Is the Sapphire 4890 Vaporx for sale yet. Where

[Cool Mike]

Thanks for the info...

 

[satelitko]

Thanks for the info...

x1

 

[Blacklash]

I got the card and compared it to my GTX 260 Core 216 card in my main rig. That card is clocked @ 711|2300. In AoC @ 1920x1200 with filtering and all enabled the Toxic is 13FPS faster. That's DX9. Under DX10 it is only +2FPS.

It idles @ 43C and loads in the mid 60s with the fan left on auto. I couldn't hear the fan at any time during gaming. It blends nicely with my cases fans which I run on low.

Now it's getting moved to my AMD rig. It's definitely a well built high quality card.

I haven't tried overclocking it yet.

It's a great card if you want stay away from multi-GPU yet have solid performance.

The reports of it being very quiet and cool are right on the money.

 

[Cool Mike]

Blacklash, I am interested in your Toxic overclocking abilities. Core and memory. Is it possible to get us some numbers. most have been getting 1050-1060 core and 1200+ memory. I have heard of 1250 memory, but that hard to believe. Thanks

 

[erocker]

Don't push your ram that far unless you want to burn up your card.

 

[Cool Mike]

With respect to all. When overclocking, the core or memory will not be damaged as long as the Core/memory voltages are maintained at factory spec's. Now when you push the core and/or memory to fast you will pixilate and the card will go out of sync and even reset. I have never had a card die. Many enthusiasts will try increasing the core and or memory voltages (by modifing the board) to give them higher amplitude clocks which in turn can significantly raise core temps. This is what kills the core and or DDR memory if you don't know what you're doing. too high voltages.
Now the vaporx series of cards are among the best cooling solutions. That is why sapphire gives you a 1100 core and 1300 memory max in CCC. They would not give this capability if it would damage the product. Overclocking not only depends on the graphics card but other hardware related factors: Adeqate/quality of power supply, Motherboard quality, Proper cooling, clean driver install are a few.
Just the basics... only my opinion and knowlege as a design engineer.

 

[erocker]

With respect to all. When overclocking, the core or memory will not be damaged as long as the Core/memory voltages are maintained at factory spec's. Now when you push the core and/or memory to fast you will pixilate and the card will go out of sync and even reset. I have never had a card die. Many enthusiasts will try increasing the core and or memory voltages (by modifing the board) to give them higher amplitude clocks which in turn can significantly raise core temps. This is what kills the core and or DDR memory if you don't know what you're doing. too high voltages.
Now the vaporx series of cards are among the best cooling solutions. That is why sapphire gives you a 1100 core and 1300 memory max in CCC. They would not give this capability if it would damage the product. Overclocking not only depends on the graphics card but other hardware related factors: Adeqate/quality of power supply, Motherboard quality, Proper cooling, clean driver install are a few.
Just the basics... only my opinion and knowlege as a design engineer.

You make very good sense. The reason I'm saying not to overclock Qimonda GDDR5 too much is because of previous issues with 4870's and people blowing out their ram chips/vregs by pushing them too far. On many 4870's/4890's the ram vregs are already quite hot at stock settings. Regardless of what parameters a company gives to to overclock, damage can still be done. My opinion as an "amateur" mechanical engineer and an experienced ATi HD 4XXX user who's wrecked a 4850 and a 4870 already.

*Bah, I should add that it's fine to try, but make sure you have something reading your GDDR5 temps and VREG temps. If they are getting too high, back down or it will kill them.

 

[Cool Mike]

good point. I believe this has been an issue with numerous vendors with the 4870's. Wasnt this addressed with the 4890's? Thinking the toxic's regulator's are heat sinked? Excessive heat will reduce the life of the card. But, for bench testing you would br alright.
I wish I could get my hands on the schematics! With some well thought out hardware/heat sink upgrades the Toxic or Atomic could run 1100 core easy!

 

[erocker]

good point. I believe this has been an issue with numerous vendors with the 4870's. Wasnt this addressed with the 4890's? Thinking the toxic's regulator's are heat sinked? Excessive heat will reduce the life of the card. But, for bench testing you would br alright.
I wish I could get my hands on the schematics! With some well thought out hardware/heat sink upgrades the Toxic or Atomic could run 1100 core easy!

Yeah, I think getting 1100 core shouldn't be too hard with the proper mod. The Toxic cooler can probablly handle it too! The hottest chips on my 4890 right now are the ram vregs which sit at 58c idle. It will be interesting to see what Nvidia does with GDDR5. I'm also wondering if it's just a Qimonda thing and I'd like to see another manufacturer with GDDR5 for some comparisons. I did get a 18,400 3d06 score with 1000/1100 and 4.5ghz on my E8600!

 

[Blacklash]

Blacklash, I am interested in your Toxic overclocking abilities. Core and memory. Is it possible to get us some numbers. most have been getting 1050-1060 core and 1200+ memory. I have heard of 1250 memory, but that hard to believe. Thanks

My card is not as good as erocker's. I've got 1030 on the core and 1100 on the memory stable. People that are getting 1200+ on the memory are scoring darn good cards. I can do suicide runs at higher clocks, and can do quick tests @ 1060 on the core. It's not long term stable though.

