolled my OC back to 4.83/1.48125 and the encoding stability problems seem to have went away... :cool:

Don't get me wrong!
I can push it up to 5.3+ Ghz with voltages >1.525 and pass a lot of benchmarks, but real stability fades ~1.485v.
Maybe I just didn't get the best sample... Wouldn't be the first time... :rolleyes:

---

Hah, wafer brothers... :up: :rofl:

I'll start with some ranting.. mostly motherboard related.
Haven't focused too much on core clock with this chip yet, mainly because its nothing special in that area.
Have noticed that I get ht ref fluctuations (~3), which sort of make things a bit tricky to dial in.
Vcore is pretty wild, and in the first place I have to over compensate a lot to get set value at idle. But it will also increase by ~0.04v under load.
Set it up for ~1.6v in bios, get 1.5v at idle and 1.54v under load.
Now, its very much workable with some patience, and its mostly when really pushing it that these small issues pose a problem. But thats where i like it..so yea.. :D

Daily non prime stable limit is pretty much what you are getting (4.95ghz). Funny thing is, passing cine at over 5ghz have proven to be a real challenge. More and more my board seem to be the culprit, but its still hard to say. Could just as well be a super picky core that just dont want to play nice.

If I want it stable in prime, ~4.8ghz is the limit, give or take. (Vcore around 1.5v, not 100% dialed in). My cooling just yawns its way through anything I've tried. High clocked 8 core chip with 1.56vcore and 1.36v cpunb, and load temp is barley reaching above 55C? My 8120's would have melted a long time ago lol. Socket and cores show similar values, so I dare to say temp is accurate enough. All voltages are measured with dmm.

I've started using x264 fhd benchmark for a quick way of testing, and its not that far off from what it would take in prime. So I'm not surprised you are seeing issues with encoding :D
You could very well be onto something with the chip not liking high vcore. Your chip have a high VID as well? (mine is 1.375v).

Quote:

---

Originally Posted by wez

I run 1.35-1.36v for 2650mhz imc on a 8350. I'd keep pushing as long as the freq keep scaling with voltage, and heat load can be kept under control/not sacrificing cpu core freq.

Upping the voltage for nb/sb/ht can help with stability. Running high ht ref is generally where you'd notice a difference. And if you are getting weird crashes/bsods that you cant really figure out, a decent voltage bump on these might do the trick (++0.2v).

Regarding VDDR, I haven't touched it when running vishera. But I'd be surprised if it did anything noticeable, at least with air/water.

---

Yah for my 8120 I'm only at 1.2375v for CPU/nb
For 2500mhz I guess vishera is different however I leave HT ref clock on auto

A question.
So, a friend of mine who bought 8320 ,and me, were doing some quick comparisons clock for clock (at @3.5ghz)
I was curious about a gain i would have when i switch, we did fritz ,convertxtodvd (encoder i often use) In fritz we about tied (i was a bit faster), in convertx he was noticably (around 17%) faster, all looked normal to me.
But when we fired up linx, i got somewhat surprised.
I got 61Gflops, he got half that.I checked around interwebz ,and the relation seems to be true.For second there i thought that smaller number means better, i checked intel scores, a 5ghz 2600k did same as my 3.5ghz phenom, so, i think not.
Anyway, anyone could elaborate on this matter ? Why is that, what it actually does, how it compiles this number.I mean.Wth ?

if anyones interested.
Cinebench 11.5
PH II 6.02
8320 6.05 (with a 4.0 turbo, however i dont think it matters much in this test)

Quote:

---

Originally Posted by vario

A question.
So, a friend of mine who bought 8320 ,and me, were doing some quick comparisons clock for clock (at @3.5ghz)
I was curious about a gain i would have when i switch, we did fritz ,convertxtodvd (encoder i often use) In fritz we about tied (i was a bit faster), in convertx he was noticably (around 17%) faster, all looked normal to me.
But when we fired up linx, i got somewhat surprised.
I got 61Gflops, he got half that.I checked around interwebz ,and the relation seems to be true.For second there i thought that smaller number means better, i checked intel scores, a 5ghz 2600k did same as my 3.5ghz phenom, so, i think not.
Anyway, anyone could elaborate on this matter ? Why is that, what it actually does, how it compiles this number.I mean.Wth ?

if anyones interested.
Cinebench 11.5
PH II 6.02
8320 6.05 (with a 4.0 turbo, however i dont think it matters much in this test)

---

The problem is zambezi/piledriver still has a single decode to 4 FlexFPU (FPU = floating point unit)
while it is still executing the the full 128 bit of it for all 8 core it can only bring 2 way decoders to the cores and each side of Flexfpu and the Flexfpu is dual threaded to work with either core.

Steamroller addresses this by giving it two 4 way decoders one to each integer and both decoders go to the FlexFPU since it's dual threaded.

Linx is not supporting BD family of cores properly,that's why you see very low score there. It should be scoring roughly 2x of what he gets if Linx supported FMA ISA.