Of course I didn't try testing with the fan jacked up yet. I might be able to get better numbers moving it off of auto. It only runs @ 38%. I'll report back later with tests done on higher fans speeds.

EDIT: Yep that was it. Moving the fan to 60% bumped me to 1050 on the core and 1130 on the memory. I may try higher fan speeds and I think that's really going to be it for the card. I've got a decent core and average memory.

 

[erocker]

Here is the bios for those who need it.

 

[ShogoXT]

I guess you can only do so much with my cards sandwiched together. Happy that you guys are getting good results though. Nice cards right?

 

[reverze]

Love it.

Will hold me over for a while.

 

[Cool Mike]

Blacklash, looks like you have a good sample. 1050 is inline with others. Some hitting 1060 stable. My Asus Tweak ddr5runs 1125 stable with the low end reference cooler. I base stable (at this time)on 2 solid hours of Crysis warhead max settings play. Thanks for the feedback. Let me know if anyone hits DDR5 1200 stable. That would be sweet

 

[reverze]

What's everyones Vantage scores?

 

[erocker]

My GPU score was right around 11,400.

 

[Cool Mike]

I would like to see some toxic vs. Atomic comparisons. Just a thought, correct me if am wrong. Core overclocking is on par with both, But the atomic seems to overclock memory higher.
Any 4890 Atomic overclockers out there?

 

[Blacklash]

In games where my card puts up 45-65FPS @ 1920x1200 the difference between 1000|1100 and 1050|1130 is all of 2FPS. I plan on running the lower clock with 38% fan speed.

 

[Wile E]

Why is everybody so obsessed with the ram speed? Overclocking this GDDR5 isn't going to provide you any benefits. The bandwidth is already more than the card can use. Hell, both my 4870 and 4870X2 sees no difference at all between 900 mem and 1000 mem, when my core is at 900. Bumping the core by another 150 isn't going to make 1000mhz GDDR5 a bottleneck, let alone 1100Mhz and above.

 

[Cool Mike]

Excellent non reference designed cards typically lead to better core/memory speeds. As you know, there are many over-clockers out here that love squeezing every bit out at the lowest temps. even if you only gain 2fps. many ofus are trying to determine if the 4890 toxic is equal to the Atomic. Core seems to be equal, memory is questionable at this point. Overclockers review showed a 1255 memory clock speed for the Atomic, i am not sure the Toxic can achieve this. MAYBE I CAN LEARN SOMETHING IN THE PROCESS

 

[intel igent]

Why is everybody so obsessed with the ram speed? Overclocking this GDDR5 isn't going to provide you any benefits. The bandwidth is already more than the card can use. Hell, both my 4870 and 4870X2 sees no difference at all between 900 mem and 1000 mem, when my core is at 900. Bumping the core by another 150 isn't going to make 1000mhz GDDR5 a bottleneck, let alone 1100Mhz and above.

good point Wile E! from what i recall back in my ol' day's of being a bench wh0re 3dm01/3dm03/3dm05 any OC on the Vram was neglegible, the real number's came when the core freq increased

oh and i should have a 4890 vapourX by middle of next week i hope

 

[Wile E]

Excellent non reference designed cards typically lead to better core/memory speeds. As you know, there are many over-clockers out here that love squeezing every bit out at the lowest temps. even if you only gain 2fps. many ofus are trying to determine if the 4890 toxic is equal to the Atomic. Core seems to be equal, memory is questionable at this point. Overclockers review showed a 1255 memory clock speed for the Atomic, i am not sure the Toxic can achieve this. MAYBE I CAN LEARN SOMETHING IN THE PROCESS

Yeah, but upping the memory doesn't even provide a 2fps benefit. It provides no benefit, whatsoever, aside from the production of extra stress and heat. Seriously, these GDDR5 cards has so much bandwidth, that clocking the memory does nothing at all, except increase heat and epeen.

 

[Blacklash]

That is true. Core clocking will yield the most results.

Anand showed a max performance increase of 13% going from an 850 core to 1000, and a max of 4% going from 975 to 1200 on the memory. Combined results below.

http://www.anandtech.com/video/showdoc.aspx?i=3555&p=11

 

[intel igent]

holy sh1t! i remembered something! this call's for celebration

 

[Cool Mike]

Blacklash, Thanks for the charts:


Quote:
So, to find the sweet spot for your overclock, you will want to increase core clock speed as much as you can. Then bump up memory clock and see how high you can get it and remain stable. Use a real world application to test performance at each point and then use a binary search like algorithm to find the sweet spot in a short number of tests. And there you have it. We didn't do this for you, but what's better practice than a little hands on experience right? Besides, if gives readers the opportunity to compare notes in the comments on what the optimal memory clock for a 1GHz core clock on the 4890 would be. Have fun!

Good discussion.

Thanks for the input on memory overclocking.