Thanks for quick replies, its quite strange to see such results, even sandy/ivybridge is much slower than phenom II in this.
On a side note, prime errors out on stock with my friends 8320... Is this some kind of incompatibility ? (large ffts)

http://www.techpowerup.com/175902/RO...e-Records.html

AMD FX-8350 over 8600 MHz!

Quote:

---

Originally Posted by vario

Thanks for quick replies, its quite strange to see such results, even sandy/ivybridge is much slower than phenom II in this.
On a side note, prime errors out on stock with my friends 8320... Is this some kind of incompatibility ? (large ffts)

---

Yeah there is some issue with prime and Bulldozer/PD. Just ignore it,everything is fine.

http://i1149.photobucket.com/albums/.../cine83503.jpg

wez: The VID on my chip is 1.3625, so pretty close... :p:

This 8350 runs MUCH cooler than either of my 81xx's (lower power consumption too)!
Like you, I'm still testing, but my chip does seem a little touchy when it comes to voltages.

I'm nowhere near done yet, but I've worked my IMC quite a bit and scaling slows drastically ~2.5/1.25v even though temps are in check.
I've pushed v's higher but with diminishing returns, could have something to do with high mem clocks I guess . :shrug:

The x264 FHD/Maxwell Render benchmarks seem to be really good stress tests for daily stability (hours of Prime95 just seems like overkill with these chips!).
Then again Encoding video is the most stressful I do, and even they can fail when running RipBot/HandBrake... :rolleyes:

Tbone: Although it doesn't offer huge benefits (slight improvements in SSD scores & 3D benches), running HTLink speeds matched to the IMC hasn't caused any problems for me so far (unlike PHII/Thuban/BD).
Keeping it 1 or 2x lower while finding max clocks is probably a good idea, but once you get things things stable you should be able to bring it up to the IMC speeds without failures... :)

What max bus speed average on these chips
Currently running 236 and 3078 rated fsb 2800NB

is there a list of stable overclocks for the vishera (8350/8320) series?

whats average 8320 overclock 4.6-4.8?

My FX-6300 tops out at 4.5GHz @ 1.47v on IBT on Very High / Max. If it can pass that, it is considered 'stable' for me.
My chip isnt very good, but I guess that is to be expected since FX-6300s are probably 8320 rejects, which dont clock as well as 8350s.

Quote:

---

Originally Posted by tbone8ty

is there a list of stable overclocks for the vishera (8350/8320) series?

---

I haven't seen one on XS... I'm certainly not the best at pushing benchmarks, but I am pretty good at finding max stable OC's (for my gear/uses anyway). :p:
From what I've seen the 8350's tend to clock slightly better but it's still luck of the draw.
My chip is average at best and it wouldn't surprise me to see a nice 8320 beat it, guess it just comes down to how important $30-40 is to you and how much a potential few Mhz matters... ;)

I've had this 8350 since their release and had what seemed "Daily Stable" ~4.95Ghz.
After a few crashes while encoding HD video I had to back it down a little... :rolleyes:

Here's what I'm currently running for the last week or so, it's been very stable and hasn't failed even while doing multiple encodes...

http://i30.photobucket.com/albums/c3...12_Prime18.png

Here are some of my usual benches/stress tests at the above clocks (it's an old version of AIDA I don't have a modern key :rolleyes:):

http://i30.photobucket.com/albums/c3...12_Benches.png

Quote:

---

Originally Posted by tbone8ty

whats average 8320 overclock 4.6-4.8?

---

From what I've seen, with good cooling most chips should fall into that range...
Personally, I'd be surprised if you couldn't get an 8320 stable @ 4.6 on aftermarket air! ;)

nice Dave, can you test for me stability in AMD Overdrive (there is stability test, looks good and stressed every part of CPU, uncore, RAM)? I testing AMD CPUs in this test. I dont much believe with newest CPUs PRIME (my i7-3770K was PRIME stable and sometimes crashed in idle or some benchmark!)

Had a full day playing around with 8350 up to yet 4.8 1.38vcore stable. Mem 2200MHz Cas 6-10-10-27-40 1T.

http://imageshack.us/a/img685/9307/482200cas6.th.png

Pretty good.

Have you tried 6-10-7 @ 2200?

Just tryed it thanx for tip mate

http://imageshack.us/a/img11/8641/48...timings.th.png

4.97Ghz X264 run
http://img10.imageshack.us/img10/826...ghertiming.png

Uploaded with ImageShack.us

This chip is amazing 5.1Ghz X264 Run.Turning into a AMDfanboy lol.

http://img822.imageshack.us/img822/8...tighertimi.png

Uploaded with ImageShack.us

Well that is some amazing OC for sure :)

wooow Bro! Whats your batch of CPU? Awesome run with still decent voltage!

Quote:

---

Originally Posted by FlanK3r

wooow Bro! Whats your batch of CPU? Awesome run with still decent voltage!

---

I will have to let you know not checked batch number yet amazed by this chip. Currently running tests at 5.25 now:shocked:

8320 max clock around 5.1Ghz...